ZacharyZhang-NY/Kigi-CLI
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 2 Wiki Activity

Compare commits

base: ZacharyZhang-NY/Kigi-CLI:v0.1.9
ZacharyZhang-NY/Kigi-CLI:main
ZacharyZhang-NY/Kigi-CLI:v0.1.22 ZacharyZhang-NY/Kigi-CLI:v0.1.21 ZacharyZhang-NY/Kigi-CLI:v0.1.13 ZacharyZhang-NY/Kigi-CLI:v0.1.12 ZacharyZhang-NY/Kigi-CLI:v0.1.20 ZacharyZhang-NY/Kigi-CLI:v0.1.19 ZacharyZhang-NY/Kigi-CLI:v0.1.18 ZacharyZhang-NY/Kigi-CLI:v0.1.17 ZacharyZhang-NY/Kigi-CLI:v0.1.16 ZacharyZhang-NY/Kigi-CLI:v0.1.15 ZacharyZhang-NY/Kigi-CLI:v0.1.14 ZacharyZhang-NY/Kigi-CLI:v0.1.11 ZacharyZhang-NY/Kigi-CLI:v0.1.10 ZacharyZhang-NY/Kigi-CLI:v0.1.9 ZacharyZhang-NY/Kigi-CLI:v0.1.8 ZacharyZhang-NY/Kigi-CLI:v0.1.7 ZacharyZhang-NY/Kigi-CLI:v0.1.6 ZacharyZhang-NY/Kigi-CLI:v0.1.5 ZacharyZhang-NY/Kigi-CLI:v0.1.4 ZacharyZhang-NY/Kigi-CLI:v0.1.3 ZacharyZhang-NY/Kigi-CLI:v0.1.2 ZacharyZhang-NY/Kigi-CLI:v0.1.1 ZacharyZhang-NY/Kigi-CLI:v0.1.0
..
compare: ZacharyZhang-NY/Kigi-CLI:v0.1.15
ZacharyZhang-NY/Kigi-CLI:main
ZacharyZhang-NY/Kigi-CLI:v0.1.22 ZacharyZhang-NY/Kigi-CLI:v0.1.21 ZacharyZhang-NY/Kigi-CLI:v0.1.13 ZacharyZhang-NY/Kigi-CLI:v0.1.12 ZacharyZhang-NY/Kigi-CLI:v0.1.20 ZacharyZhang-NY/Kigi-CLI:v0.1.19 ZacharyZhang-NY/Kigi-CLI:v0.1.18 ZacharyZhang-NY/Kigi-CLI:v0.1.17 ZacharyZhang-NY/Kigi-CLI:v0.1.16 ZacharyZhang-NY/Kigi-CLI:v0.1.15 ZacharyZhang-NY/Kigi-CLI:v0.1.14 ZacharyZhang-NY/Kigi-CLI:v0.1.11 ZacharyZhang-NY/Kigi-CLI:v0.1.10 ZacharyZhang-NY/Kigi-CLI:v0.1.9 ZacharyZhang-NY/Kigi-CLI:v0.1.8 ZacharyZhang-NY/Kigi-CLI:v0.1.7 ZacharyZhang-NY/Kigi-CLI:v0.1.6 ZacharyZhang-NY/Kigi-CLI:v0.1.5 ZacharyZhang-NY/Kigi-CLI:v0.1.4 ZacharyZhang-NY/Kigi-CLI:v0.1.3 ZacharyZhang-NY/Kigi-CLI:v0.1.2 ZacharyZhang-NY/Kigi-CLI:v0.1.1 ZacharyZhang-NY/Kigi-CLI:v0.1.0
61 Commits
v0.1.9 .. v0.1.15
Author SHA1 Message Date
雷电芽衣 0fb971e9aa chore(legal): regenerate THIRD-PARTY-NOTICES for the 0.1.15 dependency set
Release / build (x86_64-unknown-linux-gnu) (push) Failing after 6s
Details
Release / build (aarch64-apple-darwin) (push) Canceled after 0s
Details
Release / build (aarch64-unknown-linux-gnu) (push) Canceled after 0s
Details
Release / build (x86_64-apple-darwin) (push) Canceled after 0s
Details
Release / build (x86_64-pc-windows-msvc) (push) Canceled after 0s
Details
Release / publish GitHub Release (push) Canceled after 0s
Details 
`cargo about generate about.hbs -o THIRD-PARTY-NOTICES.md` on the 0.1.15
lock: kigi crates at 0.1.15, chacha20 0.10.2.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 13:38:08 -04:00
雷电芽衣 a2ad08fb52 chore(release): bump workspace version to 0.1.15 
Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 13:37:16 -04:00
雷电芽衣 b9ca80e801 docs(agents): record the 0.1.15 upstream sync window and its known limitations 
Sync record now ends at upstream `3794978` (2026-09-09, ≈v1.0.24). Lists
the two limitations the sync left: the Windows frame writer still prints
through std's stderr lock, and the bwrap builder re-stats /data.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 13:36:58 -04:00
雷电芽衣 5f249ba6ca chore(deps): chacha20 0.10.2 replaces the yanked 0.10.1 
`cargo deny check advisories` failed on the yanked crate (via rand 0.10 →
sigstore-tsa → nono). `cargo update -p chacha20`; no other lock change.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 13:36:58 -04:00
雷电芽衣 d331b0b2a3 fix(goal-classifier): git diff headers keep a/ b/ prefixes under any host gitconfig 
`extract_changed_files` parses `diff --git a/<old> b/<new>` headers.
A host `diff.mnemonicprefix=true` or `diff.noprefix=true` rewrote them
(`c/ w/`, or none), so every tracked change vanished from `changed_files`
and the two evidence git-diff tests failed on such hosts. Both capture
paths now pass `--src-prefix=a/ --dst-prefix=b/`, which override both
config keys. Regression test sets each key in the temp repo.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 13:36:58 -04:00
雷电芽衣 09867915ef fix(sandbox): the devbox /data write-deny makes bubblewrap mandatory 
A devbox-based profile on a host with /data warned and fell back to
Landlock when bubblewrap was missing or failed to exec, and Landlock
grants the workspace under /data. requires_data_write_deny (devbox
inheritance and an existing /data; a stat error other than NotFound
counts as present) now joins requires_read_deny in the fail-closed
startup decision, and the /data ro-bind is added only under that
predicate, so a missing /data no longer makes bwrap fail to exec.

Ported from upstream (grok-build sandbox/lib.rs requires_data_write_deny).

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 13:17:52 -04:00
雷电芽衣 edce333620 fix(subagent): project-agent mcpServers follow folder trust; agent names are one path component 
A project-scoped agent definition from an untrusted repository could
declare inline stdio mcpServers and the child session started that
command, while inline hooks on the same definition were already gated on
folder trust. The mcpServers list now uses the same gate.

By-name agent lookup joined the requested name onto each searched
directory, so a name carrying a separator or ".." resolved a file outside
that directory under its trust scope; both lookup entry points now reject
any name that is not a single path component.

Ported from upstream (grok-build subagent/handle_request.rs mcpServers
gate), extended with the name guard.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 13:04:46 -04:00
雷电芽衣 d1c29a576c fix(compaction): failures carry the real error, with endpoint credentials redacted 
The pager rendered "Compaction failed." with the error dropped, and the
shell's unsuppressed auto-compaction failure sent an empty error. Both
now carry the decoded first line of the failure (retry guidance plus the
detail), decoded from the ACP data rather than its JSON-escaped Display,
propagated unchanged through the compact extension, capped for a banner.

Endpoint credentials never reach the banner: the sampler redacts the
parsed endpoint (userinfo, sensitive query values, fragment via the
repository sanitizer) in every 401 message and the appended request URL,
and the shell redacts every URL in the error text before line selection,
including URLs whose password holds whitespace or that follow a
multi-byte delimiter.

Ported from upstream 1.0.13 (grok-build dispatch/prompt.rs,
session/compaction.rs), extended with the redaction.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 12:12:00 -04:00
雷电芽衣 17d4ba08f9 fix(exit): diagnostics after the terminal may be gone are best-effort writes 
With fd 2 dead (a closed pane, an exited pipe reader) every write fails,
eprintln! panics on that failure, and under panic = "abort" the process
dies by SIGABRT with a crash report on the next launch instead of its
exit code. Exit reports, headless plain-format diagnostics, update-exit,
setup-result and leader-management messages now write through
best_effort_stderr::eprint_line: one attempt, failure reported, never a
panic.

Ported from upstream 1.0.19 (grok-build best_effort_stderr.rs).

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 10:39:20 -04:00
雷电芽衣 901203c2b2 fix(tui): pasted U+2028 and U+2029 become line breaks 
str::lines() splits on neither separator, so a paste carrying them counted
as one line and the invisible characters broke the textarea's width model.
Bare \r, U+2028 and U+2029 now normalize to \n on every paste route (the
prompt, the inline paste key and the previous-message editor); dropped
file paths keep their bytes verbatim because a separator in a filename is
payload.

Ported from upstream 1.0.18 (grok-build prompt_widget normalize_line_breaks).

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 10:16:22 -04:00
雷电芽衣 d8609267a1 fix(images): resize over-cap attachments even when the re-encode saves no bytes 
An attachment over the side or area caps whose re-encode did not shrink
the byte count was returned unchanged at its original dimensions, so the
provider still received an over-cap image. The byte-saving short-circuit
now applies only within the caps; a resize that would fall under the
vision minimums keeps the original on the existing fallback path.

Ported from upstream 1.0.13 (grok-build session/image_normalize.rs).

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 09:53:11 -04:00
雷电芽衣 a7b8438881 fix(tui): refresh the context bar when auto-compaction starts 
AutoCompactStarted carries the count the trigger fired on, but only
AutoCompactCompleted refreshed the context bar, so the "N% full" banner
sat next to a stale pre-turn number until the next stream update. Both
lifecycle updates now refresh the bar, on the root view and on subagent
views.

Ported from upstream 1.0.12 (grok-build acp_handler/session_notification.rs).

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 09:35:27 -04:00
雷电芽衣 584d0fb77e fix(tui): /compact <instructions> reach the compaction request 
The queued /compact command drained into Effect::Compact with only the
session id, so the instructions typed after the command never reached the
shell's compaction prompt (which already accepts userContext). The effect
carries the parsed instructions, trimmed so an edited queue row with
leading whitespace still counts.

Ported from upstream 1.0.22 (grok-build dispatch/queue.rs).

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 09:23:39 -04:00
雷电芽衣 4512df79e6 fix(tui): terminal escapes ride the writer queue and teardown is bounded 
A terminal that stopped draining the pty parked the writer thread inside
its tty write while it held the stderr lock; the next inline stderr write
from the event loop (mouse toggle, notification, kitty clear) deadlocked
the loop, and /quit hung on the unbounded writer join and the teardown
writes.

EscapeWriter queues out-of-band escapes on the frame writer: the mouse
capture toggle, notifications (title, progress, protocol), the auth
cursor, gboom keyboard flags, the wrap clipboard request, and the kitty
clears (queued per agent view and merged into the frame's post-flush,
nested subagent views included). The $EDITOR/$PAGER handoff aborts when
the writer does not drain. WriterThread::join_within and the teardown
run under a 2 s grace; after a stall fd 2 is silenced, the exit hints,
relaunch and update prints are skipped, and the panic hook keeps only its
log report.

Ported from upstream 1.0.19 (grok-build EscapeWriter, WriterJoin,
run_bounded_teardown), extended to the remaining inline writes.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 09:08:09 -04:00
雷电芽衣 91f14112bf fix(subagent): bound the completion-path awaits so teardown reaches Shutdown 
A completed child parked forever behind a parent that never serviced its
command channel, or behind its own frozen chat-state actor, leaking the
child session thread, fs watchers and fds. Every await in the completion
path that rides another actor is now bounded: child chat-state reads
degrade to their fallbacks after 10 s, the usage-fold and not-applied acks
give up after 30 s (an unacked mark falls back to the coordinator's sticky
mark), list_tasks and reparent_notifications are bounded, and the
post-Shutdown worktree snapshot is bounded while removal is awaited in
full so no detached deletion can race a resume.

Ported from upstream 1.0.14 (grok-build subagent/handle_request.rs
PARENT_ACK_TIMEOUT / child_actor_query), reshaped without the drain-budget
const assert.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 07:57:20 -04:00
雷电芽衣 fc7113ba3c fix(chat-state): reasoning estimate and calibration reseed on every rewrite 
A Reasoning item counted its text and its encrypted blob as two separate
payloads, inflating the estimate for retained reasoning; the estimate now
takes the larger of the two (base64 runs 4/3 over the raw bytes).

Only compaction carried the provider-confirmed ratio into the reseeded
total; a rewind or a non-compaction replace sprang the context meter back
to the raw estimate. reseed_total_tokens applies the ratio (capped at the
pre-rewrite total) to every rewrite, and the truncate path uses it too.

Ported from upstream 1.0.1x (grok-build xai-chat-state actor/state.rs,
mutations.rs, queries.rs).

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 07:24:07 -04:00
雷电芽衣 00a5e318bb fix(config): an explicit [model.X] api_backend survives same-slug propagation 
The slug-match step treated an entry whose api_backend equals the default
as unset and overwrote it with a same-slug sibling's backend, so
`api_backend = "chat_completions"` silently became messages/responses.
Config keys whose override sets api_backend are recorded and skipped by
the donor step; an unset backend still inherits.

Ported from upstream 1.0.22 (grok-build agent/config.rs
explicit_api_backend_keys).

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 07:13:47 -04:00
雷电芽衣 017a570d7d fix(sampler): a max_tokens stop after complete tool calls runs them 
The Messages stream failed every max_tokens stop with MaxTokensTruncation,
so a response that had already emitted complete tool calls before the
output budget ran out lost those calls and failed the turn. The stream
now completes with StopReason::ToolCalls when every collected call's
arguments parse (or are empty, the zero-argument convention); a cut call
or a text-only truncation still fails so compact-on-error keeps working.
The ChatCompletions and Responses streams already behaved this way.

Ported from upstream 1.0.1x LengthPolicy::CompleteToolCalls, reshaped
into the stream layer because Kigi has no caller that chooses a policy.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 07:02:05 -04:00
雷电芽衣 63d26d4a3e fix(persistence): FlushAndAck is a durability barrier that reports lost writes 
The persistence actor acked FlushAndAck unconditionally after writing its
pending merged notification: nothing was fsynced and every storage failure
was only warn-logged, so persist_ack, the reconnect flush and history
repair treated lost writes as persisted.

The barrier now writes the pending notification, reports the first storage
failure latched since the last barrier (every actor arm latches), and
fsyncs the whole session directory tree (files, directories, deletions).
Session directories are created durably (init and fork copy), and the
shared summary writer syncs its staged bytes before publishing and the
parent afterwards, so a concurrent /rename cannot publish unsynced bytes
over a synced summary.json. FlushComplete and the persist_ack, history
repair and end-of-turn callers propagate the io::Result; CopyFile fails on
a sync error but keeps the latch for the barrier.

Ported from upstream 1.0.1x (grok-build persistence.rs pending_write_error
+ durable storage helpers), reshaped without the dirty-file set.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 06:51:27 -04:00
雷电芽衣 5f7416f3ef fix(permissions): native path deny/ask rules apply to the file the tool opens 
Edit/Read/Grep rules matched only the raw argument string, so a workspace
symlink (or a quoted, `~`, `./`, `..`, or absolute spelling) reached a
denied file. The policy now resolves the path exactly as the tools do
(sanitize, expand `~`, display-cwd remap, join to the requesting session's
cwd), follows symlinks parent-first, and re-checks deny/ask rules on the
opened path and its target as absolute, `./`-relative and relative
spellings. Escalation only: no allow is granted through a link. A link
that cannot be followed forces a prompt that survives YOLO, auto mode and
the session edit grant, with event reason `unresolved_link`.

Permission requests now carry the requesting tool's cwd and display cwd
so a child worktree session sharing its parent's handle is checked where
it runs.

Ported from upstream 1.0.14 (grok-build permission/policy.rs symlink
re-check), reshaped for Kigi's lexical rule matcher.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 05:48:06 -04:00
雷电芽衣 fb3fe89d33 fix(permissions): rg exec flags cannot ride an allowlist, and argv is decoded 
Port of upstream grok-build 3794978 (snap6 "Reject ripgrep --hostname-bin
on always-safe and Auto routine paths"), widened where review showed the
same hole one layer down.

- `rg --pre` / `--hostname-bin` execute a program per file. The safe
  list refused only `--pre`, and the Auto-mode routine classifier had
  no rg flag check at all. One shared `rg_has_unsafe_flag` now refuses
  both flags (bare or `=VALUE`) on a lowercased-basename head, scanning
  every argument: rg's grammar (`--replace -- --pre`, `-ne --`) makes a
  partial arity model bypassable, so a literal search for the flag name
  costs a prompt instead of the auto-allow.
- The shell splitter handed allowlists undecoded source: a concatenation
  (`--host""name-bin`, `''--hostname-bin`, `--p're'`) kept its quotes,
  and double-quoted escapes (`"--host\<newline>name-bin"`) survived.
  `PlainCommand` words are now the decoded argv (bash double-quote
  rules, whole-node text minus the quotes so literal newlines and the
  whitespace tree-sitter folds into the closing quote survive), with a
  per-word `expandable` flag for unquoted glob/brace/tilde/backslash
  syntax the shell would still expand (`--{pre,pre}`, a glob that can
  match a file named `--pre`); the rg guard treats those as unsafe.
- Every prefix-based decision matches per argv element through one
  matcher (built-in safe list, always-safe list, dangerous list, Auto
  routine prefixes): a decoded word with inner whitespace (`rg    `,
  `rg /../payload`, one word spelled `git status`) can no longer pose
  as a prefix. Saved grants match token for word (a grant saved from a
  spaced argument prompts again). Saved denies keep the pre-0.1.15
  check verbatim on a `legacy_words` spelling (raw content and raw
  concatenations, wrappers peeled independently, space-boundary prefix)
  so no deny narrows on upgrade.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 04:47:21 -04:00
雷电芽衣 aa9a5dbe72 fix(auto-mode): git checkout/switch/stash no longer ride a blanket prefix 
Port of upstream grok-build 7581004 (1.0.22 "Auto mode no longer
instantly runs destructive git checkout -- commands"), routine_git.rs.

`ROUTINE_PREFIXES` auto-allowed `git checkout`, `git switch` and
`git stash` wholesale, so Auto mode ran `git checkout -- <file>` (a
working-tree discard) and `git stash drop`/`clear` without review.
Those verbs now go through a fail-closed operand classifier: `switch`
is routine with one ref operand and benign flags; `stash` only for
push/save/pop/apply/list/show/branch; `checkout` additionally needs a
branch-mode flag (`-b`, `-B`, `--orphan`, `--detach`, `-t`) anywhere in
the argument list before an operand counts, because a bare
`git checkout <name>` — `-` included — restores a tracked path when no
such ref resolves (a deliberate deviation from upstream, which allowed
`git checkout main` outright). Path-like operands, pathspec flags and
`--` fail closed. What is not routine goes to the model side-query, not
to a hard block.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 02:36:14 -04:00
雷电芽衣 7a60cd42ac fix(trust): gate directory-typed project hook sources 
Port of upstream grok-build 7581004 (1.0.22 window) folder_trust.rs.

`collect_repo_config_kinds` gated `.kigi/hooks` with `is_dir()` and
`.cursor/hooks.json` with `is_file()`, and `.claude/settings*.json` via
the settings-file scan, while the hook loader (`path_to_source`) takes
a directory at any of the four project hook paths as a hook source and
parses its JSON children. A clone carrying `.cursor/hooks.json/evil.json`
or `.claude/settings.json/evil.json` and nothing else resolved as
trusted through the no-config branch and ran its SessionStart hook.

The gate now treats any present path (file, directory, dangling
symlink) or an unreadable one at the four sources
`discover_hook_source_paths` loads from the git root as a hit.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 02:17:25 -04:00
雷电芽衣 97c934a647 fix(sandbox): strict profile no longer lets a child rewrite ~/.kigi 
Port of upstream grok-build 1.0.14 ("--sandbox strict now restricts
writes to ~/.grok/sessions only"), shaped for Kigi.

The strict profile granted the whole ~/.kigi read-write, so a sandboxed
command could plant a global SessionStart hook or flip
`[sandbox].profile` in config.toml for the next launch. Landlock cannot
carve a child out of a write grant, so the home is now read-only
except an explicit allowlist of state subtrees (run, sessions,
worktrees, worktree_pool, tasks, memory, logs, debug, crash, memtrace,
trace-exports); config.toml, auth.json, sandbox.toml, hooks/, bin/ and
vendor/ stay read-only. A symlinked state dir under the home is
followed only to the default ~/.kigi/sessions.

What the process itself writes there after enforcement moves or is
minted first:
- kigi-shell `apply_sandbox` mints the device id, extracts the bundled
  rg and the bundled skills BEFORE enforcement for every profile that
  restricts home writes (`restricts_home_writes`, custom profiles
  extending strict included).
- The sandbox audit log lives at ~/.kigi/sessions/sandbox-events.jsonl.
- Leader lock/socket default to ~/.kigi/run/, the leader stderr log to
  ~/.kigi/logs/leader.log; discovery scans run/ and the legacy root,
  and the default target prefers the current socket.
Root-level caches (version.json, agent_id, slash-mru.json, pager.toml,
claude_import_state.json) degrade best-effort under strict; documented.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 02:08:42 -04:00
雷电芽衣 cadc46f5dc fix(sandbox): close io_uring and sendmmsg holes in the child network filter 
Port of upstream grok-build 9684fa3 (1.0.11 window) `child_net.rs`.

The per-child seccomp filter matched seven socket syscalls by number and
allowed everything else, so a restricted child could still reach the
network through `sendmmsg` or an io_uring submission queue, and a
syscall issued through another arch table (x32 on x86_64) never matched
the deny list at all.

- Gate on the audit arch before matching `nr` and reject x32 numbers on
  x86_64 (foreign-table syscalls now get EPERM instead of passing).
- Add `sendmmsg`, `io_uring_setup`, `io_uring_enter`, `io_uring_register`.
- Build the program once (`prebuilt_child_network_filter`) and hand the
  slice to the `pre_exec` installer; the old code allocated the Vec after
  `fork`, which is not async-signal-safe. All three callers updated.
- Tests evaluate the BPF program over synthetic `seccomp_data`.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 01:19:36 -04:00
雷电芽衣 b957e12370 fix(tui): repaint the header context window on a /model switch 
After `/model`, the header kept the previous model's window ("15K /
1.0M" on a 262K model) until the next turn or a restart. The pager
caches the window inside its context snapshot (streaming `totalTokens`
updates refresh it from the catalog) and nothing refreshed the cache
when the session's model changed. Comparing cached numbers against
catalog windows cannot tell a catalog-derived total from a shell
effective limit (KIGI_DEBUG_CONTEXT_WINDOW, header upgrades), so the
shell, which installs the effective window, now reports it:

- kigi-shell: `handle_set_session_model` returns the installed window;
  `model_switch::apply` stamps it as `totalContextTokens` in the
  `session/set_model` response meta and carries it on the
  `ModelChanged` broadcast (`context_window`, serde-default: old pagers
  ignore it, old leaders omit it). The response meta previously
  serialized the actor's inner `Result` and reported success on an
  actor failure; that error now propagates.
- kigi-tui: `Effect::SwitchModel` reads the meta; `SwitchModelComplete`
  carries `Ok(Option<u64>)`; the completion handler and the follower
  `ModelChanged` arm call `AgentView::apply_context_window`, which
  re-derives the snapshot (pct, free tokens) and reports whether it
  changed so a follower redraws on a window-only change.
- Tests: the bare `/model <name>` dispatch path asserts the snapshot and
  the drawn header; follower with/without the field; same-model new
  window redraw; actor returns the window; serde round trip.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-13 00:15:14 -04:00
雷电芽衣 ff92a00fea chore(fmt): reflow three link constants rustfmt rejects 
f7a4cb6 repointed the GitHub links without running cargo fmt; the
fmt gate has been red since.

Claude-Session: https://claude.ai/code/session_01RUqV8HVGXcAETyYbZ2tPQ4
 2026-09-12 23:17:44 -04:00
ZacharyZhang-NY f7a4cb64ab Return the forge to GitHub: releases, self-update, links, CI
Release / build (x86_64-unknown-linux-gnu) (push) Failing after 6s
Details
Release / build (aarch64-apple-darwin) (push) Canceled after 0s
Details
Release / build (x86_64-apple-darwin) (push) Canceled after 0s
Details
Release / build (x86_64-pc-windows-msvc) (push) Canceled after 0s
Details
Release / publish GitHub Release (push) Canceled after 0s
Details
Release / build (aarch64-unknown-linux-gnu) (push) Canceled after 0s
Details 
The GitHub account is restored, so github.com/ZacharyZhang-NY/Kigi-CLI is
the canonical repo again. Reverse the 2026-08 Gitea migration:

- .github/workflows/{release,warm-cache}.yml: restore the GitHub-hosted
  pipeline dropped in 3e1652c, unchanged — hosted runners, rust-cache,
  softprops release with every asset attached atomically.
- .gitea/workflows/release.yml: retire the Gitea edition (the runners are
  being deregistered; git.zacharyzhang.com keeps the v0.1.12/v0.1.13
  releases for old clients).
- kigi-env: update_base_url now defaults to the GitHub Releases API. The
  updater already speaks GitHub wire format and sends a User-Agent; v0.1.14
  is published to BOTH forges with identical assets so pre-0.1.14 installs
  polling Gitea self-update onto the GitHub channel.
- install.sh / install.ps1: resolve releases from api.github.com.
- feedback, /docs, SECURITY.md, frontend worker + site: repo, issues and
  raw-file links point at github.com / raw.githubusercontent.com.
- docs/RELEASE.md: GitHub edition of the checklist; Gitea runner sections
  retired.
- version: 0.1.13 -> 0.1.14 (the cross-forge transition release).

Claude-Session: https://claude.ai/code/session_01A6RcwHhhsFYYrNdQB4ASuP
 2026-09-01 01:05:22 -04:00
ZacharyZhang-NYandClaude Fable 5 39cb1ca1bc chore(release): bump workspace version to 0.1.13
Release / prepare draft release (push) Successful in 6s
Details
Release / build (aarch64-apple-darwin) (push) Failing after 1s
Details
Release / build (x86_64-apple-darwin) (push) Failing after 1s
Details
Release / build (x86_64-pc-windows-msvc) (push) Canceled after 0s
Details
Release / publish release (push) Canceled after 0s
Details
Release / build (x86_64-unknown-linux-gnu) (push) Successful in 16m32s
Details
Release / build (aarch64-unknown-linux-gnu) (push) Successful in 16m13s
Details 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 12:43:58 -04:00
ZacharyZhang-NYandClaude Fable 5 e83610db6a chore(legal): regenerate THIRD-PARTY-NOTICES for the 0.1.13 dependency set 
cargo-about over the updated lockfile (h2 0.4.19, webbrowser 1.2.4).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 12:09:34 -04:00
ZacharyZhang-NYandClaude Fable 5 fd5cff2de6 chore(gates): finish the Windows sweep — zero clippy findings workspace-wide 
Second layer of platform gaps the mac/linux-only gate history left:
unix-only test imports/helpers now cfg-gated (mermaid, fast-worktree,
hunk-tracker, pager-render link test, three pty_e2e module decls),
clippy-suggested cleanups in Windows-cfg code (file-utils classifier,
tui console-mode restore, shared clipboard arboard-error flow, tools
terminal closure, workspace capability tail), and `cargo update` of h2
(RUSTSEC-2026-0258) and webbrowser (RUSTSEC-2026-0257) so `cargo deny
check advisories` passes again.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 12:04:32 -04:00
ZacharyZhang-NYandClaude Fable 5 427e8b6fea docs(agents): record the upstream sync baseline and the subagent spawn bound 
The 0.1.13 sync had to reconstruct the fork point by content
fingerprinting; the sync record (baseline snapshot, last-evaluated
upstream commit, and where upstream hides its per-change lists) now
lives in AGENTS.md so the next sync starts from facts.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 11:55:19 -04:00
ZacharyZhang-NYandClaude Fable 5 50669bf302 chore(gates): zero clippy warnings and a buildable sandbox example on Windows 
The full-workspace gates had never run on this platform: collapsible
ifs in the Windows shell probe, a same-type cast in the crash handler,
an unused AsRawHandle import, and the sandbox smoke-test example
building its unix-only body (libc, kernel enforcement) unconditionally.
The example now stubs main off-unix; formatting reflow from cargo fmt.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 11:55:19 -04:00
ZacharyZhang-NYandClaude Fable 5 25d1a4f980 fix(sampler): retry Cloudflare 52x and other 5xx during outages 
Port of upstream grok-build 1.0.0/1.0.1-window retry policy.

- is_retryable_api_status: 429 plus any 5xx except Cloudflare 525/526
  (origin TLS handshake / invalid certificate — never clears on its
  own), replacing the 500|502|503|504|520 allowlist so the 521-524,
  529, and 530 edge pages an outage serves actually retry.
- The generic retry path clamps a server Retry-After to
  MAX_RETRY_BACKOFF (30s) and jitters it: Cloudflare 52x pages ask for
  60-120s, which across the budget stalled a turn ~28 minutes. The 429
  path keeps the full (parse-capped) Retry-After under its low attempt
  cap, and jitter is shared via jittered().

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 11:51:36 -04:00
ZacharyZhang-NYandClaude Fable 5 a441b23a6d fix(mcp): kigi mcp add <name> <url> infers an HTTP server 
Port of upstream grok-build 1.0.9. A bare positional http(s):// URL
with no transport flag was silently added as a stdio command whose
"command" was the URL. It now infers the HTTP transport with a notice;
extra args, --env, legacy --command/--url, and every explicit flag keep
their prior meaning, and scheme-less URL-looking commands still stay
stdio with the existing warning.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 11:42:47 -04:00
ZacharyZhang-NYandClaude Fable 5 574122884a fix(windows): resolve the home directory via the standard lookup 
Port of upstream grok-build 1.0.4/1.0.5. Two paths read $HOME only,
which is usually unset on Windows outside Git Bash:

- the global-gitignore fallback (prompt/ignore.rs) now uses
  std::env::home_dir(), the crate-wide idiom;
- fast-worktree's resolve_kigi_home falls back to std::env::home_dir()
  with the same dunce canonicalization, eliminating the documented
  divergence from kigi_config::default_kigi_home. On Unix home_dir()
  reads $HOME first, so behavior is unchanged where HOME is set.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 11:34:45 -04:00
ZacharyZhang-NYandClaude Fable 5 97de661554 fix(startup): thread-exhaustion resilience — cap workers, never panic 
Port of upstream grok-build 0.2.113/0.2.114 fixes, extended to Kigi's
startup order.

- kigi-tty-utils runtime module: worker-cap policy (8 max) — tokio's
  per-core default pins 100+ thread slots on many-core shared hosts
  against RLIMIT_NPROC / pids.max, and later spawns die with EAGAIN.
- kigi-bin builds the main runtime with capped workers and reports an
  out-of-threads host with a diagnostic + exit(1) instead of a panic.
- The pre-runtime optimization threads (HTTP warmup, early model
  prefetch) spawn recoverably and degrade — their panic would have
  preempted the runtime's own diagnostic.
- The per-session actor thread and its runtime report spawn/build
  failure through the existing acp::Error / AgentBuildError boundaries
  (new RuntimeBuild variant) instead of panicking the agent host.
- fsnotify's watcher thread spawn maps into FsNotifyError::WatcherStart
  so callers degrade instead of crashing.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 11:28:17 -04:00
ZacharyZhang-NYandClaude Fable 5 bf93091416 fix(agents): bound how many subagents one session runs at once 
Port of upstream grok-build 1.0.1 (landed upstream 2026-08-06). The
coordinator fire-and-forgot every spawn, so a wide fan-out (graph
waves, swarm bursts, model-driven Task storms) could exhaust file
descriptors.

- kigi-tools task/admission: SubagentLimits — 32 concurrent per session
  by default (KIGI_MAX_CONCURRENT_SUBAGENTS), queue|fail behavior
  (KIGI_SUBAGENT_LIMIT_BEHAVIOR), zero clamped to 1, read once at the
  composition root; parse_positive_env helper.
- Each parent session gets one spawn semaphore (map entries live for
  the process so a resumed session shares its predecessor's limit);
  handle_subagent_request holds an owned permit for the whole child
  run — queue awaits FIFO with a unified-log notice at saturation,
  fail routes through send_pre_spawn_failure. After a queued wait, a
  closed parent command channel (teardown) fails the spawn instead of
  running against a dead parent.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 11:09:24 -04:00
ZacharyZhang-NYandClaude Fable 5 26d26e15c9 fix(exec): bounded post-kill reaps — D-state children can't wedge a turn 
Port of upstream grok-build 1.0.1 (landed upstream 2026-08-06).

A child wedged in an uninterruptible kernel syscall (D-state) never
observes SIGKILL, and a background descendant can hold the output pipes
open past the kill; both previously stalled tool futures without bound.

- kigi-tty-utils: KILL_REAP_TIMEOUT (5s) + reap_killed_bounded — wait
  for an already-killed child at most the bound, then abandon the corpse
  to tokio's orphan reaper.
- local_terminal: pipe readers share their buffer with the spawner and
  join under the reap bound on both exit arms (aborted readers are
  awaited so no tail bytes race the take); the whole tree is enrolled in
  a ProcessGroup and group-killed on timeout; the timeout arm returns
  the synthetic result instead of waiting on the corpse. Grandchild
  kill + prompt-return pinned by a ported unix regression test.
- kigi-tools: reap_killed_search_child wrapper; grep's four kill sites
  reap bounded; glob's stderr goes to null (an unread pipe blocks rg on
  warning backpressure) and its truncation kill reaps bounded.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 10:45:17 -04:00
ZacharyZhang-NYandClaude Fable 5 1ae803519e fix(compaction): configurable summarizer tool_choice; wider overflow detection 
Port of upstream grok-build 0.2.105 + 0.2.118.

- CompactionToolChoice (auto default | none), resolved env
  KIGI_COMPACTION_TOOL_CHOICE > [features].compaction_tool_choice >
  remote settings. Some OpenAI-compatible servers reject
  tool_choice:"none" with tools attached, failing every long-session
  compaction; auto keeps them working (the summarization prompt already
  forbids tool use). Threaded from agent + subagent spawn through
  CompactionConfig into generate_session_compact; the Responses gate
  counts hosted tools too, so a hosted-only request still carries the
  configured choice.
- is_context_length_error recognizes the "current message ... exceeds
  budget" phrasing, so that overflow is Fatal instead of burning the
  retry budget on a deterministic failure.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 10:22:18 -04:00
ZacharyZhang-NYandClaude Fable 5 6a4bff0d45 fix(agent): a racing cancel can no longer wedge the session's turn slot 
Port of upstream grok-build 0.2.103. session/prompt runs an async intake
preamble before enqueueing SessionCommand::Prompt on the actor mailbox; a
concurrent cancel could enqueue Cancel first, land on an idle actor as a
no-op, and the prompt then started a turn nothing would ever cancel.

The per-session prompt-intake lock becomes the dispatch lock: prompt()
holds it across intake as before, and cancel() now holds it around its
Cancel send, so a cancel orders behind the prompt it targets. Regression
test ported verbatim (unix-only harness, like its siblings).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 09:53:09 -04:00
ZacharyZhang-NYandClaude Fable 5 6e0464fa74 fix(auth): unique tmp per auth.json write; tighten loose credential perms 
Port of upstream grok-build 0.2.113-window fixes:
- write_auth_json_atomic tmp path gains a monotonic sequence next to the
  pid, so two concurrent in-process writers (background mint + proactive
  refresher) cannot share one tmp path and tear each other's bytes; a tmp
  write that fails partway is reclaimed instead of orphaned.
- New secure_file::ensure_owner_only_permissions (0o600 / Windows ACL)
  tightens pre-existing world-readable credential files on read, backup,
  write, replace, and restore. Best-effort with warns after durable
  content — a chmod-only failure must not masquerade as a failed write —
  and propagating in restore_prior_bytes.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf
 2026-08-27 09:31:21 -04:00
ZacharyZhang-NYandClaude Fable 5 257caed650 Add kigicli.dev frontend worker with built site assets 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012pFJn6ycA4hPRwM9sqasi6
 2026-08-25 23:04:06 -04:00
ZacharyZhang-NYandClaude Opus 5 3e1652ca98 ci: drop the inert GitHub workflows 
Gitea ignores `.github/workflows/` entirely once `.gitea/workflows/` is
non-empty, so these two files have had no effect since the release workflow
landed — and the GitHub repo they targeted is gone (404, suspended account),
so they cannot run there either.

They were not merely dead: warm-cache.yml's Monday cron kept firing on the
Gitea side before `.gitea/` existed, queueing jobs against GitHub-hosted
labels (`macos-14`, `windows-2022`, `ubuntu-24.04-arm`) that no self-hosted
runner advertises. Those sat queued forever, and its one dispatchable leg
failed on `rustup: command not found` — the same bare-image gap the Gitea
workflow now provisions around.

`.gitea/workflows/release.yml` is the only pipeline; docs/RELEASE.md already
documents it as such.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-08-10 22:11:28 -04:00
ZacharyZhang-NYandClaude Opus 5 479fb8cbbd chore(release): bump workspace version to 0.1.12
Release / prepare draft release (push) Successful in 5s
Details
Release / build (aarch64-apple-darwin) (push) Failing after 1s
Details
Release / build (x86_64-apple-darwin) (push) Failing after 2s
Details
Release / build (x86_64-pc-windows-msvc) (push) Canceled after 0s
Details
Release / publish release (push) Canceled after 0s
Details
Release / build (aarch64-unknown-linux-gnu) (push) Successful in 18m14s
Details
Release / build (x86_64-unknown-linux-gnu) (push) Successful in 17m20s
Details 
Every crate inherits `workspace.package.version`, so the single edit plus
the refreshed lockfile carries the release. The `v0.1.12` tag is what
triggers the build; release.yml gates the tag against this value, and
`cargo metadata --locked` passes so the build's `--locked` will too.

This is the first release cut on Gitea Actions. v0.1.6 through v0.1.9 were
all cancelled for want of a runner whose label matched anything, so nothing
has ever published from the new forge — `releases/latest` still 404s, which
is exactly why `kigi update` finds nothing today. Ships the link migration
in the previous commit, so the installers and self-updater published here
point at Gitea instead of a GitHub repo that no longer serves them.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-08-10 21:13:19 -04:00
ZacharyZhang-NYandClaude Opus 5 9fb82291b2 chore(migration): repoint every published link off GitHub 
The canonical repo is git.zacharyzhang.com now, so every URL a user can
actually reach had to move with it.

Self-updater: `PRODUCTION_ENDPOINTS.update_base_url` targets Gitea's
`/api/v1/repos/.../releases`, which serves the same wire format GitHub's
releases API did — `Release`/asset deserialisation is unchanged. The alpha
channel requests `?per_page=30&limit=30` because GitHub reads the page size
from `per_page` and Gitea from `limit`, and each ignores the other's, so one
URL stays correct on either.

Installers: same API base, and the documented one-liners become
`https://kigicli.dev/install.{sh,ps1}`. That domain's Worker proxies both
files from Gitea raw `main`, so the published install command survives any
future forge move without being reprinted everywhere. install.ps1 also drops
the GitHub-specific `Accept: application/vnd.github+json`.

Everything else user-facing: `/feedback` and the shell's feedback fallback
point at the Gitea issue tracker, `/docs` at the Gitea-rendered README, and
web_fetch's User-Agent at kigicli.dev. SECURITY.md now routes reports to the
issue tracker with a warning to withhold exploitable detail, because Gitea
has no equivalent of GitHub's private vulnerability reporting. Comments
naming `.github/workflows/release.yml` name `.gitea/` now.

Tests assert against these constants rather than literal URLs, so they
follow automatically; the one URL-shape assertion (production endpoints are
https) still holds.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-08-10 21:13:19 -04:00
ZacharyZhang-NYandClaude Opus 5 8bdfb551b6 chore(git): ignore prompt.txt and the kigicli.dev worker artifacts 
`/prompt.txt` is the one-shot Mac Mini runner-setup prompt. It carries a
Gitea runner registration token once filled in, and a leaked registration
token lets anyone attach a runner to the account and receive our jobs —
including release jobs. Anchored with a leading slash so a legitimately
tracked prompt.txt deeper in the tree would still be versioned.

The .wrangler/ and node_modules/ lines were already sitting uncommitted in
the working tree for the untracked frontend/ worker; they belong in this
same file, so they ride along rather than being split out.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-08-10 20:24:30 -04:00
ZacharyZhang-NYandClaude Opus 5 8864b9ecc5 ci(gitea): add the release workflow and point it at our real runners 
Gitea has no hosted runners, so every `runs-on` has to name a machine we
registered ourselves — and the labels in the first draft of this file did
not match what our machines advertise. The only online Linux runner
(`dokploy-runner-01`) offers `ubuntu-latest`/`ubuntu-24.04`/`ubuntu-22.04`,
never `linux-x86_64`, and Gitea's API cannot relabel a runner remotely, so
a `v*` tag would have queued forever against a label nothing served.

The three Linux jobs therefore run on `ubuntu-latest`. That runner is in
container mode against a bare runner-image, so those legs now provision
what they need per run: build-essential, pkg-config, unzip (protoc), tar
and jq, plus rustup when absent — installed with `--default-toolchain none`
so `rust-toolchain.toml` still decides the channel and `rustup show`
materialises it. `prepare` installs jq/curl the same way. macOS stays on
`macos-arm64` and Windows on `windows-x86_64`, both host mode with their
toolchains preinstalled, so the new provisioning step is gated on a
`linux:` matrix key rather than on `ext == 'tar.gz'`, which covers Darwin
too.

docs/RELEASE.md picks up the corrected label table, the fact that
`register --labels` is ignored once config.yaml supplies them, the
registration-token call being a POST (a GET parses as runner id 0 and
404s), and how each host-mode runner is supervised so it survives reboots
and crashes.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-08-10 20:21:22 -04:00
ZacharyZhang-NY 553ec1c620 chore(release): bump workspace version to 0.1.11
Release / build (x86_64-unknown-linux-gnu) (push) Failing after 5s
Details
Release / build (aarch64-apple-darwin) (push) Canceled after 0s
Details
Release / build (aarch64-unknown-linux-gnu) (push) Canceled after 0s
Details
Release / build (x86_64-apple-darwin) (push) Canceled after 0s
Details
Release / build (x86_64-pc-windows-msvc) (push) Canceled after 0s
Details
Release / publish GitHub Release (push) Canceled after 0s
Details 
All 62 crates inherit `workspace.package.version`, so the single edit
plus the refreshed lockfile carries the release. The `v0.1.11` tag is
what triggers the build; release.yml gates the tag against this value.

Ships the sessionless-model-switch fix (f656627): `/model` from a
non-project launch directory now starts the session its deferred switch
drains into instead of dangling silently.
 2026-07-27 16:04:51 -04:00
ZacharyZhang-NY f656627c10 fix(tui): start the session a sessionless model switch defers into 
Launching kigi in a non-project directory (~/Downloads, ~, /tmp — anywhere
`is_project_dir` rejects) leaves the agent view session-less behind the
project-picker question, which only a PLAIN prompt can open. Slash
commands still execute there, so `/model <name> <effort>` stashed its
switch in `deferred_model_switch` — a stash that assumes a create is in
flight — and with none pending it dangled forever with zero feedback:
the picker rendered, the user chose model and effort, and nothing
changed. Not a 0.1.10 regression: 0.1.9 pty-reproduces identically; the
report correlated with the upgrade only because every earlier launch
happened to be from a project directory.

Apply the QueueCommand precedent ("queued slash work bypasses the
picker, so create the deferred session or it never drains") to both
deferral sites — Action::SwitchModel in the router and
set_default_model's no-session branch — via
skip_picker_and_create_session, whose in-flight guard already makes it a
no-op while a create is pending, so the racing-create case is unchanged.
SessionCreated then applies the stash through the existing
apply_deferred_model_switch path; pty-verified end-to-end from
~/Downloads (session created, model changed, effort applied).
 2026-07-27 15:55:57 -04:00
ZacharyZhang-NY 11c3ca9803 chore(release): bump workspace version to 0.1.10
Release / build (x86_64-unknown-linux-gnu) (push) Failing after 6s
Details
Release / build (aarch64-unknown-linux-gnu) (push) Canceled after 0s
Details
Release / build (x86_64-pc-windows-msvc) (push) Canceled after 0s
Details
Release / publish GitHub Release (push) Canceled after 0s
Details
Release / build (aarch64-apple-darwin) (push) Canceled after 0s
Details
Release / build (x86_64-apple-darwin) (push) Canceled after 0s
Details 
All 62 crates inherit `workspace.package.version`, so the single edit
plus the refreshed lockfile carries the release. The `v0.1.10` tag is
what triggers the build; release.yml gates the tag against this value.

Also records the shared-SSRF-policy invariant in AGENTS.md, so a future
port does not reintroduce a second copy. The zero-egress entry is
compressed by the same number of lines to keep the file at 500.
 2026-07-27 13:26:25 -04:00
ZacharyZhang-NY bac8470c80 fix(hooks): share one SSRF policy instead of a second, stale copy 
The hook runner carried its own `is_blocked_ip`, a line-for-line copy of
the pre-hardening web_fetch predicate: loopback allowed unconditionally,
and no TEST-NET, 198.18/15, 240/4, 0.0.0.0/8, multicast, IPv6 site-local
or embedded-v4 wrapper coverage. Hook URLs come from settings, and
project settings from an untrusted repo are loaded today, so the gap is
reachable.

`kigi-hooks` already depends on `kigi-tools`, so the fix is to delete the
copy and call the shared predicate — no new crate, and no third
implementation to drift.

`allow_local` is on for hooks: a loopback hook receiver is a legitimate
local setup, which the old copy also allowed. It is now allowed only
when the URL names the host literally, so a public name resolving to
loopback is refused.

Range coverage now lives with the policy; the runner's tests pin what it
adds on top. The URL-scrubbing regression test moves off TEST-NET-1,
which the policy now blocks, onto a closed loopback port — faster, and
no real egress from a test.
 2026-07-27 13:01:50 -04:00
ZacharyZhang-NY 74402f3078 fix(pager): stop a sibling test flattening the diff-band assertion 
`committed_edit_keeps_diff_line_backgrounds` passed alone and
single-threaded but failed under parallel execution, so the workspace
gate was red.

`terminal_native_lock_paints_only_native_colors` engages the
process-global terminal-native lock. While it is held `Theme::current()`
returns `terminal_default()`, whose `diff_*_bg` are all `Color::Reset`,
so no band is painted at all — and `current_kind()` reports a nominal
`KigiNight`, which is what made the failure read as a theme mismatch.

Two changes, both at the cause:
- hold the shared theme lock via `theme_cache::pin_theme()`, the helper
  written for this and until now unused, so the lock cannot be engaged
  mid-test
- assert the invariant (two distinct non-Reset bands) instead of exact
  RGB: the entry line cache is keyed on the GLOBAL theme kind, so an
  exact-color assertion races by construction. Span-level colors stay
  pinned by the `tool::edit` tests that own them

The failure message now reports the theme, height, and the bands it
actually found, which is how the cause was located.
 2026-07-27 12:38:14 -04:00
ZacharyZhang-NY 649c5f5641 fix(models): drop the ultra reasoning effort, which no backend accepts 
The hardcoded Codex catalog advertised `ultra` for gpt-5.6-sol and
gpt-5.6-terra, so selecting it sent `reasoning.effort = "ultra"` and the
Responses endpoint answered 400 with its real menu, which ends at `max`.

The tier was kigi's invention: the enum documented it as codex-only, the
catalog was its only producer, and the upstream model snapshot never
listed it. Removing the variant as well as the catalog entries means
`/effort ultra` is rejected at the TUI instead of parsing and failing on
the wire; a persisted `ultra` degrades to the model default through
`lenient_reasoning_effort_opt`, as designed for vocabulary changes.
 2026-07-27 12:12:14 -04:00
ZacharyZhang-NY ad3840f9ec fix(web_fetch): block non-public targets by default and gate every hop 
kigi allowed loopback unconditionally and missed several non-public
ranges, and the SSRF check ran only on the initial URL.

Policy (ssrf.rs):
- loopback is blocked unless `[toolset.web_fetch] allow_local` (or
  KIGI_WEB_FETCH_ALLOW_LOCAL) is on, AND the URL names it explicitly,
  so a public name resolving to loopback stays blocked (DNS rebinding)
- add 0.0.0.0/8, 100.64/10, 192.0.0.0/24, TEST-NET-1/2/3, 198.18/15,
  240/4, IPv6 site-local and documentation prefixes
- inherit the IPv4 verdict through mapped, compatible, NAT64 and 6to4
  wrappers; network-specific NAT64 prefixes remain uncovered (see doc)

Plumbing (client.rs), where the exploitable half lived:
- re-check every redirect hop, not just the first
- compare hosts exactly; a `www` sibling has its own A records, so it
  is a cross-host redirect rather than an auto-followed hop
- run the check before the fetch service, so a blocked URL is never
  posted to an endpoint that egresses elsewhere
- exempt explicit local hosts from the https upgrade and from the
  single-label filter, and re-upgrade each followed hop

Wiring: allow_local reaches WebFetchParams from both construction
paths; documented in the config guide and the README env table.
 2026-07-27 11:44:20 -04:00
ZacharyZhang-NY ac23ebc9a1 feat(turn): halt a turn stuck repeating one identical tool call 
Ports grok-build's action stationarity breaker. A polling loop looks like
progress from inside the turn — every step succeeds — so nothing stops it
and it burns the whole budget re-reading one file. Nudge at 8 identical
batches, halt at 16, halt a pure no-op run at 4.

Three defects were found and fixed before this landed, none of which the
gates caught:

- The check runs AFTER execute_tool_calls, not before. Before it, the
  assistant's tool_calls are recorded with no results, so pushing the
  nudge triggers repair_dangling_tool_calls: synthetic "cancelled by the
  user" results get spliced in, the real results land after a user
  message, and dedup misses them — the orphan tool_result shape that
  draws a provider 400. It is Ok-gated too, since one `?` inside can
  leave a call resultless.
- `TurnOutcome::StationarityHalted` is its own variant grouped with
  Completed, not a reuse of Cancelled. As a cancellation it would report
  StopReason::Cancelled, fire the abort lifecycle, kill the turn's
  subagents, grow the goal harness back-off streak until three halts
  paused a running goal, and let completion-requirement recovery re-run
  the very turn halted for looping.
- The no-op gate is size-based, not name-based. kigi registers the shell
  tool as `run_terminal_cmd` and renames it `run_terminal_command`, so
  matching on "bash" silently disabled the tighter ceiling while every
  test still passed.
 2026-07-27 10:37:41 -04:00
ZacharyZhang-NY a07d889ad9 fix(swarm): label each member with its own item 
Every member was launched with the swarm's description, so the pager drew
N identical subagent blocks: the user could not tell which member was
running, and a failure could not be attributed to an item.

This is also why K6 does not port upstream's progress grid. A member IS
an ordinary subagent, so the pager already renders one block per member
with live status, duration, tool-call and turn counts — richer than a
grid cell. The gap was the label, not the widget; upstream's grid (and
the estimator behind it, which infers progress from tool-call rate)
answers a different TUI's shortcomings. The parent tool call already
reads "<description> (N members)" and `is_task_variant` already knows
`AgentSwarm`, so the blocking-wait spinner registers.
 2026-07-27 03:01:42 -04:00
ZacharyZhang-NY 16328e55f9 feat(swarm): /swarm mode — a standing instruction to fan work out 
The `agent_swarm` tool works with the mode off; what the mode adds is the
doctrine — decompose finely, give every member a disjoint scope, do not
do the work yourself. `/swarm`, `/swarm on|off`, `/swarm <task>`, gated
on the tool actually being in the toolset.

Two triggers, not upstream's three. Upstream's `tool` trigger exists so
invoking the tool makes the doctrine appear; kigi's tool carries its own
description and works with the mode off, so that trigger would add a
state only the tool can reach. `Manual` persists until switched off,
`Task` expires at turn end — both are real user intents.

Load-bearing decisions, each from a defect found in review:

- Expiry is a DROP guard, not a post-loop call. A user interrupt aborts
  the turn future rather than cancelling it, so post-loop code never
  runs; three `?` paths skip it too. Either way a `/swarm <task>` mode
  would leak into the next, unrelated prompt.
- `enter` is a total no-op while armed, so the `/swarm <task>` shorthand
  cannot downgrade a standing `/swarm on` into a per-turn mode that then
  disarms itself. This matches upstream; the first cut diverged, and the
  test asserting the divergence was inverted with the fix.
- An explicit `/swarm off` retracts UNCONDITIONALLY. The doctrine rides
  the conversation and so survives a resume, a fork and a compaction that
  an in-memory flag does not; trusting the flag left a session reading
  "off" with the instruction still steering and no way to clear it. That
  also retires the `reminder_live` field, whose doc comment claimed to
  know something the code cannot.
- The mode is deliberately NOT persisted, unlike /goal and /graph: those
  strand real work when lost, this is a prompt hint whose recovery is
  retyping one command. Recorded in AGENTS.md so the omission reads as a
  decision, not a gap.
- The pre-session gate is `subagents_enabled`, not a hardcoded `true`:
  the builder strips `agent_swarm` wherever it strips `task`, and
  advertising it then offers a menu entry that resolves to literal text.
- The exit reminder is as emphatic as the doctrine it revokes; a single
  weak clause is the likelier of the two to be summarised away.

The doctrine rides the existing `push_system_reminder` channel rather
than a second injection path of its own.
 2026-07-27 02:39:55 -04:00
ZacharyZhang-NY 9edb8729ef feat(swarm): agent_swarm — one prompt over many items, paced as a fleet 
Ports kimi-code's AgentSwarm: a `prompt_template` containing `{{item}}`
expanded over an `items` list into up to 128 subagents, run to completion
and returned as one aggregate. Kigi already exceeds upstream on planning,
verification, isolation and merge via /graph; what it lacked was cheap
immediate fan-out. Entirely client-side — no new backend surface.

The engine is three pure pieces plus a runner: `plan` validates and
expands (every fault reported before a single member starts — a
half-launched swarm is expensive to unwind), `schedule` is the launch
ramp as testable arithmetic, `run` drives it against the existing
`SubagentBackend`. Reuses the single-spawn coordinator rather than
inventing a batch API.

Load-bearing decisions, each the result of a defect found in review:

- `backgrounded` is its own outcome. A member that outlives the 600s
  foreground budget is detached by the coordinator and KEEPS RUNNING;
  reporting it as failed invites the model to relaunch its item, putting
  a second agent on the same files. It is never offered for resume.
- `InFlightGuard` cancels live members on Drop. Send-now cancels the turn
  WITHOUT cancelling subagents and aborts the task; the dropped receivers
  read as "parent gone" and each child re-attaches itself. There is no
  cooperative path to use instead — `Cancellation` is constructed nowhere
  in the tree — so Drop is the only seam that fires.
- Retries and wall clock are both bounded. The swarm blocks the caller's
  turn, so every wait needs a ceiling it cannot argue past; stragglers at
  the deadline are reported as still-running, with their ids.
- `ToolKind::AgentSwarm` is its own variant: `TemplateRenderer`'s
  `by_kind` map holds one tool name per kind, so sharing `Task` would
  silently redirect `${{ tools.by_kind.task }}` in other tools' prompts.
- An explicitly requested model that cannot be validated is refused, as
  the task tool already does — one loud error beats `items.len()` quiet
  ones. Depth stays capped at 1: upstream's unlimited nesting is a
  hazard, not a feature.
- `SubagentResult.rate_limited` is classified where the typed ACP error
  code is still in hand; a scheduler re-deriving it from a formatted
  string would stop adapting the day the wording changed.
- Aggregate output is clamped per member (head+tail, loss stated):
  native tool output is truncated nowhere downstream.

42 agent_swarm tests. The fake backend awaits, so the concurrency and
ordering assertions can actually fail; the cap test also proves the
fixture can exceed the cap.
 2026-07-27 01:22:52 -04:00
ZacharyZhang-NY ed8049cf77 fix(memory): scope embedding credentials to the endpoint that may receive them 
`MemoryBackendParams` carried the primary session `AuthManager` and the
session api-key provider unconditionally, while `embed_base_url` is the
CURRENT MODEL's endpoint and `AuthRetryMiddleware` stamps `Authorization`
on every request it wraps. A user who enables `[memory.embedding] model`
while running a BYOK or subscription-OAuth model therefore sent the Kimi
session bearer to that third party.

`EndpointScopedCredentials` binds the credential to the one endpoint it
may reach: `for_endpoint` drops the handle unless the caller vouches for
the URL, and `approved_for` re-checks at provider-build time in release
too, because `MemoryBackendParams` is `Clone` and callers rewrite fields
on the copy.

The shell decides through `CredentialAuthority::manager_for` rather than
a second URL predicate — it answers both whether a credential may ride
and which manager governs it, so a subscription-OAuth platform gets its
own pooled manager. The session's `SharedApiKeyProvider` is not
forwarded at all: it is hard-wired to the primary manager, so at a
pooled platform's host it would resolve the wrong bearer. A platform's
own `embed_api_key` is untouched and keeps serving its own endpoint.

The background reindex built a second provider straight from
`ApiEmbeddingProvider::from_session`, outside the chokepoint and without
401 refresh; it now embeds through the session's own params.

Test strength verified by mutation: with the guard reverted, exactly
`session_credentials_are_withheld_from_a_foreign_endpoint` and
`a_cloned_param_set_cannot_redirect_scoped_credentials` fail.
 2026-07-26 23:41:45 -04:00
ZacharyZhang-NY 867b3e110b docs(comments): trim paste-fix commentary to the crucial constraints 2026-07-24 13:09:35 -04:00
268 changed files with 16467 additions and 1717 deletions
Expand all files Collapse all files

No files matched your search

.gitignore
+7
View File
@@ -1,3 +1,10 @@
/target
**/*.rs.bk
.DS_Store
# frontend/ (kigicli.dev worker)
.wrangler/
node_modules/
# Runner-setup prompt; holds a Gitea registration token once filled in.
/prompt.txt
AGENTS.md
+81 -8
View File
@@ -14,16 +14,32 @@ official `kimi` CLI: binary `kigi`, config dir `~/.kigi`
Never read or write `~/.kimi` (except the explicit one-time read-only
import) or any `KIMI_*` env var.
Upstream sync record (update on every sync): fork baseline is upstream
snapshot `8adf901` (2026-07-16, ≈v0.2.102); bug fixes evaluated and
selectively ported through upstream `77cd7eb` (2026-08-25, ≈v1.0.10) in
the 0.1.13 cycle and through `3794978` (2026-09-09, ≈v1.0.24) in the
0.1.15 cycle. Upstream is daily "Synced from monorepo" snapshots —
the commit BODIES carry per-change bullet lists, and
`crates/codegen/xai-grok-shell/changelogs/` maps releases to dates;
start any future sync from those two, not from raw diffs.
Known limitations left by the 0.1.15 sync (fix before relying on them):
the Windows frame writer writes through `std::io::stderr`, so a console
that stops reading still blocks exit-path prints behind std's stderr lock
(unix silences fd 2 after a timed-out writer join); the bwrap command
builder re-checks `Path::exists("/data")` and can omit the devbox bind on
a stat error that the profile classification treated as present.
## Hard constraints
- **Zero egress**: outbound connections are limited to
`auth.kimi.com`, `api.kimi.com`, `api.moonshot.cn`, `api.moonshot.ai`,
GitHub Releases domains, user-configured MCP servers, the endpoints of
provider platforms the user has credentialed, and `models.dev` (model
metadata refresh — reached ONLY when an enabled platform's `/models` wire
lacks metadata, `wire_serves_metadata=false`; Kimi/Moonshot never trigger
it; `KIGI_MODELS_DEV_URL=0` disables). No telemetry, no analytics, ever.
`crates/codegen/kigi-env` is the single home of first-party endpoints.
- **Zero egress**: only `auth.kimi.com`, `api.kimi.com`, `api.moonshot.cn`,
`api.moonshot.ai`, GitHub Releases, user-configured MCP servers, credentialed
provider endpoints, and `models.dev` (metadata refresh, ONLY when an enabled
platform's `/models` lacks it — `wire_serves_metadata=false`, never
Kimi/Moonshot, `KIGI_MODELS_DEV_URL=0` disables). No telemetry, no analytics,
ever. `crates/codegen/kigi-env` is the single home of first-party endpoints.
- **One SSRF policy**: `ssrf::is_blocked_for_host` gates every model/settings
URL (`web_fetch` + hooks), per redirect hop. Never write a second copy.
- **Toolchain**: Rust 1.97.0 (rust-toolchain.toml), edition 2024.
- **Gates** (all must stay green):
`cargo check --workspace --all-targets`,
@@ -164,6 +180,63 @@ edges stay deterministic Rust. The harness appends a terminal
the replan cap; `{"ops": []}` is a respected free no-op; failures
degrade.
## Swarm (`agent_swarm` tool + `/swarm` mode)
Cheap, unstructured fan-out — the one thing `/graph` does not offer. One
`prompt_template` containing `{{item}}` is expanded over an `items` list
into up to `MAX_AGENT_SWARM_MEMBERS` (128) ordinary subagents, run to
completion, and returned as ONE aggregate. Ported from kimi-code; entirely
client-side, no backend surface.
- Engine (`kigi-tools/.../kigi/agent_swarm/`): `plan.rs` validates and
expands (every fault is reported BEFORE a member starts; expanded
prompts must be pairwise distinct), `schedule.rs` is the launch ramp as
pure arithmetic (5 immediate, then 1 per 700ms; capacity shrinks on a
provider rate limit and recovers after a quiet window;
`KIGI_AGENT_SWARM_MAX_CONCURRENCY` caps it and a malformed value is a
hard error), `run.rs` drives it against the EXISTING single-spawn
`SubagentBackend`, `tool.rs` is the tool.
- `SwarmMemberOutcome::Backgrounded` is load-bearing: a member that
outlives the 600s foreground budget is detached by the coordinator and
KEEPS RUNNING. It is never reported `Failed` and never offered for
resume — relaunching its item would put a second agent on the same
files. Members share the caller's tree with NO isolation; the
distinct-prompt rule is the only thing keeping them apart.
- `InFlightGuard` (run.rs) cancels live members when the runner's future
is dropped. Send-now cancels the turn WITHOUT cancelling subagents and
aborts the task; the dropped receivers read as "parent gone" and each
child re-attaches itself. There is no cooperative path to use instead —
`kigi_tool_runtime::Cancellation` is constructed nowhere.
- Both retries (`MAX_RATE_LIMIT_RETRIES`) and wall clock
(`MAX_SWARM_RUNTIME`) are bounded: the swarm blocks the caller's turn,
so every wait needs a ceiling. Stragglers at the deadline are reported
as still-running WITH their ids.
- Underneath every fan-out (swarm, graph, raw Task) sits the
session-scoped spawn bound: `handle_subagent_request` holds one permit
from the parent session's semaphore for the whole child run
(kigi-tools `task/admission`: 32 default via
`KIGI_MAX_CONCURRENT_SUBAGENTS`, queue|fail via
`KIGI_SUBAGENT_LIMIT_BEHAVIOR`), so wide fan-outs queue instead of
exhausting file descriptors.
- `ToolKind::AgentSwarm` is its own variant because `TemplateRenderer`'s
`by_kind` map holds ONE tool name per kind — sharing `Task` would
silently redirect `${{ tools.by_kind.task }}` in other tools' prompts.
It gates exactly as `Task` does (`capability.rs`), and `builder.rs`
strips it wherever `task` is stripped. `MAX_SUBAGENT_DEPTH` stays 1:
upstream's unlimited nesting is a hazard, not a feature.
- `/swarm | /swarm on|off | /swarm <task>` (`BuiltinGate::Swarm` = the
tool is in the toolset) arms a doctrine reminder via the existing
`push_system_reminder`. Two triggers only: `Manual` persists until
switched off, `Task` auto-expires at turn end via `SwarmTurnGuard` — a
DROP guard, because a user interrupt aborts the turn future and never
reaches post-loop code. `enter` is a total no-op while armed so the
`/swarm <task>` shorthand cannot downgrade a standing `/swarm on`.
- The mode is deliberately NOT persisted (unlike `/goal` and `/graph`,
which strand real work if lost): it is a prompt hint whose recovery is
retyping one command. The injected doctrine IS durable, so an explicit
`/swarm off` retracts UNCONDITIONALLY — a resumed or compacted session
can read as "off" with the instruction still in context.
## Provider registry & API-key auth (post-0.1.3 expansion)
- The platform registry is compiled-in spec rows in `kigi-models`
Cargo.lock
Generated
+88 -85
View File
@@ -228,7 +228,7 @@ version = "1.1.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "40c48f72fd53cd289104fc64099abca73db4166ad86ea0b4341abe65af83dadc"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.60.2",
]
[[package]]
@@ -251,7 +251,7 @@ checksum = "291e6a250ff86cd4a820112fb8898808a366d8f9f58ce16d1f538353ad55747d"
dependencies = [
"anstyle",
"once_cell_polyfill",
"windows-sys 0.61.2",
"windows-sys 0.60.2",
]
[[package]]
@@ -1626,9 +1626,9 @@ checksum = "f079e83a288787bcd14a6aea84cee5c87a67c5a3e660c30f557a3d24761b3527"
[[package]]
name = "chacha20"
version = "0.10.1"
version = "0.10.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d524456ba66e72eb8b115ff89e01e497f8e6d11d78b70b1aa13c0fbd97540a81"
checksum = "65c35e4b699c7e15ccbe7ee35c005e4fc0a278d22238a2857e6ce2dadeda1b06"
dependencies = [
"cfg-if",
"cpufeatures 0.3.0",
@@ -1810,7 +1810,7 @@ version = "3.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "faf9468729b8cbcea668e36183cb69d317348c2e08e994829fb56ebfdfbaac34"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.48.0",
]
[[package]]
@@ -2654,7 +2654,7 @@ dependencies = [
"libc",
"option-ext",
"redox_users 0.5.2",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -2980,7 +2980,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
dependencies = [
"libc",
"windows-sys 0.61.2",
"windows-sys 0.52.0",
]
[[package]]
@@ -3958,7 +3958,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b305d85504de270ad3525d726a6b69cc59ee7b2269b014387651107ab9f0755b"
dependencies = [
"bstr",
"hashbrown 0.17.1",
"hashbrown 0.15.5",
]
[[package]]
@@ -4497,9 +4497,9 @@ dependencies = [
[[package]]
name = "h2"
version = "0.4.15"
version = "0.4.19"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6cb093c84e8bd9b188d4c4a8cb6579fc016968d14c99882163cd3ff402a4f155"
checksum = "ef8e5e5a340588f4452631496976cf8636d4a7ecf600239fdc27615d2530bc16"
dependencies = [
"atomic-waker",
"bytes",
@@ -5442,7 +5442,7 @@ dependencies = [
[[package]]
name = "kigi-acp-lib"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"agent-client-protocol",
"async-trait",
@@ -5456,7 +5456,7 @@ dependencies = [
[[package]]
name = "kigi-agent"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"chrono",
"dirs 6.0.0",
@@ -5486,7 +5486,7 @@ dependencies = [
[[package]]
name = "kigi-agent-lifecycle"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"async-trait",
"tokio",
@@ -5495,7 +5495,7 @@ dependencies = [
[[package]]
name = "kigi-auth"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"async-trait",
"http 1.4.2",
@@ -5508,7 +5508,7 @@ dependencies = [
[[package]]
name = "kigi-bin"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"clap",
@@ -5543,7 +5543,7 @@ dependencies = [
[[package]]
name = "kigi-chat-state"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"indexmap",
"kigi-compaction",
@@ -5560,7 +5560,7 @@ dependencies = [
[[package]]
name = "kigi-codebase-graph"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"ahash",
"clap",
@@ -5596,7 +5596,7 @@ dependencies = [
[[package]]
name = "kigi-compaction"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"async-trait",
@@ -5609,7 +5609,7 @@ dependencies = [
[[package]]
name = "kigi-config"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"base64",
"blake3",
@@ -5632,7 +5632,7 @@ dependencies = [
[[package]]
name = "kigi-config-types"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"agent-client-protocol",
"indexmap",
@@ -5646,7 +5646,7 @@ dependencies = [
[[package]]
name = "kigi-crash-handler"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"backtrace",
"libc",
@@ -5657,7 +5657,7 @@ dependencies = [
[[package]]
name = "kigi-env"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"tracing",
"url",
@@ -5665,7 +5665,7 @@ dependencies = [
[[package]]
name = "kigi-fast-worktree"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"bytes",
@@ -5697,7 +5697,7 @@ dependencies = [
[[package]]
name = "kigi-file-utils"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"aws-config",
@@ -5721,7 +5721,7 @@ dependencies = [
[[package]]
name = "kigi-fsnotify"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"criterion",
"dunce",
@@ -5742,7 +5742,7 @@ dependencies = [
[[package]]
name = "kigi-gix-status"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"gix",
"kigi-test-utils",
@@ -5752,7 +5752,7 @@ dependencies = [
[[package]]
name = "kigi-hooks"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"fastrand",
"kigi-config",
@@ -5771,7 +5771,7 @@ dependencies = [
[[package]]
name = "kigi-hooks-plugins-types"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"serde",
"serde_json",
@@ -5779,7 +5779,7 @@ dependencies = [
[[package]]
name = "kigi-http"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"kigi-auth",
"kigi-log",
@@ -5794,7 +5794,7 @@ dependencies = [
[[package]]
name = "kigi-hunk-tracker"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"chrono",
"dunce",
@@ -5815,14 +5815,14 @@ dependencies = [
[[package]]
name = "kigi-interjection-core"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"serde",
]
[[package]]
name = "kigi-log"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"chrono",
@@ -5840,7 +5840,7 @@ dependencies = [
[[package]]
name = "kigi-markdown"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anstyle",
"anstyle-lossy",
@@ -5864,14 +5864,14 @@ dependencies = [
[[package]]
name = "kigi-markdown-core"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"pulldown-cmark",
]
[[package]]
name = "kigi-mcp"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"agent-client-protocol",
"async-trait",
@@ -5908,7 +5908,7 @@ dependencies = [
[[package]]
name = "kigi-memory"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"arc-swap",
@@ -5942,7 +5942,7 @@ dependencies = [
[[package]]
name = "kigi-mermaid"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"fontdb",
"image",
@@ -5960,7 +5960,7 @@ dependencies = [
[[package]]
name = "kigi-models"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"kigi-env",
"serde",
@@ -5970,7 +5970,7 @@ dependencies = [
[[package]]
name = "kigi-pager-minimal"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"chrono",
"crossterm",
@@ -5987,7 +5987,7 @@ dependencies = [
[[package]]
name = "kigi-pager-pty-harness"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"alacritty_terminal",
"anyhow",
@@ -6012,7 +6012,7 @@ dependencies = [
[[package]]
name = "kigi-pager-render"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"agent-client-protocol",
"anstyle",
@@ -6064,7 +6064,7 @@ dependencies = [
[[package]]
name = "kigi-paths"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"camino",
"serde",
@@ -6074,7 +6074,7 @@ dependencies = [
[[package]]
name = "kigi-prompt-queue"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"serde",
"serde_json",
@@ -6082,7 +6082,7 @@ dependencies = [
[[package]]
name = "kigi-proto-build"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"pbjson-build",
@@ -6093,7 +6093,7 @@ dependencies = [
[[package]]
name = "kigi-ratatui-inline"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"ansi-width",
"anstyle-parse 0.2.7",
@@ -6110,7 +6110,7 @@ dependencies = [
[[package]]
name = "kigi-ratatui-textarea"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"arboard",
"chrono",
@@ -6131,7 +6131,7 @@ dependencies = [
[[package]]
name = "kigi-sampler"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"async-openai",
"async-stream",
@@ -6141,6 +6141,7 @@ dependencies = [
"futures-util",
"indexmap",
"kigi-sampling-types",
"kigi-secrets",
"kigi-test-support",
"kigi-version",
"reqwest 0.12.28",
@@ -6154,7 +6155,7 @@ dependencies = [
[[package]]
name = "kigi-sampling-types"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"assert_matches",
"async-openai",
@@ -6171,7 +6172,7 @@ dependencies = [
[[package]]
name = "kigi-sandbox"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"chrono",
@@ -6186,13 +6187,14 @@ dependencies = [
"serde",
"serde_json",
"serial_test",
"tempfile",
"toml",
"tracing",
]
[[package]]
name = "kigi-secrets"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"regex",
"serde_json",
@@ -6230,7 +6232,7 @@ dependencies = [
[[package]]
name = "kigi-shell"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"agent-client-protocol",
"anyhow",
@@ -6293,6 +6295,7 @@ dependencies = [
"kigi-sampler",
"kigi-sampling-types",
"kigi-sandbox",
"kigi-secrets",
"kigi-shared",
"kigi-shell-base",
"kigi-sqlite-journal",
@@ -6367,7 +6370,7 @@ dependencies = [
[[package]]
name = "kigi-shell-base"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"chrono",
@@ -6392,7 +6395,7 @@ dependencies = [
[[package]]
name = "kigi-sqlite-journal"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"libc",
"rusqlite",
@@ -6403,7 +6406,7 @@ dependencies = [
[[package]]
name = "kigi-subagent-resolution"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"kigi-sampling-types",
"kigi-tool-types",
@@ -6418,7 +6421,7 @@ dependencies = [
[[package]]
name = "kigi-system-power"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"windows-sys 0.59.0",
"zbus",
@@ -6426,7 +6429,7 @@ dependencies = [
[[package]]
name = "kigi-test-support"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"agent-client-protocol",
"anyhow",
@@ -6448,7 +6451,7 @@ dependencies = [
[[package]]
name = "kigi-test-utils"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"runfiles",
"tracing",
@@ -6457,11 +6460,11 @@ dependencies = [
[[package]]
name = "kigi-token-estimation"
version = "0.1.9"
version = "0.1.15"
[[package]]
name = "kigi-tool-protocol"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"kigi-tool-types",
"serde",
@@ -6472,7 +6475,7 @@ dependencies = [
[[package]]
name = "kigi-tool-runtime"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"async-trait",
@@ -6490,7 +6493,7 @@ dependencies = [
[[package]]
name = "kigi-tool-types"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"minijinja",
"schemars 1.2.1",
@@ -6500,7 +6503,7 @@ dependencies = [
[[package]]
name = "kigi-tools"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"arc-swap",
@@ -6577,7 +6580,7 @@ dependencies = [
[[package]]
name = "kigi-tools-api"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"kigi-proto-build",
"kigi-tool-protocol",
@@ -6590,11 +6593,11 @@ dependencies = [
[[package]]
name = "kigi-tracing-macros"
version = "0.1.9"
version = "0.1.15"
[[package]]
name = "kigi-tty-utils"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"libc",
"nix 0.30.1",
@@ -6604,7 +6607,7 @@ dependencies = [
[[package]]
name = "kigi-tui"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"agent-client-protocol",
"ansi-to-tui",
@@ -6691,7 +6694,7 @@ dependencies = [
[[package]]
name = "kigi-update"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"dunce",
@@ -6720,14 +6723,14 @@ dependencies = [
[[package]]
name = "kigi-version"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"semver",
]
[[package]]
name = "kigi-workspace"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"agent-client-protocol",
"anyhow",
@@ -6806,7 +6809,7 @@ dependencies = [
[[package]]
name = "kigi-workspace-types"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"base64",
"chrono",
@@ -7305,7 +7308,7 @@ version = "0.6.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "536bfad37a309d62069485248eeaba1e8d9853aaf951caaeaed0585a95346f08"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.60.2",
]
[[package]]
@@ -7576,7 +7579,7 @@ version = "0.50.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7957b9740744892f114936ab4a57b3f487491bbeafaf8083688b16841a4240e5"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -7997,7 +8000,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7d8fae84b431384b68627d0f9b3b1245fcf9f46f6c0e3dc902e9dce64edd1967"
dependencies = [
"libc",
"windows-sys 0.61.2",
"windows-sys 0.48.0",
]
[[package]]
@@ -8840,7 +8843,7 @@ dependencies = [
[[package]]
name = "ptyctl"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"alacritty_terminal",
"anyhow",
@@ -8858,7 +8861,7 @@ dependencies = [
[[package]]
name = "ptyctl-cli"
version = "0.1.9"
version = "0.1.15"
dependencies = [
"anyhow",
"axum",
@@ -8992,7 +8995,7 @@ dependencies = [
"once_cell",
"socket2",
"tracing",
"windows-sys 0.61.2",
"windows-sys 0.52.0",
]
[[package]]
@@ -9617,7 +9620,7 @@ dependencies = [
"errno",
"libc",
"linux-raw-sys 0.12.1",
"windows-sys 0.61.2",
"windows-sys 0.52.0",
]
[[package]]
@@ -9687,7 +9690,7 @@ dependencies = [
"security-framework 3.7.0",
"security-framework-sys",
"webpki-root-certs",
"windows-sys 0.61.2",
"windows-sys 0.52.0",
]
[[package]]
@@ -10536,7 +10539,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c3d1e2c7f27f8d4cb10542a02c49005dbd6e93095799d6f3be745fae9f8fedd4"
dependencies = [
"libc",
"windows-sys 0.61.2",
"windows-sys 0.60.2",
]
[[package]]
@@ -10937,7 +10940,7 @@ dependencies = [
"getrandom 0.4.3",
"once_cell",
"rustix 1.1.4",
"windows-sys 0.61.2",
"windows-sys 0.52.0",
]
[[package]]
@@ -11882,7 +11885,7 @@ checksum = "f2f6fb2847f6742cd76af783a2a2c49e9375d0a111c7bef6f71cd9e738c72d6e"
dependencies = [
"memoffset",
"tempfile",
"windows-sys 0.61.2",
"windows-sys 0.60.2",
]
[[package]]
@@ -12414,15 +12417,15 @@ dependencies = [
[[package]]
name = "webbrowser"
version = "1.2.1"
version = "1.2.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0fc95580916af1e68ff6a7be07446fc5db73ebf71cf092de939bbf5f7e189f72"
checksum = "62c35be770821a214dbc362fc26908c853e776c0004294d0b10b8a6bad582f94"
dependencies = [
"core-foundation 0.10.1",
"jni",
"log",
"ndk-context",
"objc2 0.6.4",
"objc2-app-kit",
"objc2-foundation 0.3.2",
"url",
"web-sys",
@@ -12561,7 +12564,7 @@ version = "0.1.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.48.0",
]
[[package]]
Cargo.toml
+1 -1
View File
@@ -76,7 +76,7 @@ members = [
]
[workspace.package]
version = "0.1.9"
version = "0.1.15"
edition = "2024"
license = "Apache-2.0"
README.md
+4 -4
View File
@@ -36,12 +36,12 @@ key? OpenAI, Anthropic, Google, DeepSeek, Groq, Moonshot and
```sh
# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/ZacharyZhang-NY/Kigi-CLI/main/install.sh | bash
curl -fsSL https://kigicli.dev/install.sh | bash
```
```powershell
# Windows PowerShell
irm https://raw.githubusercontent.com/ZacharyZhang-NY/Kigi-CLI/main/install.ps1 | iex
irm https://kigicli.dev/install.ps1 | iex
```
```sh
@@ -176,8 +176,8 @@ keyring entry, and its own `KIGI_*` env vars, and never touches what the
official `kimi` CLI installed. `kigi import-kimi` copies your old config over
once, read-only.
**Zero telemetry.** It talks to the APIs you configured, GitHub Releases, and
your own MCP servers. Nothing else.
**Zero telemetry.** It talks to the APIs you configured, the Kigi release
server, and your own MCP servers. Nothing else.
## License
SECURITY.md
+5 -3
View File
@@ -1,8 +1,10 @@
# Security Policy
Kigi is a community build; report vulnerabilities via
[GitHub private vulnerability reporting](https://github.com/ZacharyZhang-NY/Kigi-CLI/security/advisories/new)
on this repository. Please do not open public issues for security reports.
Kigi is a community build; report vulnerabilities on the
[Kigi issue tracker](https://github.com/ZacharyZhang-NY/Kigi-CLI/issues).
Issues there are public, so please leave out any detail that would make a
report directly exploitable — open with the affected area and impact, and a
maintainer will follow up for the rest.
Dependency advisories are gated in CI with `cargo deny check advisories`;
documented exceptions live in `deny.toml`.
THIRD-PARTY-NOTICES.md
+1511 -81
View File
File diff suppressed because it is too large. Load diff
crates/codegen/kigi-agent/src/builder.rs
+33 -4
View File
@@ -711,9 +711,19 @@ impl AgentBuilder {
kigi_tools::types::tool::ToolNamespace::Kigi,
"task"
);
// The swarm is a fan-out of subagent spawns, so it lives and dies with
// the task tool: any condition that leaves no subagent to spawn leaves
// the swarm with nothing to fan out to.
let swarm_tool_id = format!(
"{}:{}",
kigi_tools::types::tool::ToolNamespace::Kigi,
"agent_swarm"
);
let mut task_stripped = false;
if !self.subagents_enabled {
tool_config.tools.retain(|tc| tc.id != task_tool_id);
tool_config
.tools
.retain(|tc| tc.id != task_tool_id && tc.id != swarm_tool_id);
task_stripped = true;
} else {
let subagents = crate::discovery::all_subagents_with_plugins(
@@ -722,7 +732,9 @@ impl AgentBuilder {
self.plugin_registry.as_deref(),
);
if subagents.is_empty() {
tool_config.tools.retain(|tc| tc.id != task_tool_id);
tool_config
.tools
.retain(|tc| tc.id != task_tool_id && tc.id != swarm_tool_id);
task_stripped = true;
} else if self.prompt_audience == crate::prompt::context::PromptAudience::Subagent {
if let Some(task_tc) = tool_config
@@ -810,7 +822,13 @@ impl AgentBuilder {
.tools
.iter()
.any(|t| AGENT_TASK_CLASSIFIER_RE.is_match(t));
let task_deps = ["task", "get_task_output", "kill_task", "wait_tasks"];
let task_deps = [
"task",
"agent_swarm",
"get_task_output",
"kill_task",
"wait_tasks",
];
let registered_tool_ids = tool_bridge_builder.known_tool_ids();
let present_kinds: std::collections::HashSet<ToolKind> =
tool_config.tools.iter().filter_map(|tc| tc.kind).collect();
@@ -925,7 +943,13 @@ impl AgentBuilder {
}
}
if definition.allowed_subagent_types.as_deref() == Some(&[]) {
let task_deps = ["task", "get_task_output", "kill_task", "wait_tasks"];
let task_deps = [
"task",
"agent_swarm",
"get_task_output",
"kill_task",
"wait_tasks",
];
tool_config
.tools
.retain(|tc| !task_deps.contains(&short_tool_name(&tc.id)));
@@ -1601,6 +1625,11 @@ mod tests {
has_task, *subagents,
"[{label}] spawn_subagent presence should match subagents_enabled={subagents}; got tools: {names:?}"
);
let has_swarm = names.contains(&"agent_swarm");
assert_eq!(
has_swarm, *subagents,
"[{label}] agent_swarm fans out to subagents, so it must follow the same gate as spawn_subagent; got tools: {names:?}"
);
assert!(
names.contains(&"enter_plan_mode"),
"[{label}] enter_plan_mode must always be present (TUI plan-mode keybind needs it); got tools: {names:?}"
crates/codegen/kigi-agent/src/config.rs
+10
View File
@@ -154,6 +154,11 @@ fn task_tool_config() -> ToolConfig {
.with_name("spawn_subagent")
.with_param_rename("run_in_background", "background")
}
/// Swarm tool. Keeps its registry name: unlike `task` it has no CLI-specific
/// alias, and the name is what the model is told to reuse for `resume_agent_ids`.
fn agent_swarm_tool_config() -> ToolConfig {
ToolConfig::from(&kigi::AgentSwarmTool)
}
/// Task output tool renamed for clarity:
/// `get_task_output` → `get_command_or_subagent_output`.
fn task_output_tool_config() -> ToolConfig {
@@ -270,6 +275,7 @@ fn default_kigi_toolset() -> ToolServerConfig {
task_output_tool_config(),
wait_tasks_tool_config(),
task_tool_config(),
agent_swarm_tool_config(),
(&kigi::SchedulerCreateTool).into(),
(&kigi::SchedulerDeleteTool).into(),
(&kigi::SchedulerListTool).into(),
@@ -318,6 +324,7 @@ pub fn kigi_hashline_toolset(
task_output_tool_config(),
wait_tasks_tool_config(),
task_tool_config(),
agent_swarm_tool_config(),
(&kigi::WebSearchTool).into(),
(&kigi::SchedulerCreateTool).into(),
(&kigi::SchedulerDeleteTool).into(),
@@ -400,6 +407,7 @@ fn kigi_plan_toolset() -> ToolServerConfig {
(&kigi::TodoWriteTool).into(),
task_output_tool_config(),
task_tool_config(),
agent_swarm_tool_config(),
(&kigi::SchedulerCreateTool).into(),
(&kigi::SchedulerDeleteTool).into(),
(&kigi::SchedulerListTool).into(),
@@ -428,6 +436,7 @@ fn orchestrator_toolset() -> ToolServerConfig {
(&kigi::ListDirTool).into(),
(&kigi::GrepTool).into(),
task_tool_config(),
agent_swarm_tool_config(),
task_output_tool_config(),
wait_tasks_tool_config(),
kill_task_tool_config(),
@@ -497,6 +506,7 @@ fn kigi_ask_user_toolset() -> ToolServerConfig {
task_output_tool_config(),
wait_tasks_tool_config(),
task_tool_config(),
agent_swarm_tool_config(),
(&kigi::SchedulerCreateTool).into(),
(&kigi::SchedulerDeleteTool).into(),
(&kigi::SchedulerListTool).into(),
crates/codegen/kigi-agent/src/discovery.rs
+37
View File
@@ -252,11 +252,20 @@ pub fn by_name(name: &str) -> Option<AgentDefinition> {
by_name_with_home(name, dirs::home_dir().as_deref(), kigi.as_deref())
}
/// An agent name is one file stem: a separator or `..` would resolve a file outside the searched directory with that directory's trust scope.
fn is_single_path_component(name: &str) -> bool {
!name.is_empty() && name != "." && name != ".." && !name.contains(['/', '\\'])
}
fn by_name_with_home(
name: &str,
home: Option<&Path>,
kigi_home: Option<&Path>,
) -> Option<AgentDefinition> {
if !is_single_path_component(name) {
tracing::warn!(name, "rejected agent name: not a single path component");
return None;
}
// Check built-ins first — type-safe via BuiltinAgentName strum enum.
// Legacy pre-rebrand agent types (persisted in old session files) are
// mapped onto their current names first; see `canonical_agent_type`.
@@ -298,6 +307,10 @@ fn by_name_in_cwd_with_home(
home: Option<&Path>,
kigi_home: Option<&Path>,
) -> Option<AgentDefinition> {
if !is_single_path_component(name) {
tracing::warn!(name, "rejected agent name: not a single path component");
return None;
}
if let Some(def) = load_project_definition_by_name(name, cwd) {
return Some(def);
}
@@ -827,6 +840,30 @@ mod tests {
assert!(def.is_none());
}
/// `../outside` must not load a file beyond the agents directory under that directory's scope.
#[test]
fn by_name_rejects_names_that_leave_the_agents_directory() {
let tmp = tempfile::tempdir().unwrap();
let project = tmp.path().join("project");
let agents_dir = project.join(".kigi").join("agents");
fs::create_dir_all(&agents_dir).unwrap();
write_agent_file(
&project,
"outside.md",
"outside",
"Lives outside the agents dir",
);
write_agent_file(&agents_dir, "inside.md", "inside", "Lives inside");
assert!(by_name_in_cwd_with_home("../../outside", &project, None, None).is_none());
assert!(by_name_in_cwd_with_home("..", &project, None, None).is_none());
assert!(by_name_in_cwd_with_home("a/b", &project, None, None).is_none());
assert!(by_name_with_home("../../outside", None, None).is_none());
assert_eq!(
by_name_in_cwd_with_home("inside", &project, None, None).map(|d| d.name),
Some("inside".to_string())
);
}
#[test]
fn test_discover_finds_md_files() {
let tmp = tempfile::tempdir().unwrap();
crates/codegen/kigi-agent/src/error.rs
+5
View File
@@ -17,6 +17,11 @@ pub enum AgentBuildError {
#[error("IO error during agent construction: {0}")]
IoError(#[from] std::io::Error),
/// The per-session runtime could not be built — fd or thread
/// exhaustion (EMFILE/EAGAIN), not a definition problem.
#[error("failed to build the session runtime: {0}")]
RuntimeBuild(std::io::Error),
/// Carries template line numbers and surrounding context.
#[error("template rendering error: {0}")]
MiniJinjaError(#[from] minijinja::Error),
crates/codegen/kigi-agent/src/prompt/ignore.rs
+1 -1
View File
@@ -32,5 +32,5 @@ fn get_global_gitignore_path() -> Option<PathBuf> {
git2::Config::open_default()
.ok()
.and_then(|cfg| cfg.get_path("core.excludesFile").ok())
.or_else(|| std::env::var_os("HOME").map(|h| PathBuf::from(h).join(".gitignore")))
.or_else(|| std::env::home_dir().map(|h| h.join(".gitignore")))
}
crates/codegen/kigi-bin/src/main.rs
+80 -29
View File
@@ -146,27 +146,33 @@ async fn run_setup_command(json: bool) {
.expect("setup report has no non-serializable values");
println!("{out}");
if !report.configured {
eprintln!(
"Your team doesn't have a managed configuration yet. Ask a team admin to provision one."
kigi_tui::best_effort_stderr::eprint_line(
"Your team doesn't have a managed configuration yet. Ask a team admin to provision one.",
);
}
}
Err(e) => {
eprintln!("Couldn't fetch managed configuration. {e}");
kigi_tui::best_effort_stderr::eprint_line(&format!(
"Couldn't fetch managed configuration. {e}"
));
std::process::exit(1);
}
}
return;
}
match managed_config::run_setup().await {
SetupOutcome::Installed => eprintln!("Applied managed configuration."),
SetupOutcome::Installed => {
kigi_tui::best_effort_stderr::eprint_line("Applied managed configuration.")
}
SetupOutcome::NothingConfigured => {
eprintln!(
"Your team doesn't have a managed configuration yet. Ask a team admin to provision one."
kigi_tui::best_effort_stderr::eprint_line(
"Your team doesn't have a managed configuration yet. Ask a team admin to provision one.",
);
}
SetupOutcome::Failed(e) => {
eprintln!("Couldn't apply managed configuration. {e}");
kigi_tui::best_effort_stderr::eprint_line(&format!(
"Couldn't apply managed configuration. {e}"
));
std::process::exit(1);
}
}
@@ -208,8 +214,8 @@ async fn run_leader_mgmt(args: LeaderMgmtArgs) -> Result<()> {
println!("{info:#?}");
} else {
print_leader_descriptor(&descriptor);
eprintln!(
" (detailed info unavailable — leader does not advertise control capabilities)"
kigi_tui::best_effort_stderr::eprint_line(
" (detailed info unavailable — leader does not advertise control capabilities)",
);
}
client.cancel();
@@ -220,7 +226,7 @@ async fn run_leader_mgmt(args: LeaderMgmtArgs) -> Result<()> {
async fn kill_leaders() -> Result<()> {
let leaders = kigi_shell::leader::discover_leaders().await;
if leaders.is_empty() {
eprintln!("No leader candidates found.");
kigi_tui::best_effort_stderr::eprint_line("No leader candidates found.");
return Ok(());
}
let mut killed = 0u32;
@@ -231,7 +237,9 @@ async fn kill_leaders() -> Result<()> {
};
if !kigi_shell::util::is_kigi_process(pid) {
if let Some(ref lock) = d.lock_path {
eprintln!(" PID {pid} is not a kigi process, removing stale lock");
kigi_tui::best_effort_stderr::eprint_line(&format!(
" PID {pid} is not a kigi process, removing stale lock"
));
let _ = std::fs::remove_file(lock);
cleaned += 1;
}
@@ -240,19 +248,23 @@ async fn kill_leaders() -> Result<()> {
}
continue;
}
eprintln!(" Killing leader PID {pid}");
kigi_tui::best_effort_stderr::eprint_line(&format!(" Killing leader PID {pid}"));
if let Err(e) = kigi_shell::util::kill_process_by_pid(pid) {
eprintln!(" warning: failed to terminate PID {pid}: {e}");
kigi_tui::best_effort_stderr::eprint_line(&format!(
" warning: failed to terminate PID {pid}: {e}"
));
continue;
}
killed += 1;
}
if killed > 0 {
eprintln!("Killed {killed} leader process(es).");
kigi_tui::best_effort_stderr::eprint_line(&format!("Killed {killed} leader process(es)."));
} else if cleaned > 0 {
eprintln!("No live leader processes found (cleaned up {cleaned} stale lock(s)).");
kigi_tui::best_effort_stderr::eprint_line(&format!(
"No live leader processes found (cleaned up {cleaned} stale lock(s))."
));
} else {
eprintln!("No live leader processes found.");
kigi_tui::best_effort_stderr::eprint_line("No live leader processes found.");
}
Ok(())
}
@@ -295,7 +307,7 @@ fn print_leader_descriptor(d: &LeaderDescriptor) {
.map(|p| p.display().to_string())
.unwrap_or_else(|| "?".into());
let state = format!("{:?}", d.classification);
eprintln!(" PID {pid} ({state}) -- {sock}");
kigi_tui::best_effort_stderr::eprint_line(&format!(" PID {pid} ({state}) -- {sock}"));
}
fn leader_descriptor_json(d: &LeaderDescriptor) -> serde_json::Value {
serde_json::json!(
@@ -1221,15 +1233,34 @@ fn main() {
"Found crashed sessions from a previous run"
);
}
let runtime = tokio::runtime::Builder::new_multi_thread()
// Capped workers: tokio's per-core default pins 100+ thread slots on
// many-core shared hosts and later spawns die with EAGAIN.
let runtime = match tokio::runtime::Builder::new_multi_thread()
.worker_threads(kigi_tty_utils::runtime::capped_worker_threads().get())
.enable_all()
.build()
.unwrap_or_else(|e| panic!("failed to start tokio runtime: {e}"));
{
Ok(runtime) => runtime,
Err(e) => {
// A host out of threads (RLIMIT_NPROC / pids.max) is an
// environment problem, not a bug — report it, don't panic.
kigi_tty_utils::restore_native_stderr();
kigi_tui::best_effort_stderr::eprint_line(&format!(
"Error: failed to start the async runtime: {e}"
));
kigi_tui::best_effort_stderr::eprint_line(
"The host may be out of threads (RLIMIT_NPROC or the cgroup pids ceiling); \
close other processes or raise the limit, then retry.",
);
std::process::exit(1);
}
};
let result = run_and_shutdown(runtime, async_main(), RUNTIME_SHUTDOWN_GRACE);
kigi_log::debug_log::flush();
if let Err(e) = result {
kigi_tty_utils::restore_native_stderr();
eprintln!("Error: {e:#}");
// fd 2 may be a closed pane: one attempt, never a panic under panic = "abort".
kigi_tui::best_effort_stderr::eprint_line(&format!("Error: {e:#}"));
std::process::exit(1);
}
}
@@ -1580,16 +1611,25 @@ async fn async_main() -> Result<()> {
let result = kigi_tui::app::run(args, bg_update_rx).await;
kigi_sandbox::flush();
match result {
Ok(true) => {
Ok(exit) if exit.quit_for_update => {
let adopted = bg_update_wait.lock().await.take();
if finish_update_on_exit(adopted, &update_config).await {
eprintln!("Update installed. Run `kigi` to start.");
} else {
eprintln!("Update did not complete. Run `kigi update` to retry.");
let installed =
finish_update_on_exit(adopted, &update_config, exit.terminal_reading).await;
// A terminal that stopped reading at teardown would block these prints.
if exit.terminal_reading {
if installed {
kigi_tui::best_effort_stderr::eprint_line(
"Update installed. Run `kigi` to start.",
);
} else {
kigi_tui::best_effort_stderr::eprint_line(
"Update did not complete. Run `kigi update` to retry.",
);
}
}
Ok(())
}
Ok(false) => Ok(()),
Ok(_) => Ok(()),
Err(e) => Err(e),
}
}
@@ -1605,10 +1645,15 @@ async fn async_main() -> Result<()> {
async fn finish_update_on_exit(
adopted: Option<tokio::task::JoinHandle<std::io::Result<std::process::ExitStatus>>>,
update_config: &UpdateConfig,
terminal_reading: bool,
) -> bool {
let run_blocking = |reason: Option<String>| async move {
// The blocking updater prints and its child inherits stderr: not on a stalled terminal.
if !terminal_reading {
return false;
}
if let Some(reason) = reason {
eprintln!("{reason}");
kigi_tui::best_effort_stderr::eprint_line(&reason);
}
auto_update::run_update_if_available(
auto_update::UpdateRunMode::Blocking,
@@ -1620,7 +1665,11 @@ async fn finish_update_on_exit(
};
match adopted {
Some(handle) => {
eprintln!("Waiting for the update download to finish...");
if terminal_reading {
kigi_tui::best_effort_stderr::eprint_line(
"Waiting for the update download to finish...",
);
}
match handle.await {
Ok(Ok(status)) if status.success() => true,
Ok(Ok(status)) => {
@@ -1793,7 +1842,9 @@ async fn signal_leaders_to_relaunch(installed_version: &str) {
to_version,
..
})) => {
eprintln!(" ↻ Relaunching shared session (leader {from_version} → {to_version})…");
kigi_tui::best_effort_stderr::eprint_line(&format!(
" ↻ Relaunching shared session (leader {from_version} → {to_version})…"
));
}
Ok(Ok(kigi_shell::leader::ControlPayload::RelaunchDeclined { reason })) => {
tracing::debug!(% reason, "Leader declined relaunch");
crates/codegen/kigi-chat-state/src/actor/mutations.rs
+20 -22
View File
@@ -390,14 +390,24 @@ impl ChatStateActor {
});
}
/// Replace the entire conversation, persist, re-estimate `total_tokens`,
/// and emit reset + token-update events.
///
/// Compaction replaces carry the provider-side overhead forward as a
/// *ratio* (`base_estimate × provider_total ÷ estimate_at_last_response`,
/// capped at the pre-compaction total; `base_estimate` when that estimate is
/// 0) so the reseed neither springs back nor over-counts (see
/// `COMPACTION.md`).
/// Reseed after a rewrite: the raw estimate scaled by the provider-confirmed ratio, capped at the pre-rewrite total (see COMPACTION.md).
pub(super) fn reseed_total_tokens(&self, base_estimate: u64) -> u64 {
let pre_replace_total = self.state.total_tokens;
let mut estimated_tokens =
if pre_replace_total > 0 && self.state.estimate_at_last_response > 0 {
let ratio = pre_replace_total as f64 / self.state.estimate_at_last_response as f64;
(base_estimate as f64 * ratio).round() as u64
} else {
base_estimate
};
// Over-counting fires premature compaction; an under-count self-heals on the next response.
if pre_replace_total > 0 {
estimated_tokens = estimated_tokens.min(pre_replace_total);
}
estimated_tokens
}
/// Replace the conversation, persist, reseed `total_tokens` and emit reset + token events; `is_compaction` only marks the turn capture.
pub(super) fn replace_conversation(
&mut self,
items: Vec<ConversationItem>,
@@ -407,28 +417,16 @@ impl ChatStateActor {
if is_compaction && let Some(cap) = &mut self.state.turn_capture {
cap.compaction_occurred = true;
}
let pre_replace_total = self.state.total_tokens;
// `harness_trace_buffer` / `harness_trace_turns` intentionally untouched:
// the planner/verifier subagents ran, so their sealed trace turns survive
// a conversation replace (same intent as the `TruncateToPromptIndex` arm).
self.persistence.replace_history(&items);
let base_estimate = super::state::estimate_conversation_tokens(&items);
let mut estimated_tokens =
if is_compaction && pre_replace_total > 0 && self.state.estimate_at_last_response > 0 {
let ratio = pre_replace_total as f64 / self.state.estimate_at_last_response as f64;
(base_estimate as f64 * ratio).round() as u64
} else {
base_estimate
};
// Compaction must never appear to increase usage.
if is_compaction && pre_replace_total > 0 {
estimated_tokens = estimated_tokens.min(pre_replace_total);
}
let estimated_tokens = self.reseed_total_tokens(base_estimate);
self.state.conversation = items;
self.state.estimated_tokens_since_model = 0;
self.state.total_tokens = estimated_tokens;
self.state.estimate_at_last_response =
super::state::estimate_conversation_tokens(&self.state.conversation);
self.state.estimate_at_last_response = base_estimate;
self.rebase_turn_capture_offset();
self.send_event(ChatStateEvent::ConversationReset {
new_len: self.state.conversation.len(),
crates/codegen/kigi-chat-state/src/actor/queries.rs
+3 -3
View File
@@ -68,10 +68,10 @@ impl ChatStateActor {
self.state.conversation.truncate(truncate_at);
self.state.prompt_texts.truncate(target_prompt_index);
self.state.prompt_index = target_prompt_index;
self.state.total_tokens =
super::state::estimate_conversation_tokens(&self.state.conversation);
let base_estimate = super::state::estimate_conversation_tokens(&self.state.conversation);
self.state.total_tokens = self.reseed_total_tokens(base_estimate);
self.state.estimated_tokens_since_model = 0;
self.state.estimate_at_last_response = self.state.total_tokens;
self.state.estimate_at_last_response = base_estimate;
self.persistence.replace_history(&self.state.conversation);
crates/codegen/kigi-chat-state/src/actor/state.rs
+2 -4
View File
@@ -68,12 +68,10 @@ pub fn estimate_item_tokens(item: &ConversationItem) -> u64 {
kigi_token_estimation::estimate_tokens(&b.text_summary())
}
ConversationItem::Reasoning(r) => {
// Summary + content text follow the standard bytes-per-token
// estimate; encrypted blobs are base64 and don't survive
// tokenization 1:1, so estimate at len/4 as well.
// Text and ciphertext are the same reasoning twice: the larger wins (base64 is 4/3 over raw).
let text_bytes = kigi_sampling_types::reasoning_item_text(r).len();
let enc_bytes = r.encrypted_content.as_deref().map(str::len).unwrap_or(0);
((text_bytes + enc_bytes) as u64) / kigi_token_estimation::BYTES_PER_TOKEN
(text_bytes.max(enc_bytes * 3 / 4) as u64) / kigi_token_estimation::BYTES_PER_TOKEN
}
}
}
crates/codegen/kigi-chat-state/src/actor/tests.rs
+48 -4
View File
@@ -658,22 +658,66 @@ async fn compaction_reseed_without_provider_count_matches_plain_estimate() {
}
#[tokio::test]
async fn non_compaction_replace_does_not_carry_overhead() {
async fn non_compaction_replace_carries_confirmed_total() {
// A rewind/mode-switch replace scales from the provider-confirmed count, not the raw estimate.
let h = TestHarness::new();
h.handle
.push_user_message(ConversationItem::user("x".repeat(4000)));
h.handle.record_token_usage(51_000);
h.handle.record_token_usage(500);
// Estimate at last response = 1_000, confirmed = 500 → ratio 0.5.
h.handle
.replace_conversation(vec![ConversationItem::user("q".repeat(4000))]);
let total = h.handle.get_total_tokens().await;
assert_eq!(
total, 1_000,
"non-compaction replace (e.g. rewind) keeps the plain estimate"
total, 500,
"same-size replace carries the provider-confirmed count"
);
}
#[tokio::test]
async fn replace_never_increases_total_tokens() {
// A growing replace is capped at the confirmed total; the under-count self-heals on the next usage.
let h = TestHarness::new();
h.handle
.push_user_message(ConversationItem::user("x".repeat(4000)));
h.handle.record_token_usage(1_500);
h.handle
.replace_conversation(vec![ConversationItem::user("q".repeat(40_000))]);
assert_eq!(h.handle.get_total_tokens().await, 1_500);
}
#[tokio::test]
async fn truncate_scales_from_confirmed_total() {
// Rewind path: TruncateToPromptIndex must use the same carry as replace_conversation.
let h = TestHarness::new();
h.handle
.push_user_message(ConversationItem::user("x".repeat(4000)));
h.handle.increment_prompt_index();
h.handle
.push_user_message(ConversationItem::user("y".repeat(4000)));
h.handle.increment_prompt_index();
h.handle.record_token_usage(1_000);
// Estimate at last response = 2_000, confirmed = 1_000 → ratio 0.5.
h.handle.truncate_to_prompt_index(1).await;
// Keeps the first user item (raw estimate 1_000) → scaled to 500.
assert_eq!(h.handle.get_total_tokens().await, 500);
}
#[test]
fn reasoning_estimate_takes_the_larger_of_text_and_ciphertext() {
// "thinking for r1" is 15 bytes; 800 base64 bytes are ~600 raw bytes.
let encrypted = reasoning_sibling("r1", Some(&"A".repeat(800)));
assert_eq!(crate::estimate_conversation_tokens(&[encrypted]), 150);
let plain = reasoning_sibling("r1", None);
assert_eq!(crate::estimate_conversation_tokens(&[plain]), 3);
}
#[tokio::test]
async fn flush_calls_persistence_flush() {
let mut h = TestHarness::new();
crates/codegen/kigi-config-types/src/lib.rs
+5
View File
@@ -668,6 +668,11 @@ pub struct RemoteSettings {
/// remote settings verbatim-input flag; env (`KIGI_COMPACTION_VERBATIM_INPUT`) and config override it. `None` = default (true).
#[serde(default)]
pub compaction_verbatim_input: Option<bool>,
/// Compaction summarizer `tool_choice` (`auto` | `none`) from remote
/// settings. Env (`KIGI_COMPACTION_TOOL_CHOICE`) and config override it.
/// Some backends reject `tool_choice: "none"` with tools attached.
#[serde(default)]
pub compaction_tool_choice: Option<String>,
/// Master switch for jemalloc heap sampling + threshold dumps.
/// `Some(true)` enables, `Some(false)` kill-switch, `None` = client default off.
#[serde(default)]
crates/codegen/kigi-config/src/shell.rs
+11 -13
View File
@@ -78,11 +78,10 @@ pub fn detect_windows_shell() -> &'static WindowsShell {
kigi_tty_utils::detach_std_command(&mut cmd);
cmd.arg("pwsh.exe").stdin(std::process::Stdio::null());
cmd.output()
} {
if output.status.success() {
tracing::info!("Windows shell: pwsh");
return WindowsShell::Pwsh;
}
} && output.status.success()
{
tracing::info!("Windows shell: pwsh");
return WindowsShell::Pwsh;
}
// powershell.exe (Windows PowerShell 5.1).
@@ -131,14 +130,13 @@ fn find_git_bash() -> Option<String> {
kigi_tty_utils::detach_std_command(&mut cmd);
cmd.arg("bash.exe").stdin(std::process::Stdio::null());
cmd.output()
} {
if output.status.success() {
let stdout = String::from_utf8_lossy(&output.stdout);
for line in stdout.lines() {
let line = line.trim();
if line.to_ascii_lowercase().contains("git") {
return Some(line.to_string());
}
} && output.status.success()
{
let stdout = String::from_utf8_lossy(&output.stdout);
for line in stdout.lines() {
let line = line.trim();
if line.to_ascii_lowercase().contains("git") {
return Some(line.to_string());
}
}
}
crates/codegen/kigi-crash-handler/src/handler.rs
+1 -1
View File
@@ -616,7 +616,7 @@ mod win {
}
let signal = exception_to_signal(exception_code);
let si_code = exception_code as i32;
let si_code = exception_code;
// ExceptionInformation[1] holds the faulting address for ACCESS_VIOLATION.
let si_addr = if exception_code == EXCEPTION_ACCESS_VIOLATION
crates/codegen/kigi-env/src/lib.rs
+3 -2
View File
@@ -11,7 +11,8 @@ pub struct KigiEndpoints {
pub coding_api_base_url: &'static str,
/// OAuth device-flow host for Kimi Code subscription login.
pub oauth_host: &'static str,
/// GitHub Releases API endpoint the self-updater polls.
/// Releases API endpoint the self-updater polls (GitHub-Releases-shaped;
/// Gitea's `/api/v1` serves the same wire format).
pub update_base_url: &'static str,
/// Human-facing page for subscription upgrade guidance.
pub upgrade_page_url: &'static str,
@@ -29,7 +30,7 @@ pub const CODE_BASE_URL_ENV: &str = "KIGI_CODE_BASE_URL";
/// Env var overriding [`oauth_host`] (PRD F1).
pub const OAUTH_HOST_ENV: &str = "KIGI_OAUTH_HOST";
/// Env var overriding [`update_base_url`] (PRD F8). Points the self-updater
/// at an alternate GitHub-Releases-shaped API (mirrors, tests).
/// at an alternate GitHub/Gitea-Releases-shaped API (mirrors, tests).
pub const UPDATE_BASE_URL_ENV: &str = "KIGI_UPDATE_BASE_URL";
fn resolve(var: &str, compiled: &'static str) -> String {
crates/codegen/kigi-fast-worktree/src/copy/cow.rs
+1
View File
@@ -76,6 +76,7 @@ mod tests {
assert_eq!(std::fs::read(&dest).unwrap(), data);
}
#[cfg(unix)]
#[test]
fn test_clone_preserves_permissions() {
use std::os::unix::fs::PermissionsExt;
crates/codegen/kigi-fast-worktree/src/db/mod.rs
+8 -6
View File
@@ -341,12 +341,14 @@ pub fn resolve_kigi_home() -> Result<PathBuf> {
if let Ok(v) = std::env::var("KIGI_SHARE_DIR") {
return Ok(PathBuf::from(v));
}
let home =
PathBuf::from(std::env::var("HOME").context("neither $KIGI_SHARE_DIR nor $HOME is set")?);
// Canonicalize the home dir so worktree paths share the same physical .kigi
// tree as trust/hooks even when it is symlinked. The dunce canonicalization
// must stay in sync with kigi_config::default_kigi_home();
// home resolution deliberately differs ($HOME here vs std::env::home_dir()).
// Standard home-dir lookup (`USERPROFILE`/Known Folders on Windows, where
// `HOME` is usually unset outside Git Bash) — the same resolution as
// kigi_config::default_kigi_home(), so worktree paths land in the same
// physical .kigi tree as trust/hooks.
let home = std::env::home_dir()
.context("neither $KIGI_SHARE_DIR nor a home directory could be resolved")?;
// Canonicalize so a symlinked home still maps to one physical tree; the
// dunce canonicalization must stay in sync with default_kigi_home().
Ok(dunce::canonicalize(&home).unwrap_or(home).join(".kigi"))
}
crates/codegen/kigi-fast-worktree/src/worktree/mod.rs
+1
View File
@@ -37,6 +37,7 @@ pub(crate) fn execute_plan(plan: WorktreePlan) -> Result<CreateWorktreeResult> {
#[cfg(test)]
mod tests {
#[cfg(unix)]
use super::*;
use crate::{IgnoredFilesMode, WorkingTreeMode, WorktreeBuilder};
use kigi_test_utils::git::{git_commit_all, init_git_repo};
crates/codegen/kigi-file-utils/src/workspace_classifier.rs
+5 -5
View File
@@ -89,10 +89,10 @@ fn is_platform_system_dir(cwd: &Path) -> bool {
#[cfg(target_os = "windows")]
fn is_platform_system_dir(cwd: &Path) -> bool {
if let Ok(temp) = std::env::var("TEMP").or_else(|_| std::env::var("TMP")) {
if cwd.starts_with(&temp) {
return true;
}
if let Ok(temp) = std::env::var("TEMP").or_else(|_| std::env::var("TMP"))
&& cwd.starts_with(&temp)
{
return true;
}
let path_lower = cwd.to_string_lossy().to_lowercase();
@@ -103,7 +103,7 @@ fn is_platform_system_dir(cwd: &Path) -> bool {
return true;
}
if cwd.parent().map_or(false, |p| p.parent().is_none()) && cwd.to_string_lossy().len() <= 3 {
if cwd.parent().is_some_and(|p| p.parent().is_none()) && cwd.to_string_lossy().len() <= 3 {
return true;
}
crates/codegen/kigi-fsnotify/src/watcher.rs
+8 -2
View File
@@ -971,7 +971,7 @@ pub(crate) fn start_with_timeout(
let progress_for_thread = progress.clone();
let thread = std::thread::spawn(move || {
let watcher_loop = move || {
let update_stage = |stage: &'static str| {
if let Ok(mut p) = progress_for_thread.lock() {
p.set_stage(stage);
@@ -1351,7 +1351,13 @@ pub(crate) fn start_with_timeout(
let _ = ready_tx.send(Err(Box::new(e)));
}
}
});
};
// Recoverable spawn: a host out of threads must surface as an error the
// caller can degrade on, not a `std::thread::spawn` panic.
let thread = std::thread::Builder::new()
.name("fsnotify-watcher".into())
.spawn(watcher_loop)
.map_err(|e| crate::FsNotifyError::WatcherStart(Box::new(e)))?;
// Wait for watcher to be ready (with timeout)
if let Ok(mut p) = progress.lock() {
crates/codegen/kigi-hooks/src/runner/http.rs
+56 -135
View File
@@ -26,68 +26,14 @@ struct HttpHookOutput {
reason: Option<String>,
}
/// CWE-918: Returns `true` if an IP address is in a private, link-local,
/// or cloud metadata range that should be blocked to prevent SSRF attacks.
/// CWE-918: whether this address is blocked for a hook.
///
/// Loopback (`127.x` / `::1`) is allowed for local development servers.
fn is_blocked_ip(ip: &IpAddr) -> bool {
match ip {
IpAddr::V4(v4) => {
let octets = v4.octets();
if octets[0] == 127 {
// loopback — allowed for local dev
return false;
}
if octets[0] == 10 {
// RFC 1918: 10.0.0.0/8
return true;
}
if octets[0] == 172 && (16..=31).contains(&octets[1]) {
// RFC 1918: 172.16.0.0/12
return true;
}
if octets[0] == 192 && octets[1] == 168 {
// RFC 1918: 192.168.0.0/16
return true;
}
if octets[0] == 169 && octets[1] == 254 {
// RFC 3927: 169.254.0.0/16 (link-local, cloud metadata)
return true;
}
if octets[0] == 100 && (64..=127).contains(&octets[1]) {
// RFC 6598: 100.64.0.0/10 (CGNAT)
return true;
}
if v4.is_unspecified() {
// 0.0.0.0
return true;
}
false
}
IpAddr::V6(v6) => {
if v6.is_loopback() {
// ::1 — allowed for local dev
return false;
}
if v6.is_unspecified() {
// ::
return true;
}
if let Some(v4) = v6.to_ipv4_mapped() {
return is_blocked_ip(&IpAddr::V4(v4));
}
let segments = v6.segments();
if segments[0] & 0xffc0 == 0xfe80 {
// fe80::/10 — link-local
return true;
}
if segments[0] & 0xfe00 == 0xfc00 {
// fc00::/7 — unique local (ULA)
return true;
}
false
}
}
/// Delegates to the `web_fetch` policy, so one implementation
/// governs every outbound URL. `allow_local` is on: a loopback hook
/// receiver is legitimate, but only when named literally, so
/// rebinding through a public name stays blocked.
fn is_blocked_ip(ip: &IpAddr, host: &str) -> bool {
kigi_tools::implementations::kigi::web_fetch::ssrf::is_blocked_for_host(ip, host, true)
}
/// CWE-918: Validate a hook URL to prevent SSRF.
@@ -112,7 +58,7 @@ async fn validate_hook_url(url: &str) -> Result<(), String> {
// If host is a literal IP, check it directly.
if let Ok(ip) = host.parse::<IpAddr>() {
if is_blocked_ip(&ip) {
if is_blocked_ip(&ip, host) {
return Err(format!("URL resolves to blocked private/internal IP: {ip}"));
}
return Ok(());
@@ -131,7 +77,7 @@ async fn validate_hook_url(url: &str) -> Result<(), String> {
}
for addr in &addrs {
if is_blocked_ip(&addr.ip()) {
if is_blocked_ip(&addr.ip(), host) {
return Err(format!(
"URL host {host} resolves to blocked private/internal IP: {}",
addr.ip()
@@ -571,76 +517,53 @@ mod tests {
}
}
// SSRF protection: is_blocked_ip tests
// SSRF protection: hook-side policy tests
//
// Range coverage lives with the shared predicate in `web_fetch::ssrf`;
// these pin what the runner adds on top.
/// Every range the shared policy knows is refused here.
#[test]
fn ssrf_blocks_rfc1918_10x() {
assert!(is_blocked_ip(&"10.0.0.1".parse().unwrap()));
assert!(is_blocked_ip(&"10.255.255.255".parse().unwrap()));
fn ssrf_delegates_to_the_shared_policy() {
for ip in [
"10.0.0.1",
"172.16.0.1",
"192.168.0.1",
"169.254.169.254",
"100.64.0.1",
"0.0.0.0",
"::",
"fe80::1",
"fc00::1",
"::ffff:10.0.0.1",
// Ranges the old hand-rolled copy missed entirely.
"198.18.0.1",
"192.0.2.1",
"203.0.113.1",
"240.0.0.1",
"64:ff9b::a9fe:a9fe",
] {
let ip: IpAddr = ip.parse().unwrap();
assert!(is_blocked_ip(&ip, &ip.to_string()), "{ip}");
}
for ip in ["1.1.1.1", "8.8.8.8", "172.32.0.1", "100.63.0.1"] {
let ip: IpAddr = ip.parse().unwrap();
assert!(!is_blocked_ip(&ip, &ip.to_string()), "{ip}");
}
}
/// A local hook receiver stays reachable when named literally.
#[test]
fn ssrf_blocks_rfc1918_172x() {
assert!(is_blocked_ip(&"172.16.0.1".parse().unwrap()));
assert!(is_blocked_ip(&"172.31.255.255".parse().unwrap()));
assert!(!is_blocked_ip(&"172.15.0.1".parse().unwrap()));
assert!(!is_blocked_ip(&"172.32.0.1".parse().unwrap()));
fn ssrf_allows_a_literal_loopback_hook_target() {
assert!(!is_blocked_ip(&"127.0.0.1".parse().unwrap(), "127.0.0.1"));
assert!(!is_blocked_ip(&"::1".parse().unwrap(), "localhost"));
}
/// A public name resolving to loopback is rebinding.
#[test]
fn ssrf_blocks_rfc1918_192168() {
assert!(is_blocked_ip(&"192.168.0.1".parse().unwrap()));
assert!(is_blocked_ip(&"192.168.255.255".parse().unwrap()));
}
#[test]
fn ssrf_blocks_link_local_metadata() {
assert!(is_blocked_ip(&"169.254.0.1".parse().unwrap()));
assert!(is_blocked_ip(&"169.254.169.254".parse().unwrap()));
}
#[test]
fn ssrf_blocks_cgnat() {
assert!(is_blocked_ip(&"100.64.0.1".parse().unwrap()));
assert!(is_blocked_ip(&"100.127.255.255".parse().unwrap()));
assert!(!is_blocked_ip(&"100.63.0.1".parse().unwrap()));
}
#[test]
fn ssrf_blocks_unspecified() {
assert!(is_blocked_ip(&"0.0.0.0".parse().unwrap()));
assert!(is_blocked_ip(&"::".parse().unwrap()));
}
#[test]
fn ssrf_allows_loopback() {
assert!(!is_blocked_ip(&"127.0.0.1".parse().unwrap()));
assert!(!is_blocked_ip(&"::1".parse().unwrap()));
}
#[test]
fn ssrf_allows_public_ips() {
assert!(!is_blocked_ip(&"1.1.1.1".parse().unwrap()));
assert!(!is_blocked_ip(&"8.8.8.8".parse().unwrap()));
}
#[test]
fn ssrf_blocks_ipv6_link_local() {
assert!(is_blocked_ip(&"fe80::1".parse().unwrap()));
}
#[test]
fn ssrf_blocks_ipv6_unique_local() {
assert!(is_blocked_ip(&"fc00::1".parse().unwrap()));
assert!(is_blocked_ip(&"fd00::1".parse().unwrap()));
}
#[test]
fn ssrf_blocks_ipv4_mapped_ipv6_private() {
assert!(is_blocked_ip(&"::ffff:10.0.0.1".parse::<IpAddr>().unwrap()));
assert!(is_blocked_ip(
&"::ffff:192.168.1.1".parse::<IpAddr>().unwrap()
));
fn ssrf_blocks_a_public_name_that_resolves_to_loopback() {
let ip: IpAddr = "127.0.0.1".parse().unwrap();
assert!(is_blocked_ip(&ip, "evil.example.com"));
}
// SSRF protection: validate_hook_url tests
@@ -839,14 +762,15 @@ mod tests {
/// the secret does NOT appear in the returned error message.
#[tokio::test]
async fn run_http_hook_scrubs_url_from_reqwest_error() {
// Use a TEST-NET-1 host (RFC 5737, "MUST NOT be used in
// public networks"). It is not RFC1918 so SSRF validation
// will let it through, but no real DNS or connection will
// succeed -- reqwest will surface a connection error whose
// default Display includes the URL.
// A closed loopback port: allowed as a literal local host,
// and refused at once. TEST-NET-1 is now blocked by policy.
let dead = {
let l = std::net::TcpListener::bind("127.0.0.1:0").unwrap();
l.local_addr().unwrap()
};
let secret = "ghp_VERY_REAL_SECRET_TOKEN_42";
let mut extra_env = std::collections::HashMap::new();
extra_env.insert("RUNTIME_HOST".to_string(), "192.0.2.1".to_string());
extra_env.insert("RUNTIME_HOST".to_string(), dead.to_string());
extra_env.insert("MY_TOKEN".to_string(), secret.to_string());
let raw = "https://${RUNTIME_HOST}/check?token=${MY_TOKEN}";
@@ -928,10 +852,7 @@ mod tests {
// debugging). The wire-DTO consumer must prefer raw_url for
// display -- documented in the HttpInfo rustdoc.
let info = info.expect("HttpInfo should be present for connection failures too");
assert_eq!(
info.url,
"https://192.0.2.1/check?token=ghp_VERY_REAL_SECRET_TOKEN_42"
);
assert_eq!(info.url, format!("https://{dead}/check?token={secret}"));
assert_eq!(info.raw_url.as_deref(), Some(raw));
}
crates/codegen/kigi-hunk-tracker/src/actor/file_utils.rs
+1
View File
@@ -318,6 +318,7 @@ mod tests {
assert!(matches!(state, FileContentState::TooLarge { .. }));
}
#[cfg(unix)]
#[tokio::test]
async fn test_read_file_bounded_symlink() {
let dir = tempfile::tempdir().unwrap();
crates/codegen/kigi-memory/src/backend.rs
+203 -77
View File
@@ -18,6 +18,76 @@ use super::embedding::EmbeddingProvider as _;
use super::storage::MemoryStorage;
use super::watcher::MemoryFileWatcher;
/// The session's embedding credentials, bound to the one endpoint they may
/// reach. Only [`Self::for_endpoint`] retains a live handle; the default fails
/// closed.
///
/// `embed_base_url` is the CURRENT MODEL's endpoint, so a session on a BYOK or
/// subscription-OAuth model aims memory embeddings at that provider's host —
/// and [`kigi_auth::AuthRetryMiddleware`] stamps `Authorization` on every
/// request it wraps, with no idea where the request is going. The caller that
/// owns the credential rule decides `trusted` once, here; a platform's own
/// `embed_api_key` is unaffected and keeps serving its own endpoint.
#[derive(Clone, Default)]
pub struct EndpointScopedCredentials {
endpoint: Option<reqwest::Url>,
auth_credentials: Option<Arc<dyn kigi_auth::AuthCredentialProvider>>,
}
// Redacts the credential handles; only their presence is printable.
impl std::fmt::Debug for EndpointScopedCredentials {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_struct("EndpointScopedCredentials")
.field("endpoint", &self.endpoint)
.field("has_auth_credentials", &self.auth_credentials.is_some())
.finish()
}
}
impl EndpointScopedCredentials {
/// No session credential may ride — the state every non-session caller wants.
pub fn none() -> Self {
Self::default()
}
/// Retains the handles only for a `trusted`, parsable `endpoint`.
pub fn for_endpoint(
endpoint: &str,
trusted: bool,
auth_credentials: Option<Arc<dyn kigi_auth::AuthCredentialProvider>>,
) -> Self {
if trusted && let Ok(url) = reqwest::Url::parse(endpoint) {
return Self {
endpoint: Some(url),
auth_credentials,
};
}
if auth_credentials.is_some() {
tracing::info!(
target: kigi_log::memory_log::TARGET,
endpoint,
"memory embeddings: session credentials withheld from this endpoint; \
its own key, if any, still applies"
);
}
Self::none()
}
pub fn is_empty(&self) -> bool {
self.auth_credentials.is_none()
}
/// Enforced at request-build time, in release too: [`MemoryBackendParams`]
/// is `Clone` and callers rewrite fields on the copy, so construction-time
/// scoping alone would not survive a rewritten `embed_base_url`.
fn approved_for(&self, base_url: &str) -> bool {
match &self.endpoint {
None => self.is_empty(),
Some(endpoint) => reqwest::Url::parse(base_url).is_ok_and(|url| &url == endpoint),
}
}
}
/// All configuration needed to build a fully-wired [`MemoryBackendImpl`] for a live session.
///
/// Grouping these in one struct ensures every call site — ToolBridge, first-turn
@@ -30,7 +100,8 @@ pub struct MemoryBackendParams {
pub session_id: String,
/// Embedding provider config — `None` forces FTS-only fallback everywhere.
pub embed_config: Option<kigi_config_types::MemoryEmbeddingConfig>,
/// Base URL for embedding API calls (CLI proxy).
/// Base URL for embedding API calls (CLI proxy). Must match the endpoint
/// `embedding_credentials` was scoped to; a mismatch fails closed.
pub embed_base_url: String,
/// API key for embedding API calls.
pub embed_api_key: Option<String>,
@@ -47,10 +118,8 @@ pub struct MemoryBackendParams {
/// - `"injection"` — first-turn memory context injection
/// - `"compaction_recovery"` — post-compaction context re-injection
pub search_source: &'static str,
/// Dynamic API key provider — when set, `make_embedding_provider()` resolves
/// the key per-call instead of using the static `embed_api_key`.
pub api_key_provider: Option<kigi_tools::types::SharedApiKeyProvider>,
pub auth_credentials: Option<Arc<dyn kigi_auth::AuthCredentialProvider>>,
/// The session credentials, and the single endpoint they may reach.
pub embedding_credentials: EndpointScopedCredentials,
}
impl MemoryBackendParams {
@@ -59,8 +128,7 @@ impl MemoryBackendParams {
pub async fn make_embedding_provider(&self) -> Option<super::embedding::ApiEmbeddingProvider> {
build_embedding_provider(
self.embed_config.as_ref(),
self.auth_credentials.as_ref(),
self.api_key_provider.as_ref(),
&self.embedding_credentials,
self.embed_api_key.as_deref(),
&self.embed_base_url,
)
@@ -70,8 +138,7 @@ impl MemoryBackendParams {
async fn build_embedding_provider(
config: Option<&kigi_config_types::MemoryEmbeddingConfig>,
auth_credentials: Option<&Arc<dyn kigi_auth::AuthCredentialProvider>>,
api_key_provider: Option<&kigi_tools::types::SharedApiKeyProvider>,
credentials: &EndpointScopedCredentials,
static_api_key: Option<&str>,
base_url: &str,
) -> Option<super::embedding::ApiEmbeddingProvider> {
@@ -80,9 +147,19 @@ async fn build_embedding_provider(
return None;
}
let approved = credentials.approved_for(base_url);
if !approved {
tracing::error!(
target: kigi_log::memory_log::TARGET,
base_url,
approved_endpoint = ?credentials.endpoint,
"memory embeddings: scoped credentials do not match the request URL; dropping them"
);
}
// Prefer the refresh-capable credential provider — the middleware gives
// 401 retry for free without any per-call key resolution.
if let Some(creds) = auth_credentials {
if approved && let Some(creds) = credentials.auth_credentials.as_ref() {
let client = super::embedding::build_middleware_client(creds.clone());
return super::embedding::ApiEmbeddingProvider::from_config(
config,
@@ -91,14 +168,13 @@ async fn build_embedding_provider(
);
}
// Fallback: resolve API key per-call, wrap in a static middleware client
// (no 401 refresh, but auth header is still stamped by middleware).
let api_key = match api_key_provider {
Some(p) => p.current_api_key_async().await,
None => None,
}
.or_else(|| static_api_key.map(|s| s.to_owned()))?;
super::embedding::ApiEmbeddingProvider::from_session(config, base_url.to_owned(), api_key)
// The platform's own configured key, wrapped in a static middleware client
// (no 401 refresh, but the auth header is still stamped by middleware).
super::embedding::ApiEmbeddingProvider::from_session(
config,
base_url.to_owned(),
static_api_key?.to_owned(),
)
}
/// `MemoryBackend` implementation backed by hybrid search (FTS5 + vector KNN).
@@ -129,10 +205,8 @@ pub struct MemoryBackendImpl {
/// Only the ToolBridge backend's counter is shared back to the session actor;
/// injection and compaction-recovery backends use their own local counters.
pub search_counter: std::sync::Arc<std::sync::atomic::AtomicU64>,
/// Dynamic API key provider for embedding requests.
api_key_provider: Option<kigi_tools::types::SharedApiKeyProvider>,
/// Refresh-capable credential provider for embedding HTTP middleware.
auth_credentials: Option<Arc<dyn kigi_auth::AuthCredentialProvider>>,
/// The session credentials, and the single endpoint they may reach.
embedding_credentials: EndpointScopedCredentials,
}
impl MemoryBackendImpl {
@@ -150,8 +224,7 @@ impl MemoryBackendImpl {
stale_claim_secs: 60,
session_id: String::new(),
search_source: "tool",
api_key_provider: None,
auth_credentials: None,
embedding_credentials: EndpointScopedCredentials::none(),
search_counter: std::sync::Arc::new(std::sync::atomic::AtomicU64::new(0)),
}
}
@@ -200,8 +273,7 @@ impl MemoryBackendImpl {
async fn make_embedding_provider(&self) -> Option<super::embedding::ApiEmbeddingProvider> {
build_embedding_provider(
self.embed_config.as_ref(),
self.auth_credentials.as_ref(),
self.api_key_provider.as_ref(),
&self.embedding_credentials,
self.embed_api_key.as_deref(),
&self.embed_base_url,
)
@@ -232,8 +304,7 @@ impl MemoryBackendImpl {
if let Some(w) = &params.watcher {
backend = backend.with_watcher(w.clone(), params.stale_claim_secs);
}
backend.api_key_provider = params.api_key_provider.clone();
backend.auth_credentials = params.auth_credentials.clone();
backend.embedding_credentials = params.embedding_credentials.clone();
backend
}
}
@@ -479,8 +550,7 @@ mod factory_tests {
watcher: None,
stale_claim_secs: 60,
search_source: "tool",
api_key_provider: None,
auth_credentials: None,
embedding_credentials: EndpointScopedCredentials::none(),
}
}
@@ -1038,72 +1108,128 @@ mod factory_tests {
);
}
/// Regression: provider build must use `current_api_key_async`,
/// never sync. Prevents memory_search 401s on rotated tokens.
#[tokio::test]
async fn make_embedding_provider_uses_async_api_key_resolution() {
use kigi_tools::types::ApiKeyProvider;
use std::sync::atomic::{AtomicU32, Ordering};
struct AsyncProbe {
sync_calls: Arc<AtomicU32>,
async_calls: Arc<AtomicU32>,
struct ProbeCredentials;
impl kigi_auth::HttpAuth for ProbeCredentials {
fn apply(
&self,
builder: reqwest::RequestBuilder,
_base_url: &str,
) -> reqwest::RequestBuilder {
builder.bearer_auth("session-bearer")
}
impl ApiKeyProvider for AsyncProbe {
fn current_api_key(&self) -> Option<String> {
self.sync_calls.fetch_add(1, Ordering::SeqCst);
Some("sync-stale".into())
}
fn current_api_key_async(
&self,
) -> std::pin::Pin<Box<dyn std::future::Future<Output = Option<String>> + Send + '_>>
{
let counter = self.async_calls.clone();
Box::pin(async move {
counter.fetch_add(1, Ordering::SeqCst);
Some("async-fresh".into())
})
}
#[async_trait::async_trait]
impl kigi_auth::AuthCredentialProvider for ProbeCredentials {
fn snapshot(&self) -> kigi_auth::CredentialSnapshot {
kigi_auth::CredentialSnapshot {
token: Some("session-bearer".into()),
..Default::default()
}
}
async fn refresh_after_unauthorized(&self) -> bool {
false
}
}
let sync_calls = Arc::new(AtomicU32::new(0));
let async_calls = Arc::new(AtomicU32::new(0));
let probe: kigi_tools::types::SharedApiKeyProvider = Arc::new(AsyncProbe {
sync_calls: sync_calls.clone(),
async_calls: async_calls.clone(),
});
const SESSION_ENDPOINT: &str = "https://api.kimi.com/coding/v1";
const FOREIGN_ENDPOINT: &str = "https://api.anthropic.com/v1";
let params = MemoryBackendParams {
fn params_at(base_url: &str, credentials: EndpointScopedCredentials) -> MemoryBackendParams {
MemoryBackendParams {
session_id: "s1".into(),
embed_config: Some(MemoryEmbeddingConfig {
model: Some("test-embed-model".into()),
..Default::default()
}),
embed_base_url: "http://example/v1".into(),
embed_api_key: Some("static-fallback".into()),
embed_base_url: base_url.into(),
embed_api_key: None,
search_config: MemorySearchConfig::default(),
watcher: None,
stale_claim_secs: 60,
search_source: "tool",
api_key_provider: Some(probe),
// No auth_credentials — forces the api_key_provider fallback path.
auth_credentials: None,
};
embedding_credentials: credentials,
}
}
let provider = params.make_embedding_provider().await;
/// The session bearer must never ride to a third-party embedding host.
///
/// `embed_base_url` is the CURRENT MODEL's endpoint, so a session on a BYOK
/// or subscription-OAuth model aims memory embeddings at that provider —
/// and the auth middleware stamps `Authorization` unconditionally.
#[tokio::test]
async fn session_credentials_are_withheld_from_a_foreign_endpoint() {
let scoped = EndpointScopedCredentials::for_endpoint(
FOREIGN_ENDPOINT,
false,
Some(Arc::new(ProbeCredentials)),
);
assert!(
scoped.is_empty(),
"an untrusted endpoint must drop both handles"
);
let provider = params_at(FOREIGN_ENDPOINT, scoped)
.make_embedding_provider()
.await;
assert!(
provider.is_none(),
"no credential may ride to a foreign endpoint, and there is no static key to fall back to"
);
}
/// The trusted-endpoint path still builds a credentialed provider.
#[tokio::test]
async fn session_credentials_ride_their_own_endpoint() {
let scoped = EndpointScopedCredentials::for_endpoint(
SESSION_ENDPOINT,
true,
Some(Arc::new(ProbeCredentials)),
);
assert!(
!scoped.is_empty(),
"a trusted endpoint must retain the handles"
);
let provider = params_at(SESSION_ENDPOINT, scoped)
.make_embedding_provider()
.await;
assert!(
provider.is_some(),
"provider must be built when model is set"
"the session endpoint keeps its refresh-capable credential"
);
assert_eq!(
async_calls.load(Ordering::SeqCst),
1,
"must call current_api_key_async exactly once per provider build"
}
/// The runtime re-check, not construction alone, is what guards the wire:
/// `MemoryBackendParams` is `Clone` and callers rewrite fields on the copy.
#[tokio::test]
async fn a_cloned_param_set_cannot_redirect_scoped_credentials() {
let scoped = EndpointScopedCredentials::for_endpoint(
SESSION_ENDPOINT,
true,
Some(Arc::new(ProbeCredentials)),
);
assert_eq!(
sync_calls.load(Ordering::SeqCst),
0,
"sync current_api_key must NOT be called — the async path is the contract"
let redirected = MemoryBackendParams {
embed_base_url: FOREIGN_ENDPOINT.into(),
..params_at(SESSION_ENDPOINT, scoped)
};
assert!(
redirected.make_embedding_provider().await.is_none(),
"credentials scoped to one endpoint must not follow a rewritten base_url"
);
}
/// A platform's own API key is not a session credential: it is resolved for
/// that platform and must keep serving that platform's endpoint.
#[tokio::test]
async fn a_platform_api_key_still_serves_its_own_endpoint() {
let params = MemoryBackendParams {
embed_api_key: Some("platform-key".into()),
..params_at(FOREIGN_ENDPOINT, EndpointScopedCredentials::none())
};
assert!(
params.make_embedding_provider().await.is_some(),
"withholding the session credential must not disable BYOK embeddings"
);
}
}
crates/codegen/kigi-memory/src/lib.rs
+1 -1
View File
@@ -31,7 +31,7 @@ pub mod storage;
pub mod text_utils;
pub mod watcher;
pub use backend::{MemoryBackendImpl, MemoryBackendParams};
pub use backend::{EndpointScopedCredentials, MemoryBackendImpl, MemoryBackendParams};
pub use index::{MemoryIndex, init_sqlite_vec};
pub use storage::{MemoryScope, MemoryStorage};
crates/codegen/kigi-mermaid/src/mmdc.rs
+1
View File
@@ -150,6 +150,7 @@ fn write_private(path: &Path, contents: &str) -> std::io::Result<()> {
#[cfg(test)]
mod tests {
use super::*;
#[cfg(unix)]
use std::time::Instant;
/// Write an executable `#!/bin/sh` fake `mmdc` and return (dir-guard, path).
crates/codegen/kigi-mermaid/src/subprocess.rs
+3
View File
@@ -179,9 +179,12 @@ fn reap_process_group(_child: &Child) {
#[cfg(test)]
mod tests {
use super::*;
#[cfg(unix)]
use std::process::Stdio;
#[cfg(unix)]
use std::time::Instant;
#[cfg(unix)]
fn detached(mut cmd: Command) -> Command {
cmd.stdin(Stdio::null())
.stdout(Stdio::null())
crates/codegen/kigi-models/src/lib.rs
+6 -6
View File
@@ -1574,7 +1574,7 @@ impl PlatformId {
/// `models_cache.json`, the `visibility=="list"` AND `supported_in_api==true`
/// set) because OpenAI exposes no stable public models endpoint for the
/// ChatGPT Codex backend. Each entry carries context window + per-model
/// selectable reasoning efforts (incl. the codex-only `xhigh`/`max`/`ultra`
/// selectable reasoning efforts (incl. the codex-only `xhigh`/`max`
/// tiers), so the fetch path maps them through the SAME
/// `platform_wire_model_to_entry` output as a live listing — no new type.
pub fn hardcoded_catalog(self) -> Option<Vec<WireModel>> {
@@ -1617,13 +1617,13 @@ fn openai_codex_wire_models() -> Vec<WireModel> {
codex_wire_model(
"gpt-5.6-sol",
"GPT-5.6-Sol",
&["low", "medium", "high", "xhigh", "max", "ultra"],
&["low", "medium", "high", "xhigh", "max"],
"low",
),
codex_wire_model(
"gpt-5.6-terra",
"GPT-5.6-Terra",
&["low", "medium", "high", "xhigh", "max", "ultra"],
&["low", "medium", "high", "xhigh", "max"],
"medium",
),
codex_wire_model(
@@ -2655,7 +2655,7 @@ mod tests {
/// The HARDCODED openai-codex catalog is exactly the 4 supported+listed
/// models, keyed by slug, ctx 272000, each exposing its exact supported
/// efforts (incl. the codex-only `xhigh`/`max`/`ultra` tiers). The
/// efforts (incl. the codex-only `xhigh`/`max` tiers). The
/// list-but-broken / hidden models are absent. Every other platform serves
/// NO hardcoded catalog (its models come from the live wire).
#[test]
@@ -2702,11 +2702,11 @@ mod tests {
};
assert_eq!(
efforts("gpt-5.6-sol"),
["low", "medium", "high", "xhigh", "max", "ultra"]
["low", "medium", "high", "xhigh", "max"]
);
assert_eq!(
efforts("gpt-5.6-terra"),
["low", "medium", "high", "xhigh", "max", "ultra"]
["low", "medium", "high", "xhigh", "max"]
);
assert_eq!(
efforts("gpt-5.6-luna"),
crates/codegen/kigi-pager-minimal/src/commit.rs
+16 -11
View File
@@ -1251,6 +1251,10 @@ mod tests {
use ratatui::layout::Rect;
use similar::ChangeTag;
// A sibling test engages the terminal-native lock, which flattens
// every diff background process-wide. Hold the shared theme lock.
let _theme = kigi_tui::theme::cache::pin_theme();
let hunk = vec![
DiffLine {
text: "let x = 1;\n".into(),
@@ -1293,23 +1297,24 @@ mod tests {
// The committed edit uses a flat background (terminal transparency), but
// must still paint the per-line diff backgrounds — otherwise an added /
// removed line is indistinguishable from context.
let mut saw_insert = false;
let mut saw_delete = false;
//
// Theme-agnostic: the line cache is keyed on the global theme,
// so an exact-RGB match would race. Colors pinned in tool::edit.
let mut bands = std::collections::BTreeSet::new();
for y in 0..h {
for x in 0..width {
if let Some(cell) = buf.cell((x, y)) {
saw_insert |= cell.bg == theme.diff_insert_bg;
saw_delete |= cell.bg == theme.diff_delete_bg;
if let Some(cell) = buf.cell((x, y))
&& cell.bg != ratatui::style::Color::Reset
{
bands.insert(format!("{:?}", cell.bg));
}
}
}
assert!(
saw_insert,
"committed edit lost the insert (green) diff background"
);
assert!(
saw_delete,
"committed edit lost the delete (red) diff background"
bands.len() >= 2,
"committed edit lost its insert/delete diff bands: \
theme={:?} h={h} bands={bands:?}",
Theme::current_kind(),
);
}
crates/codegen/kigi-pager-render/src/link_opener.rs
+1
View File
@@ -214,6 +214,7 @@ pub fn open_url_if_safe(url: &str, filter: SchemeFilter) -> bool {
mod tests {
use super::*;
#[cfg(not(target_os = "windows"))]
#[test]
fn open_path_command_passes_path_as_a_single_arg() {
// Path with spaces must be one argument, never shell-interpolated.
crates/codegen/kigi-pager-render/src/render/draw.rs
+132 -9
View File
@@ -146,6 +146,10 @@ impl TermWriter {
pub fn writer_sync(&self) -> &WriterSync {
&self.sync
}
/// A cloneable handle onto this writer's queue for out-of-band escapes.
pub fn escape_writer(&self) -> EscapeWriter {
EscapeWriter::new(self.tx.clone(), self.sync.clone())
}
}
impl Write for TermWriter {
fn write(&mut self, data: &[u8]) -> std::io::Result<usize> {
@@ -166,6 +170,56 @@ impl Drop for TermWriter {
let _ = self.flush();
}
}
/// Out-of-band escapes ride the frame queue (an inline stderr write deadlocks on a stalled tty); drop every clone before `join_within`.
#[derive(Clone)]
pub struct EscapeWriter {
tx: mpsc::Sender<Vec<u8>>,
sync: WriterSync,
}
impl EscapeWriter {
pub fn new(tx: mpsc::Sender<Vec<u8>>, sync: WriterSync) -> Self {
Self { tx, sync }
}
/// No writer thread behind it: sends are dropped. For tests and views built before the terminal exists.
pub fn disconnected() -> Self {
let (tx, _rx) = mpsc::channel::<Vec<u8>>();
Self {
tx,
sync: WriterSync::new(),
}
}
/// Never blocks; covered by [`WriterSync::wait_drained`].
pub fn emit(&self, bytes: impl Into<Vec<u8>>) {
let data: Vec<u8> = bytes.into();
if data.is_empty() {
return;
}
self.sync.mark_queued();
let _ = self.tx.send(data);
}
/// Winapi-only commands run synchronously: a console API call, not a tty write.
pub fn emit_command(&self, command: impl crossterm::Command) {
#[cfg(windows)]
if !command.is_ansi_code_supported() {
let _ = command.execute_winapi();
return;
}
let mut ansi = String::new();
if command.write_ansi(&mut ansi).is_ok() {
self.emit(ansi);
}
}
}
/// Outcome of [`WriterThread::join_within`].
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum WriterJoin {
/// The thread drained its queue and exited.
Joined,
/// Still running at the deadline (tty blocked, or a sender alive); detached.
TimedOut,
}
/// Teardown waits this long for the writer thread before detaching it, so a stalled tty cannot hang `/quit`.
pub const WRITER_JOIN_GRACE: Duration = Duration::from_secs(2);
/// Handle for the background writer thread.
///
/// Joining ensures all queued frames have been written to the terminal
@@ -174,19 +228,45 @@ pub struct WriterThread {
handle: Option<std::thread::JoinHandle<()>>,
}
impl WriterThread {
/// Block until the writer thread has processed all pending frames and
/// exited. The [`mpsc::Sender`] must be dropped *before* calling this,
/// otherwise the thread will never see the channel close.
pub fn join(mut self) {
if let Some(h) = self.handle.take() {
let _ = h.join();
/// Wait up to `grace` for the thread to drain and exit, else detach it; every sender (`EscapeWriter` clones too) must be dropped first.
pub fn join_within(mut self, grace: Duration) -> WriterJoin {
let Some(handle) = self.handle.take() else {
return WriterJoin::Joined;
};
if wait_finished(&handle, grace) {
let _ = handle.join();
WriterJoin::Joined
} else {
tracing::warn!(
grace_ms = grace.as_millis() as u64,
"term-writer thread still running at teardown; detaching"
);
WriterJoin::TimedOut
}
}
#[cfg(test)]
fn for_test(handle: std::thread::JoinHandle<()>) -> Self {
Self {
handle: Some(handle),
}
}
}
fn wait_finished(handle: &std::thread::JoinHandle<()>, grace: Duration) -> bool {
let deadline = Instant::now() + grace;
while !handle.is_finished() {
if Instant::now() >= deadline {
return false;
}
std::thread::sleep(Duration::from_millis(1));
}
true
}
impl Drop for WriterThread {
fn drop(&mut self) {
if let Some(h) = self.handle.take() {
let _ = h.join();
if let Some(handle) = self.handle.take()
&& wait_finished(&handle, WRITER_JOIN_GRACE)
{
let _ = handle.join();
}
}
}
@@ -197,7 +277,7 @@ impl Drop for WriterThread {
/// `BufWriter`. The [`WriterSync`] must be shared with every [`TermWriter`]
/// built on the sender so [`WriterSync::wait_drained`] tracks the queue.
/// Drop the sender to signal the thread to exit, then call
/// [`WriterThread::join`] to wait for it.
/// [`WriterThread::join_within`] to wait for it.
pub fn spawn_writer_thread() -> (mpsc::Sender<Vec<u8>>, WriterSync, WriterThread) {
let (tx, rx) = mpsc::channel::<Vec<u8>>();
let sync = WriterSync::new();
@@ -570,4 +650,47 @@ mod tests {
"Round-tripped bytes do not decode to original UTF-8 string"
);
}
/// Frames and escapes share one queue, so the writer sees them in call order.
#[test]
#[cfg(not(windows))]
fn escapes_share_the_frame_queue_in_call_order() {
let (tx, rx) = mpsc::channel::<Vec<u8>>();
let sync = WriterSync::new();
let mut frames = TermWriter::new(tx, sync.clone());
let escapes = frames.escape_writer();
frames.write_all(b"frame-1").unwrap();
frames.flush().unwrap();
escapes.emit("\x1b[?1000h");
frames.write_all(b"frame-2").unwrap();
frames.flush().unwrap();
escapes.emit_command(crossterm::event::DisableMouseCapture);
let payloads: Vec<Vec<u8>> = std::iter::from_fn(|| rx.try_recv().ok()).collect();
assert_eq!(payloads.len(), 4);
assert_eq!(payloads[0], b"frame-1");
assert_eq!(payloads[1], b"\x1b[?1000h");
assert_eq!(payloads[2], b"frame-2");
assert!(payloads[3].starts_with(b"\x1b["), "{:?}", payloads[3]);
assert!(!sync.is_drained(), "four payloads queued, none written");
}
/// A writer thread parked in a blocked tty write must not hang teardown.
#[test]
fn join_within_times_out_on_a_stuck_writer_thread() {
let (release_tx, release_rx) = mpsc::channel::<()>();
let handle = std::thread::spawn(move || {
let _ = release_rx.recv();
});
let started = Instant::now();
let outcome = WriterThread::for_test(handle).join_within(Duration::from_millis(50));
assert_eq!(outcome, WriterJoin::TimedOut);
assert!(started.elapsed() < Duration::from_secs(2));
let _ = release_tx.send(());
}
#[test]
fn join_within_joins_a_finished_writer_thread() {
let handle = std::thread::spawn(|| {});
assert_eq!(
WriterThread::for_test(handle).join_within(Duration::from_secs(5)),
WriterJoin::Joined
);
}
}
crates/codegen/kigi-pager-render/src/terminal/probe.rs
+1
View File
@@ -10,6 +10,7 @@
//! portable re-injection exists (TIOCSTI is blocked); accepted loss.
use std::io::Write;
#[cfg(unix)]
use std::time::Duration;
/// Bounds the reply buffer against terminals that stream without a terminator.
crates/codegen/kigi-sampler/Cargo.toml
+1
View File
@@ -18,6 +18,7 @@ eventsource-stream = { workspace = true }
futures-util = { workspace = true }
indexmap = { workspace = true, features = ["serde"] }
reqwest = { workspace = true, features = ["stream"] }
kigi-secrets = { path = "../kigi-secrets" }
serde = { workspace = true, features = ["derive"] }
serde_json = { workspace = true }
tokio = { workspace = true, features = ["rt", "macros", "time", "sync"] }
crates/codegen/kigi-sampler/src/client.rs
+49 -6
View File
@@ -193,6 +193,16 @@ fn extract_context_total(value: &serde_json::Value) -> Option<u32> {
Some(i.saturating_add(o))
}
/// The endpoint as the HTTP client parses it, minus userinfo, sensitive query values and fragment; an unparseable one is not echoed.
fn redacted_endpoint(endpoint: &str) -> String {
match reqwest::Url::parse(endpoint) {
Ok(mut url) => {
kigi_secrets::redact_url(&mut url);
url.to_string()
}
Err(_) => "<unparseable endpoint>".to_owned(),
}
}
/// Record `success=false` + `error` on the active inference span when a stream
/// request fails before any response (transport/connect/TLS errors). Without
/// this the `#[instrument]` span closes with both fields Empty, so an outage
@@ -806,7 +816,7 @@ impl SamplingClient {
let server_message_lower = server_message.to_lowercase();
let mut context_parts = vec![server_message.to_string()];
context_parts.push(format!("\nRequest URL: {}", endpoint));
context_parts.push(format!("\nRequest URL: {}", redacted_endpoint(endpoint)));
// Show headers if error mentions headers
if server_message_lower.contains("header") {
@@ -1006,7 +1016,8 @@ impl SamplingClient {
let endpoint = self.endpoint("chat/completions");
let server_message = response.text().await.unwrap_or_default();
return Err(SamplingError::Auth(format!(
"Unauthorized (401) from {endpoint}: {server_message}"
"Unauthorized (401) from {}: {server_message}",
redacted_endpoint(&endpoint)
)));
}
@@ -1196,7 +1207,8 @@ impl SamplingClient {
let endpoint = self.endpoint("responses");
let server_message = parse_error_bytes(bytes.as_ref());
return Err(SamplingError::Auth(format!(
"Unauthorized (401) from {endpoint}: {server_message}"
"Unauthorized (401) from {}: {server_message}",
redacted_endpoint(&endpoint)
)));
}
@@ -1363,7 +1375,8 @@ impl SamplingClient {
let endpoint = self.endpoint("responses");
let server_message = response.text().await.unwrap_or_default();
return Err(SamplingError::Auth(format!(
"Unauthorized (401) from {endpoint}: {server_message}"
"Unauthorized (401) from {}: {server_message}",
redacted_endpoint(&endpoint)
)));
}
let model_metadata = extract_model_metadata(response.headers());
@@ -1542,7 +1555,8 @@ impl SamplingClient {
let endpoint = self.endpoint("messages");
let server_message = parse_error_bytes(bytes.as_ref());
return Err(SamplingError::Auth(format!(
"Unauthorized (401) from {endpoint}: {server_message}"
"Unauthorized (401) from {}: {server_message}",
redacted_endpoint(&endpoint)
)));
}
@@ -1657,7 +1671,8 @@ impl SamplingClient {
let endpoint = self.endpoint("messages");
let server_message = response.text().await.unwrap_or_default();
return Err(SamplingError::Auth(format!(
"Unauthorized (401) from {endpoint}: {server_message}"
"Unauthorized (401) from {}: {server_message}",
redacted_endpoint(&endpoint)
)));
}
let model_metadata = extract_model_metadata(response.headers());
@@ -3048,3 +3063,31 @@ mod tests {
));
}
}
#[cfg(test)]
mod redacted_endpoint_tests {
/// Every spelling the client accepts (LF inside, one slash, no slash) loses its userinfo.
#[test]
fn redacted_endpoint_drops_userinfo_in_every_accepted_spelling() {
for spelling in [
"http://u:TEST_PASSWORD@127.0.0.1:1/v1/chat/completions",
"http://u:TEST PASSWORD@127.0.0.1:1/v1/chat/completions",
"http://u:TEST_PASSWORD\n@127.0.0.1:1/v1/chat/completions",
"http://u:TEST_PASSWORD\r\n@127.0.0.1:1/v1/chat/completions",
"http:/u:TEST_PASSWORD@127.0.0.1:1/v1/chat/completions",
"http:u:TEST_PASSWORD@127.0.0.1:1/v1/chat/completions",
"http:\\u:TEST_PASSWORD@127.0.0.1:1/v1/chat/completions",
] {
let redacted = super::redacted_endpoint(spelling);
assert_eq!(
redacted, "http://127.0.0.1:1/v1/chat/completions",
"{spelling:?}"
);
}
let gateway = super::redacted_endpoint(
"http://127.0.0.1:1/?access_token=TEST_QUERY_TOKEN&route=x#frag",
);
assert!(!gateway.contains("TEST_QUERY_TOKEN"), "{gateway}");
assert!(gateway.contains("route=x"), "{gateway}");
}
}
crates/codegen/kigi-sampler/src/retry.rs
+47 -7
View File
@@ -5,8 +5,11 @@
//!
//! # Retry behavior summary
//!
//! **Retried** (up to [`DEFAULT_MAX_RETRIES`] = 15, ~6 min with 30s backoff cap):
//! - 500, 502, 503, 504, 520 (server errors)
//! **Retried** (up to [`DEFAULT_MAX_RETRIES`] = 15, ~6 min with the
//! [`MAX_RETRY_BACKOFF`] cap — a server `Retry-After` is clamped to it too):
//! - Any 5xx except Cloudflare 525/526 — covers the Cloudflare edge pages
//! (520-524 origin unreachable/timed out, 530 edge 1xxx) and upstream
//! overload (529); see `is_retryable_api_status`
//! - Connection errors (timeout, refused, reset)
//! - `EventStreamError` / `StreamError` (mid-stream failures)
//! - `EmptyResponse` (model returned no content/tool calls)
@@ -77,16 +80,35 @@ pub fn doom_loop_backoff(retry_count: u32) -> Duration {
Duration::from_millis(hasher.finish() % 251)
}
/// Exponential backoff (2s, 4s, 8s, ..., capped 30s) with +/-20% jitter
/// to prevent thundering-herd retry storms.
/// Longest single wait on the generic retry path — the exponential-backoff
/// ceiling, and the clamp for a server `Retry-After`. Cloudflare answers 52x
/// with `Retry-After: 60`-`120`; honoring that verbatim across the retry
/// budget would stall a turn ~28 min instead of the ~6 min budget. The 429
/// path deliberately waits the full `Retry-After` instead, bounded by
/// [`RATE_LIMIT_RETRY_THRESHOLD`] attempts (and by the parse-level 120s cap
/// on the header).
pub const MAX_RETRY_BACKOFF: Duration = Duration::from_secs(30);
/// Exponential backoff (2s, 4s, 8s, ..., capped at [`MAX_RETRY_BACKOFF`])
/// with +/-20% jitter to prevent thundering-herd retry storms.
pub fn retry_backoff_with_jitter(retry_count: u32) -> Duration {
let shift = retry_count.saturating_sub(1);
let base_ms = 2000u64
.checked_shl(shift)
.unwrap_or(u64::MAX)
.min(MAX_RETRY_BACKOFF.as_millis() as u64);
jittered(Duration::from_millis(base_ms))
}
/// +/-20% jitter around `base`, de-syncing clients that failed at the
/// same instant (e.g. a mass Cloudflare 52x event during an origin outage).
fn jittered(base: Duration) -> Duration {
use std::hash::{Hash, Hasher};
use std::sync::atomic::{AtomicU64, Ordering};
static JITTER_SEQ: AtomicU64 = AtomicU64::new(0);
let shift = retry_count.saturating_sub(1);
let base_ms = 2000u64.checked_shl(shift).unwrap_or(u64::MAX).min(30_000);
let base_ms = base.as_millis() as u64;
let jitter_range = base_ms / 5;
let mut hasher = std::hash::DefaultHasher::new();
JITTER_SEQ.fetch_add(1, Ordering::Relaxed).hash(&mut hasher);
@@ -210,9 +232,11 @@ pub fn classify_error(
if next_attempt >= max_retries {
return RetryDecision::Fatal(clone_error(err));
}
// A server `Retry-After` is clamped: Cloudflare 52x pages ask for
// 60-120s, which across the budget would stall a turn ~28 min.
let backoff = err
.retry_after()
.map(Duration::from_secs)
.map(|secs| jittered(Duration::from_secs(secs).min(MAX_RETRY_BACKOFF)))
.unwrap_or_else(|| retry_backoff_with_jitter(next_attempt));
if next_attempt == 1 {
return RetryDecision::RetryWithClientRebuild { backoff };
@@ -595,6 +619,22 @@ mod tests {
}
}
#[test]
fn classify_5xx_retry_after_is_clamped_to_backoff_cap() {
// Cloudflare answers 52x with Retry-After: 60-120; honoring it
// verbatim across the retry budget would stall a turn ~28 min.
let err = api_err_with_retry_after(StatusCode::from_u16(522).unwrap(), 120);
match classify_error(&err, 1, 15, RATE_LIMIT_RETRY_THRESHOLD) {
RetryDecision::Retry { backoff } => {
assert!(
backoff <= MAX_RETRY_BACKOFF.mul_f64(1.2),
"52x Retry-After must clamp to ~MAX_RETRY_BACKOFF, got {backoff:?}"
);
}
other => panic!("expected Retry, got {other:?}"),
}
}
#[test]
fn classify_5xx_first_retry_rebuilds_client() {
let err = api_err(StatusCode::INTERNAL_SERVER_ERROR, "boom");
crates/codegen/kigi-sampler/src/stream/messages.rs
+21 -5
View File
@@ -55,6 +55,15 @@ enum BlockType {
Thinking,
}
/// Executable after a `max_tokens` stop: every call's arguments parse or are empty (zero-arg convention).
fn tool_calls_complete(tool_calls: &[ToolCall]) -> bool {
!tool_calls.is_empty()
&& tool_calls.iter().all(|tc| {
tc.arguments.trim().is_empty()
|| serde_json::from_str::<serde::de::IgnoredAny>(&tc.arguments).is_ok()
})
}
/// Transform a raw Anthropic Messages API stream into a stream of
/// [`SamplingEvent`]s.
///
@@ -454,11 +463,18 @@ pub fn stream_messages<'a>(
}
if final_stop_reason == Some(StopReason::Length) {
yield SamplingEvent::Failed {
request_id: request_id.clone(),
error: SamplingErrorInfo::from(&SamplingError::MaxTokensTruncation),
};
return;
if !tool_calls_complete(&assistant_tool_calls) {
yield SamplingEvent::Failed {
request_id: request_id.clone(),
error: SamplingErrorInfo::from(&SamplingError::MaxTokensTruncation),
};
return;
}
tracing::info!(
request_id = %request_id,
tool_calls = assistant_tool_calls.len(),
"max_tokens stop after complete tool calls; completing the turn so they run"
);
}
// Build the final response
crates/codegen/kigi-sampler/src/stream/messages_tests.rs
+87
View File
@@ -653,3 +653,90 @@ async fn pure_cache_hit_with_zero_uncached_still_emits_usage() {
assert_eq!(usage.cached_prompt_tokens, 2500);
assert_eq!(usage.total_tokens, 2501);
}
fn tool_use_start(index: u32, id: &str) -> MessageStreamEvent {
MessageStreamEvent::ContentBlockStart {
index,
content_block: ContentBlock::ToolUse {
id: id.into(),
name: "do_thing".into(),
input: serde_json::json!({}),
},
}
}
fn arg_delta(index: u32, partial_json: &str) -> MessageStreamEvent {
MessageStreamEvent::ContentBlockDelta {
index,
delta: StreamDelta::InputJsonDelta {
partial_json: partial_json.into(),
},
}
}
async fn max_tokens_after(events: Vec<MessageStreamEvent>) -> Vec<SamplingEvent> {
let mut all = vec![Ok(message_start())];
all.extend(events.into_iter().map(Ok));
all.push(Ok(message_delta_with_stop(messages::StopReason::MaxTokens)));
all.push(Ok(MessageStreamEvent::MessageStop));
let raw = stream::iter(all).boxed();
collect(stream_messages(raw, None, rid(), Duration::from_secs(60))).await
}
/// Complete tool calls before a max_tokens stop complete the turn as ToolCalls.
#[tokio::test]
async fn max_tokens_after_complete_tool_calls_completes_as_tool_calls() {
let evs = max_tokens_after(vec![
tool_use_start(0, "call_done"),
arg_delta(0, "{\"x\": 1}"),
block_stop(0),
])
.await;
match evs.last().unwrap() {
SamplingEvent::Completed { response, .. } => {
assert_eq!(response.stop_reason, Some(StopReason::ToolCalls));
assert_eq!(response.tool_calls().len(), 1);
assert_eq!(response.tool_calls()[0].arguments.as_ref(), "{\"x\": 1}");
}
other => panic!("expected Completed(ToolCalls), got {other:?}"),
}
}
/// A tool_use block with no argument deltas is the zero-argument convention, not truncation.
#[tokio::test]
async fn max_tokens_after_tool_call_without_arguments_completes() {
let evs = max_tokens_after(vec![tool_use_start(0, "call_no_args"), block_stop(0)]).await;
match evs.last().unwrap() {
SamplingEvent::Completed { response, .. } => {
assert_eq!(response.stop_reason, Some(StopReason::ToolCalls));
assert_eq!(response.tool_calls()[0].arguments.as_ref(), "");
}
other => panic!("expected Completed(ToolCalls), got {other:?}"),
}
}
/// A call cut mid-arguments is truncation: the turn fails so the caller can compact.
#[tokio::test]
async fn max_tokens_with_truncated_tool_arguments_fails() {
let evs = max_tokens_after(vec![
tool_use_start(0, "call_ok"),
arg_delta(0, "{}"),
block_stop(0),
tool_use_start(1, "call_cut"),
arg_delta(1, "{\"x\": \"trunc"),
block_stop(1),
])
.await;
assert!(
!evs.iter()
.any(|e| matches!(e, SamplingEvent::Completed { .. })),
"a cut tool call must not complete: {evs:?}"
);
match evs.last().unwrap() {
SamplingEvent::Failed { error, .. } => assert_eq!(
error.kind,
crate::events::SamplingErrorKind::MaxTokensTruncation
),
other => panic!("expected Failed(MaxTokensTruncation), got {other:?}"),
}
}
crates/codegen/kigi-sampling-types/src/error.rs
+33 -3
View File
@@ -239,9 +239,7 @@ impl SamplingError {
SamplingError::InvalidConfiguration(_) => false,
SamplingError::Http(err) => is_retryable_reqwest(err),
SamplingError::Serialization(_) => false,
SamplingError::Api { status, .. } => {
matches!(status.as_u16(), 429 | 500 | 502 | 503 | 504 | 520)
}
SamplingError::Api { status, .. } => is_retryable_api_status(status.as_u16()),
SamplingError::EventStreamError(_) => true,
SamplingError::StreamError { .. } => true,
SamplingError::IdleTimeout { .. } => false,
@@ -342,6 +340,15 @@ pub fn try_parse_stream_error(data: &str) -> Option<SamplingError> {
})
}
/// 429 and any 5xx except Cloudflare 525/526 (origin TLS handshake /
/// invalid certificate — a broken origin certificate never clears on its
/// own). The 5xx sweep covers the Cloudflare edge pages (520-524 origin
/// unreachable/timed out, 530 edge 1xxx) and upstream overload (529),
/// which an outage serves in place of the API's own statuses.
pub fn is_retryable_api_status(status: u16) -> bool {
status == 429 || ((500..=599).contains(&status) && !matches!(status, 525 | 526))
}
/// True when an error message indicates a context-window overflow. Backends report
/// this inconsistently with no stable error code, so we match the message text; it's
/// deterministic (re-sending the same payload always fails), so callers must not retry.
@@ -352,6 +359,7 @@ pub fn is_context_length_error(message: &str) -> bool {
|| m.contains("maximum prompt length")
|| m.contains("maximum context length")
|| m.contains("context_length_exceeded")
|| (m.contains("current message") && m.contains("exceeds budget"))
}
/// Decide whether a [`reqwest::Error`] is worth retrying.
@@ -378,6 +386,27 @@ pub fn is_retryable_reqwest(err: &reqwest::Error) -> bool {
mod tests {
use super::*;
#[test]
fn retryable_api_status_covers_edge_5xx_and_excludes_broken_tls() {
for status in [
429u16, 500, 502, 503, 504, 520, 521, 522, 523, 524, 529, 530, 599,
] {
assert!(is_retryable_api_status(status), "should retry {status}");
}
// Cloudflare 525/526: origin TLS handshake / invalid certificate —
// a broken origin certificate never clears on its own.
for status in [525u16, 526, 400, 401, 403, 404, 408, 422, 200] {
assert!(!is_retryable_api_status(status), "must not retry {status}");
}
let api = SamplingError::Api {
status: StatusCode::from_u16(522).unwrap(),
message: "cloudflare: connection timed out".into(),
model_metadata: None,
retry_after_secs: None,
};
assert!(api.is_retryable(), "522 edge page must be retryable");
}
#[test]
fn context_length_error_matches_backend_variants() {
for msg in [
@@ -387,6 +416,7 @@ mod tests {
"This model's maximum context length is 200000 tokens",
"invalid_request_error: prompt is too long: 300000 tokens > 200000 maximum",
"error type: context_length_exceeded",
"the current message exceeds budget: 120000 < 250000",
] {
assert!(is_context_length_error(msg), "should match: {msg}");
}
crates/codegen/kigi-sampling-types/src/types.rs
+12 -36
View File
@@ -911,11 +911,6 @@ pub enum ReasoningEffort {
/// Messages both accept `xhigh` AND `max` as separate levels in 2026;
/// the Kimi wire spells its top tier `max` with no `xhigh`).
Max,
/// Codex-only top tier above `max` (the ChatGPT Codex backend exposes an
/// `ultra` reasoning effort on its flagship models). Reachable ONLY via a
/// model's server-declared effort menu (openai-codex); no built-in fallback
/// menu offers it, so other providers never emit it.
Ultra,
}
impl ReasoningEffort {
@@ -944,15 +939,12 @@ impl ReasoningEffort {
Self::High => "high",
Self::Xhigh => "xhigh",
Self::Max => "max",
Self::Ultra => "ultra",
}
}
/// Anthropic Messages API effort string; `None` for unsupported variants.
/// `xhigh` and `max` are distinct levels on the 2026 Messages API (both
/// appear in `GET /v1/models` `capabilities.effort`). `ultra` is codex-only
/// and never selected on an Anthropic model, but maps to its own string for
/// completeness (the Responses path writes effort via `as_str`, not this).
/// appear in `GET /v1/models` `capabilities.effort`).
pub fn to_messages_api(self) -> Option<&'static str> {
match self {
Self::None | Self::Minimal => None,
@@ -961,7 +953,6 @@ impl ReasoningEffort {
Self::High => Some("high"),
Self::Xhigh => Some("xhigh"),
Self::Max => Some("max"),
Self::Ultra => Some("ultra"),
}
}
}
@@ -984,9 +975,8 @@ impl std::str::FromStr for ReasoningEffort {
"high" => Ok(Self::High),
"xhigh" => Ok(Self::Xhigh),
"max" => Ok(Self::Max),
"ultra" => Ok(Self::Ultra),
_ => Err(format!(
"invalid reasoning effort: {s:?} (expected one of: none, minimal, low, medium, high, xhigh, max, ultra)"
"invalid reasoning effort: {s:?} (expected one of: none, minimal, low, medium, high, xhigh, max)"
)),
}
}
@@ -1878,30 +1868,16 @@ mod tests {
assert_eq!(ReasoningEffort::Max.to_messages_api(), Some("max"));
}
/// The codex-only `ultra` tier parses, serializes, and patches onto a
/// Responses body as `reasoning.effort = "ultra"` (the crux of surfacing a
/// codex model's full thinking menu). It is a DISTINCT level above `max`.
/// `ultra` is not a tier any backend accepts.
///
/// The Codex Responses endpoint rejects it with a 400 listing its menu,
/// which tops out at `max`. Parsing it would only send it again.
#[test]
fn reasoning_effort_ultra_is_a_distinct_codex_tier() {
assert_eq!(
"ultra".parse::<ReasoningEffort>().unwrap(),
ReasoningEffort::Ultra
);
assert_eq!(
"ULTRA".parse::<ReasoningEffort>().unwrap(),
ReasoningEffort::Ultra
);
assert_ne!(ReasoningEffort::Ultra, ReasoningEffort::Max);
assert_eq!(ReasoningEffort::Ultra.as_str(), "ultra");
let json = serde_json::to_string(&ReasoningEffort::Ultra).unwrap();
assert_eq!(json, "\"ultra\"");
assert_eq!(
serde_json::from_str::<ReasoningEffort>("\"ultra\"").unwrap(),
ReasoningEffort::Ultra
);
let mut body = serde_json::json!({ "model": "gpt-5.6-sol" });
patch_reasoning_effort(&mut body, Some(ReasoningEffort::Ultra));
assert_eq!(body["reasoning"]["effort"], "ultra");
fn reasoning_effort_ultra_is_not_a_tier() {
assert!("ultra".parse::<ReasoningEffort>().is_err());
assert!("ULTRA".parse::<ReasoningEffort>().is_err());
assert!(serde_json::from_str::<ReasoningEffort>("\"ultra\"").is_err());
assert_eq!(ReasoningEffort::Max.as_str(), "max");
}
/// The account id is decoded STATELESSLY from the bearer JWT payload's
@@ -2122,7 +2098,7 @@ mod tests {
);
let bad_type = as_map(serde_json::json!({"reasoningEffort": 3}));
assert_eq!(parse_reasoning_effort_meta(Some(&bad_type)), None);
// `ultra` is a real codex tier; a genuinely-unknown token still None.
// A genuinely-unknown token is dropped, not guessed at.
let unknown = as_map(serde_json::json!({"reasoningEffort": "MEGA"}));
assert_eq!(parse_reasoning_effort_meta(Some(&unknown)), None);
}
crates/codegen/kigi-sandbox/Cargo.toml
+1
View File
@@ -51,6 +51,7 @@ enforce = []
[dev-dependencies]
serial_test = { workspace = true }
tempfile = { workspace = true }
# Compiles the hand-rolled macOS deny-glob regex in tests to assert it matches
# EXACTLY the set globset (the Linux backend) matches — the cross-platform parity
# guard for the glob dialect.
crates/codegen/kigi-sandbox/examples/sandbox_smoke_test.rs
+12
View File
@@ -14,9 +14,19 @@
//! cargo run -p kigi-sandbox --example sandbox_smoke_test -- read-only
//! ```
// Kernel enforcement is unix-only (Seatbelt / Landlock); Windows builds a
// stub so `--all-targets` stays green everywhere.
#[cfg(unix)]
use kigi_sandbox::{ProfileName, SandboxManager};
#[cfg(unix)]
use std::path::Path;
#[cfg(not(unix))]
fn main() {
eprintln!("sandbox_smoke_test exercises kernel enforcement and only runs on Unix.");
}
#[cfg(unix)]
fn main() {
let profile_name = std::env::args()
.nth(1)
@@ -121,6 +131,7 @@ fn main() {
println!("\n✅ Smoke test complete");
}
#[cfg(unix)]
fn test_read(label: &str, path: &Path) {
if path.is_file() {
match std::fs::read(path) {
@@ -150,6 +161,7 @@ fn test_read(label: &str, path: &Path) {
}
}
#[cfg(unix)]
fn test_write(label: &str, path: &Path) {
match std::fs::write(path, b"sandbox-test") {
Ok(()) => {
crates/codegen/kigi-sandbox/src/child_net.rs
+210 -74
View File
@@ -1,91 +1,130 @@
//! Per-child seccomp network filter. No-op on non-Linux.
/// Install seccomp BPF filter blocking network syscalls.
#[cfg(target_os = "linux")]
mod bpf {
use libc::sock_filter;
pub(super) const SECCOMP_RET_ALLOW: u32 = 0x7fff_0000;
pub(super) const SECCOMP_RET_ERRNO: u32 = 0x0005_0000;
pub(super) const EPERM_VAL: u32 = 1;
#[cfg(target_arch = "x86_64")]
pub(super) const X32_SYSCALL_BIT: u32 = 0x4000_0000;
/// `seccomp_data` field offsets.
pub(super) const OFF_NR: u32 = 0;
pub(super) const OFF_ARCH: u32 = 4;
#[cfg(target_arch = "x86_64")]
pub(super) const EXPECTED_ARCH: u32 = 0xc000_003e; // AUDIT_ARCH_X86_64
#[cfg(target_arch = "aarch64")]
pub(super) const EXPECTED_ARCH: u32 = 0xc000_00b7; // AUDIT_ARCH_AARCH64
#[cfg(not(any(target_arch = "x86_64", target_arch = "aarch64")))]
pub(super) const EXPECTED_ARCH: u32 = 0;
pub(super) fn stmt(code: u32, k: u32) -> sock_filter {
sock_filter {
code: code as u16,
jt: 0,
jf: 0,
k,
}
}
pub(super) fn jump(code: u32, k: u32, jt: u8, jf: u8) -> sock_filter {
sock_filter {
code: code as u16,
jt,
jf,
k,
}
}
/// Reject a foreign-arch (and, on x86_64, x32) syscall number before
/// matching `nr`, so a syscall reached through another table cannot
/// alias past the deny list.
pub(super) fn push_arch_nr_gate(f: &mut Vec<sock_filter>) {
use libc::{BPF_ABS, BPF_JEQ, BPF_JMP, BPF_K, BPF_LD, BPF_RET, BPF_W};
f.push(stmt(BPF_LD | BPF_W | BPF_ABS, OFF_ARCH));
f.push(jump(BPF_JMP | BPF_JEQ | BPF_K, EXPECTED_ARCH, 1, 0));
f.push(stmt(BPF_RET | BPF_K, SECCOMP_RET_ERRNO | EPERM_VAL));
f.push(stmt(BPF_LD | BPF_W | BPF_ABS, OFF_NR));
#[cfg(target_arch = "x86_64")]
{
f.push(jump(
BPF_JMP | libc::BPF_JSET | BPF_K,
X32_SYSCALL_BIT,
0,
1,
));
f.push(stmt(BPF_RET | BPF_K, SECCOMP_RET_ERRNO | EPERM_VAL));
}
}
}
/// Syscalls that open or send on a socket, including the `io_uring` and
/// `sendmmsg` paths that bypass the classic entry points.
#[cfg(target_os = "linux")]
fn child_network_blocked_syscalls() -> [u32; 11] {
use libc::{
SYS_accept, SYS_accept4, SYS_bind, SYS_connect, SYS_io_uring_enter, SYS_io_uring_register,
SYS_io_uring_setup, SYS_listen, SYS_sendmmsg, SYS_sendmsg, SYS_sendto,
};
[
SYS_connect as u32,
SYS_bind as u32,
SYS_sendto as u32,
SYS_sendmsg as u32,
SYS_sendmmsg as u32,
SYS_listen as u32,
SYS_accept as u32,
SYS_accept4 as u32,
SYS_io_uring_setup as u32,
SYS_io_uring_enter as u32,
SYS_io_uring_register as u32,
]
}
#[cfg(target_os = "linux")]
fn build_child_network_filter() -> Vec<libc::sock_filter> {
use bpf::{EPERM_VAL, SECCOMP_RET_ALLOW, SECCOMP_RET_ERRNO, jump, stmt};
use libc::{BPF_JEQ, BPF_JMP, BPF_K, BPF_RET};
let blocked = child_network_blocked_syscalls();
let mut f = Vec::with_capacity(blocked.len() + 10);
bpf::push_arch_nr_gate(&mut f);
let n = blocked.len();
for (i, &sys) in blocked.iter().enumerate() {
let remaining = n - i - 1;
f.push(jump(BPF_JMP | BPF_JEQ | BPF_K, sys, remaining as u8 + 1, 0));
}
f.push(stmt(BPF_RET | BPF_K, SECCOMP_RET_ALLOW));
f.push(stmt(BPF_RET | BPF_K, SECCOMP_RET_ERRNO | EPERM_VAL));
f
}
/// Built once: `pre_exec` runs after `fork`, where allocating is unsafe.
#[cfg(target_os = "linux")]
pub fn prebuilt_child_network_filter() -> &'static [libc::sock_filter] {
static FILTER: std::sync::OnceLock<Vec<libc::sock_filter>> = std::sync::OnceLock::new();
FILTER.get_or_init(build_child_network_filter)
}
/// Install the seccomp BPF filter blocking network syscalls.
///
/// # Safety
///
/// Must be called in a `pre_exec` context (after `fork`, before `exec`).
#[cfg(target_os = "linux")]
pub unsafe fn install_child_network_filter() -> std::io::Result<()> {
use libc::{
BPF_ABS, BPF_JEQ, BPF_JMP, BPF_K, BPF_LD, BPF_RET, BPF_W, PR_SET_NO_NEW_PRIVS,
PR_SET_SECCOMP, SECCOMP_MODE_FILTER, SYS_accept, SYS_accept4, SYS_bind, SYS_connect,
SYS_listen, SYS_sendmsg, SYS_sendto, prctl, sock_filter, sock_fprog,
};
const SECCOMP_RET_ALLOW: u32 = 0x7fff_0000;
const SECCOMP_RET_ERRNO: u32 = 0x0005_0000;
// libc::EPERM
const EPERM_VAL: u32 = 1;
macro_rules! bpf_stmt {
($code:expr, $k:expr) => {
sock_filter {
code: $code as u16,
jt: 0,
jf: 0,
k: $k as u32,
}
};
}
macro_rules! bpf_jump {
($code:expr, $k:expr, $jt:expr, $jf:expr) => {
sock_filter {
code: $code as u16,
jt: $jt,
jf: $jf,
k: $k as u32,
}
};
}
// seccomp_data.nr offset
const NR_OFFSET: u32 = 0;
let blocked_syscalls: &[i64] = &[
SYS_connect,
SYS_bind,
SYS_sendto,
SYS_sendmsg,
SYS_listen,
SYS_accept,
SYS_accept4,
];
let mut filter: Vec<sock_filter> = Vec::new();
let total_checks = blocked_syscalls.len();
filter.push(bpf_stmt!(BPF_LD | BPF_W | BPF_ABS, NR_OFFSET));
for (i, &syscall) in blocked_syscalls.iter().enumerate() {
let remaining = total_checks - i - 1;
filter.push(bpf_jump!(
BPF_JMP | BPF_JEQ | BPF_K,
syscall,
// match: jump to ERRNO
remaining as u8 + 1,
// no match: check next
0
));
}
filter.push(bpf_stmt!(BPF_RET | BPF_K, SECCOMP_RET_ALLOW));
filter.push(bpf_stmt!(BPF_RET | BPF_K, SECCOMP_RET_ERRNO | EPERM_VAL));
pub unsafe fn install_child_network_filter(filter: &[libc::sock_filter]) -> std::io::Result<()> {
use libc::{PR_SET_NO_NEW_PRIVS, PR_SET_SECCOMP, SECCOMP_MODE_FILTER, prctl, sock_fprog};
let prog = sock_fprog {
len: filter.len() as u16,
filter: filter.as_mut_ptr(),
filter: filter.as_ptr().cast_mut(),
};
// Must set PR_SET_NO_NEW_PRIVS before applying seccomp filter
// SAFETY: prctl with PR_SET_NO_NEW_PRIVS is safe in pre_exec context.
if unsafe { prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) } != 0 {
return Err(std::io::Error::last_os_error());
}
// SAFETY: prog is a valid sock_fprog pointing to our filter array.
// SAFETY: prog is a valid sock_fprog pointing to a filter the kernel copies.
if unsafe {
prctl(
PR_SET_SECCOMP,
@@ -98,7 +137,6 @@ pub unsafe fn install_child_network_filter() -> std::io::Result<()> {
{
return Err(std::io::Error::last_os_error());
}
Ok(())
}
@@ -109,3 +147,101 @@ pub unsafe fn install_child_network_filter() -> std::io::Result<()> {
pub unsafe fn install_child_network_filter() -> std::io::Result<()> {
Ok(())
}
#[cfg(all(test, target_os = "linux"))]
mod tests {
use super::bpf::*;
use super::{build_child_network_filter, child_network_blocked_syscalls};
use libc::sock_filter;
/// Minimal classic-BPF interpreter over synthetic `seccomp_data` fields.
fn eval(filter: &[sock_filter], arch: u32, nr: u32) -> u32 {
use libc::{BPF_ABS, BPF_JEQ, BPF_JMP, BPF_JSET, BPF_K, BPF_LD, BPF_RET, BPF_W};
let mut pc = 0usize;
let mut a = 0u32;
for _ in 0..filter.len().saturating_mul(2) {
let insn = &filter[pc];
let op = insn.code as u32;
if op == (BPF_LD | BPF_W | BPF_ABS) {
a = match insn.k {
OFF_NR => nr,
OFF_ARCH => arch,
_ => 0,
};
pc += 1;
} else if op == (BPF_JMP | BPF_JEQ | BPF_K) {
pc = if a == insn.k {
pc + 1 + insn.jt as usize
} else {
pc + 1 + insn.jf as usize
};
} else if op == (BPF_JMP | BPF_JSET | BPF_K) {
pc = if a & insn.k != 0 {
pc + 1 + insn.jt as usize
} else {
pc + 1 + insn.jf as usize
};
} else if op == (BPF_RET | BPF_K) {
return insn.k;
} else {
panic!("unsupported opcode {:#x} at {pc}", insn.code);
}
assert!(pc < filter.len(), "pc out of range");
}
panic!("filter did not RET");
}
fn is_allow(r: u32) -> bool {
r == SECCOMP_RET_ALLOW
}
fn is_eperm(r: u32) -> bool {
r == (SECCOMP_RET_ERRNO | EPERM_VAL)
}
#[test]
fn blocks_every_network_syscall_and_allows_read_and_socket() {
let f = build_child_network_filter();
for sys in child_network_blocked_syscalls() {
assert!(
is_eperm(eval(&f, EXPECTED_ARCH, sys)),
"syscall {sys} must be EPERM"
);
}
assert!(is_allow(eval(&f, EXPECTED_ARCH, libc::SYS_read as u32)));
assert!(is_allow(eval(&f, EXPECTED_ARCH, libc::SYS_socket as u32)));
}
#[test]
fn covers_io_uring_and_sendmmsg() {
let blocked = child_network_blocked_syscalls();
for sys in [
libc::SYS_sendmmsg,
libc::SYS_io_uring_setup,
libc::SYS_io_uring_enter,
libc::SYS_io_uring_register,
] {
assert!(blocked.contains(&(sys as u32)), "{sys} missing");
}
}
#[test]
fn wrong_arch_and_x32_are_denied_before_the_nr_match() {
let f = build_child_network_filter();
assert!(is_eperm(eval(&f, 0xdead_beef, libc::SYS_read as u32)));
#[cfg(target_arch = "x86_64")]
assert!(is_eperm(eval(
&f,
EXPECTED_ARCH,
(libc::SYS_read as u32) | X32_SYSCALL_BIT
)));
}
#[test]
fn prebuilt_filter_is_shared() {
let a = super::prebuilt_child_network_filter();
let b = super::prebuilt_child_network_filter();
assert!(std::ptr::eq(a, b));
assert_eq!(a.len(), build_child_network_filter().len());
}
}
crates/codegen/kigi-sandbox/src/lib.rs
+116 -4
View File
@@ -333,6 +333,27 @@ fn is_devbox_based(profile: &ProfileName, config: &SandboxConfig) -> bool {
_ => false,
}
}
/// Whether an existing `/data` needs the devbox bwrap read-only bind; a filesystem error other than NotFound conservatively requires it.
#[cfg(target_os = "linux")]
pub fn requires_data_write_deny(profile: &ProfileName, workspace: &Path) -> bool {
requires_data_write_deny_for(
profile,
&profiles::load_sandbox_config(workspace),
data_path_requires_bind(Path::new("/data")),
)
}
#[cfg(target_os = "linux")]
fn requires_data_write_deny_for(
profile: &ProfileName,
config: &SandboxConfig,
is_data_present: bool,
) -> bool {
is_devbox_based(profile, config) && is_data_present
}
#[cfg(target_os = "linux")]
fn data_path_requires_bind(path: &Path) -> bool {
path.try_exists().unwrap_or(true)
}
/// Whether kernel read-deny enforcement is required. The single source of truth
/// for this classification so callers (e.g. the shell's fail-closed startup path)
/// cannot drift and silently fail open.
@@ -360,6 +381,23 @@ pub fn requires_read_deny(profile: &ProfileName, workspace: &Path) -> bool {
pub fn requires_read_deny(_profile: &ProfileName, _workspace: &Path) -> bool {
false
}
/// Whether the resolved profile leaves `~/.kigi` read-only outside its
/// state dirs (strict and every profile extending it), so files first use
/// would write there must be minted before enforcement.
#[cfg(all(feature = "enforce", unix))]
pub fn restricts_home_writes(profile: &ProfileName, workspace: &Path) -> bool {
if *profile == ProfileName::Off {
return false;
}
let config = profiles::load_sandbox_config(workspace);
profile
.resolve_profile(workspace, &config)
.is_ok_and(|resolved| !resolved.read_write.contains(&paths::kigi_home()))
}
#[cfg(not(all(feature = "enforce", unix)))]
pub fn restricts_home_writes(_profile: &ProfileName, _workspace: &Path) -> bool {
false
}
/// A profile's resolved bwrap deny plan: read-only mounts (`deny_write`),
/// bound-over unreadable placeholders (`deny_read`), and whether the profile
/// carries deny globs (`has_globs`, so the re-exec proceeds even with zero
@@ -382,7 +420,11 @@ struct BwrapDenyPlan {
#[cfg(all(feature = "enforce", target_os = "linux"))]
fn bwrap_deny_plan(profile: &ProfileName, workspace: &Path) -> Option<BwrapDenyPlan> {
let config = profiles::load_sandbox_config(workspace);
let deny_write: Vec<String> = if is_devbox_based(profile, &config) {
let deny_write: Vec<String> = if requires_data_write_deny_for(
profile,
&config,
data_path_requires_bind(Path::new("/data")),
) {
vec!["/data".to_string()]
} else {
Vec::new()
@@ -424,7 +466,11 @@ fn bwrap_deny_plan(profile: &ProfileName, workspace: &Path) -> Option<BwrapDenyP
#[cfg(all(not(feature = "enforce"), target_os = "linux"))]
fn bwrap_deny_plan(profile: &ProfileName, workspace: &Path) -> Option<BwrapDenyPlan> {
let config = profiles::load_sandbox_config(workspace);
let deny_write: Vec<String> = if is_devbox_based(profile, &config) {
let deny_write: Vec<String> = if requires_data_write_deny_for(
profile,
&config,
data_path_requires_bind(Path::new("/data")),
) {
vec!["/data".to_string()]
} else {
Vec::new()
@@ -595,6 +641,43 @@ mod tests {
);
}
#[test]
#[cfg(target_os = "linux")]
fn data_write_deny_requires_devbox_and_a_present_data_dir() {
let tmp = tempfile::tempdir().unwrap();
std::fs::create_dir_all(tmp.path().join(".kigi")).unwrap();
std::fs::write(
tmp.path().join(".kigi/sandbox.toml"),
"[profiles.devext]\nextends = \"devbox\"\n\n[profiles.plain]\nextends = \"workspace\"\n",
)
.unwrap();
let config = profiles::load_sandbox_config(tmp.path());
assert!(requires_data_write_deny_for(
&ProfileName::Devbox,
&config,
true
));
assert!(!requires_data_write_deny_for(
&ProfileName::Devbox,
&config,
false
));
assert!(requires_data_write_deny_for(
&ProfileName::Custom("devext".into()),
&config,
true
));
assert!(!requires_data_write_deny_for(
&ProfileName::Custom("plain".into()),
&config,
true
));
assert!(!requires_data_write_deny_for(
&ProfileName::Workspace,
&config,
true
));
}
#[test]
fn configured_profile_is_recorded() {
set_configured_profile("read-only");
assert_eq!(configured_profile_name(), Some("read-only"));
@@ -700,10 +783,11 @@ mod tests {
"devbox-empty",
"[profiles.devempty]\nextends = \"devbox\"\n",
);
assert!(
assert_eq!(
bwrap_reexec_for_profile(&ProfileName::Custom("devempty".to_string()), &ws_empty)
.is_some(),
"devbox-extending custom must compose the /data write-deny re-exec"
Path::new("/data").exists(),
"devbox-extending custom composes the /data write-deny re-exec exactly when /data exists"
);
let _ = std::fs::remove_dir_all(&ws_empty);
let ws_ws = temp_workspace_with_sandbox_toml(
@@ -717,3 +801,31 @@ mod tests {
let _ = std::fs::remove_dir_all(&ws_ws);
}
}
#[cfg(all(test, feature = "enforce", unix))]
mod home_write_tests {
use super::{ProfileName, restricts_home_writes};
#[test]
fn strict_restricts_home_writes_and_workspace_does_not() {
let workspace = std::env::temp_dir();
assert!(restricts_home_writes(&ProfileName::Strict, &workspace));
assert!(!restricts_home_writes(&ProfileName::Workspace, &workspace));
assert!(!restricts_home_writes(&ProfileName::ReadOnly, &workspace));
assert!(!restricts_home_writes(&ProfileName::Off, &workspace));
}
#[test]
fn custom_profile_extending_strict_inherits_the_restriction() {
let tmp = tempfile::tempdir().unwrap();
let project = tmp.path().join(".kigi");
std::fs::create_dir_all(&project).unwrap();
std::fs::write(
project.join("sandbox.toml"),
"[profiles.audit-home-write-test]\nextends = \"strict\"\n",
)
.unwrap();
let custom = ProfileName::Custom("audit-home-write-test".to_string());
assert!(restricts_home_writes(&custom, tmp.path()));
}
}
crates/codegen/kigi-sandbox/src/logging.rs
+6 -2
View File
@@ -1,7 +1,7 @@
//! Sandbox event logger.
//!
//! Events (profile applied, violations, bypasses) are buffered in memory and
//! flushed as JSONL to `~/.kigi/sandbox-events.jsonl`.
//! flushed as JSONL to `~/.kigi/sessions/sandbox-events.jsonl`.
use std::path::PathBuf;
use std::sync::Mutex;
@@ -87,8 +87,12 @@ impl SandboxLogger {
Ok(())
}
/// Under `sessions/`: the only home subtree a strict sandbox leaves
/// writable, and this flush runs after enforcement.
fn log_file_path() -> PathBuf {
kigi_config::kigi_home().join("sandbox-events.jsonl")
kigi_config::kigi_home()
.join("sessions")
.join("sandbox-events.jsonl")
}
}
crates/codegen/kigi-sandbox/src/paths.rs
+31
View File
@@ -96,6 +96,37 @@ pub(crate) fn essential_writable_paths(workspace: &Path) -> Vec<PathBuf> {
paths
}
/// Home subtrees the process keeps writing after enforcement: session and
/// worktree state, memory, diagnostics. Landlock cannot carve a child out of
/// a write grant, so the rest of `~/.kigi` (`config.toml`, `auth.json`,
/// `sandbox.toml`, `hooks/`, `bin/`, `vendor/`, `plugin-data/`) stays
/// read-only and a sandboxed child cannot arm the next launch.
#[cfg(all(feature = "enforce", unix))]
const HOME_STATE_DIRS: &[&str] = &[
"run",
"sessions",
"worktrees",
"worktree_pool",
"tasks",
"memory",
"logs",
"debug",
"crash",
"memtrace",
"trace-exports",
];
/// Writable paths for the strict profile: the workspace, the home state
/// subtrees, and temp.
#[cfg(all(feature = "enforce", unix))]
pub(crate) fn essential_writable_paths_strict(workspace: &Path) -> Vec<PathBuf> {
let home = kigi_home();
let mut paths = vec![workspace.to_path_buf()];
paths.extend(HOME_STATE_DIRS.iter().map(|dir| home.join(dir)));
paths.extend(temp_writable_paths());
paths
}
/// Writable directory paths for the read-only profile (minimal: just ~/.kigi + temp).
/// Device files are handled separately via `allow_file` in `to_capability_set_with_config`.
#[cfg(all(feature = "enforce", unix))]
crates/codegen/kigi-sandbox/src/profiles.rs
+113 -6
View File
@@ -18,6 +18,7 @@ use crate::paths::kigi_home;
#[cfg(all(feature = "enforce", unix))]
use crate::paths::{
DEVICE_DIRS, DEVICE_FILES, essential_writable_paths, essential_writable_paths_minimal,
essential_writable_paths_strict,
};
/// A resolved sandbox profile ready to be converted to a `CapabilitySet`.
@@ -186,6 +187,32 @@ fn load_config_file(path: &Path) -> Option<SandboxConfig> {
#[cfg(all(feature = "enforce", unix))]
impl ProfileName {
/// Convert this profile into a nono `CapabilitySet` for the given workspace.
/// The directory a `read_write` entry really grants. A symlink under
/// `~/.kigi` is followed only to the default `~/.kigi/sessions` (the
/// `KIGI_SHARE_DIR` override case); any other target would widen the
/// grant to wherever the link points. Missing paths are created.
fn read_write_grant_path(path: &Path, home: &Path) -> Option<PathBuf> {
match std::fs::symlink_metadata(path) {
Ok(meta) if meta.file_type().is_symlink() => {
if path == home || !path.starts_with(home) {
return Some(path.to_path_buf());
}
let real = dunce::canonicalize(path).ok()?;
let default_sessions = kigi_config::default_kigi_home().join("sessions");
(real.is_dir()
&& path.file_name() == Some(std::ffi::OsStr::new("sessions"))
&& real == default_sessions)
.then_some(real)
}
Ok(_) => Some(path.to_path_buf()),
Err(err) if err.kind() == std::io::ErrorKind::NotFound => {
std::fs::create_dir_all(path).ok()?;
Some(path.to_path_buf())
}
Err(_) => None,
}
}
pub fn to_capability_set(&self, workspace: &Path) -> anyhow::Result<CapabilitySet> {
let config = load_sandbox_config(workspace);
self.to_capability_set_with_config(workspace, &config)
@@ -231,13 +258,14 @@ impl ProfileName {
// apply time (it opens an O_PATH fd), but new files within it can
// be created freely after the sandbox is applied. Pre-create
// directories like ~/.kigi/ that may not exist on first run.
let home = kigi_home();
for path in &profile.read_write {
if !path.exists() && std::fs::create_dir_all(path).is_err() {
tracing::warn!(path = ?path, "read_write path does not exist and could not be created, skipping");
let Some(grant) = Self::read_write_grant_path(path, &home) else {
tracing::warn!(path = ?path, "read_write grant skipped");
continue;
}
let Some(path_str) = path.to_str() else {
tracing::warn!(path = ?path, "Skipping non-UTF8 read_write path");
};
let Some(path_str) = grant.to_str() else {
tracing::warn!(path = ?grant, "Skipping non-UTF8 read_write path");
continue;
};
caps = caps.allow_path(path_str, AccessMode::ReadWrite)?;
@@ -380,12 +408,15 @@ impl ProfileName {
.chain(std::iter::once(home.join("Library")))
.filter(|p| p.exists())
.chain(std::iter::once(workspace.to_path_buf()))
// Read-only: config, auth and global hooks are read after
// enforcement; only sessions/ below stays writable.
.chain(std::iter::once(kigi_home()))
.collect();
Ok(SandboxProfile {
name: "strict".to_string(),
read_only: system_read,
read_write: essential_writable_paths(workspace),
read_write: essential_writable_paths_strict(workspace),
deny: vec![],
default_read: false,
restrict_network: true,
@@ -544,6 +575,82 @@ mod tests {
}
}
#[test]
#[cfg(all(feature = "enforce", unix))]
fn strict_read_write_is_workspace_home_state_and_temp() {
let workspace = std::env::temp_dir();
let profile = ProfileName::Strict
.resolve_profile(&workspace, &SandboxConfig::default())
.expect("strict resolves");
let home = kigi_home();
assert_eq!(profile.read_write[0], workspace);
for dir in ["sessions", "worktrees", "memory", "logs", "debug"] {
assert!(
profile.read_write.contains(&home.join(dir)),
"{dir} must stay writable: {:?}",
profile.read_write
);
}
for denied in [
home.clone(),
home.join("hooks"),
home.join("vendor"),
home.join("bin"),
] {
assert!(
!profile.read_write.contains(&denied),
"strict must not grant {}: {:?}",
denied.display(),
profile.read_write
);
}
assert!(
profile.read_only.contains(&home),
"config/auth/hooks are read after enforcement: {:?}",
profile.read_only
);
}
#[test]
#[cfg(all(feature = "enforce", unix))]
fn read_write_grant_follows_real_dirs_and_creates_missing_ones() {
let tmp = tempfile::tempdir().unwrap();
let home = tmp.path().join("home");
std::fs::create_dir_all(home.join("sessions")).unwrap();
assert_eq!(
ProfileName::read_write_grant_path(&home.join("sessions"), &home),
Some(home.join("sessions"))
);
let missing = home.join("fresh");
assert_eq!(
ProfileName::read_write_grant_path(&missing, &home),
Some(missing.clone())
);
assert!(missing.is_dir());
}
#[test]
#[cfg(all(feature = "enforce", unix))]
fn read_write_grant_refuses_a_home_symlink_pointing_elsewhere() {
let tmp = tempfile::tempdir().unwrap();
let home = tmp.path().join("home");
let outside = tmp.path().join("outside");
std::fs::create_dir_all(&home).unwrap();
std::fs::create_dir_all(&outside).unwrap();
std::os::unix::fs::symlink(&outside, home.join("sessions")).unwrap();
assert_eq!(
ProfileName::read_write_grant_path(&home.join("sessions"), &home),
None
);
// A symlinked workspace outside the home is the user's own choice.
let ws_link = tmp.path().join("ws");
std::os::unix::fs::symlink(&outside, &ws_link).unwrap();
assert_eq!(
ProfileName::read_write_grant_path(&ws_link, &home),
Some(ws_link.clone())
);
}
#[test]
#[cfg(all(feature = "enforce", unix))]
fn base_profile_capability_set_builds() {
crates/codegen/kigi-shared/src/clipboard.rs
+9 -10
View File
@@ -1164,6 +1164,7 @@ mod platform {
#[cfg(not(target_os = "macos"))]
mod platform {
use super::ImageData;
#[cfg(target_os = "linux")]
use std::process::{Command, Stdio};
/// No subprocess-free pasteboard probe exists off-macOS.
@@ -1926,19 +1927,18 @@ mod platform {
// Public API
pub fn get_text() -> anyhow::Result<Option<String>> {
let mut arboard_error = None;
match arboard_get_text() {
let arboard_error = match arboard_get_text() {
Ok(Some(text)) => return Ok(Some(text)),
// Wayland-only: arboard's empty answer is not authoritative,
// fall through to wl-paste (see `wayland_tool_selected`).
#[cfg(target_os = "linux")]
Ok(None) if wayland_tool_selected(linux_tool_spec()) => {}
Ok(None) if wayland_tool_selected(linux_tool_spec()) => None,
Ok(None) => return Ok(None),
Err(e) => {
tracing::debug!("arboard get_text failed: {e}");
arboard_error = Some(e);
Some(e)
}
}
};
#[cfg(target_os = "linux")]
if let Some(spec) = linux_tool_spec() {
let bytes = run_capture_out_checked(spec.read_text, CLI_READ_WAIT)?;
@@ -2039,19 +2039,18 @@ mod platform {
}
pub fn get_image() -> anyhow::Result<Option<ImageData>> {
let mut arboard_error = None;
match arboard_get_image() {
let arboard_error = match arboard_get_image() {
Ok(Some(image)) => return Ok(Some(image)),
// Wayland-only: arboard's empty answer is not authoritative,
// fall through to wl-paste (see `wayland_tool_selected`).
#[cfg(target_os = "linux")]
Ok(None) if wayland_tool_selected(linux_tool_spec()) => {}
Ok(None) if wayland_tool_selected(linux_tool_spec()) => None,
Ok(None) => return Ok(None),
Err(e) => {
tracing::debug!("arboard get_image failed: {e}");
arboard_error = Some(e);
Some(e)
}
}
};
#[cfg(target_os = "linux")]
if let Some(spec) = linux_tool_spec()
&& let Some(argv) = spec.read_png
crates/codegen/kigi-shell-base/src/util/secure_file.rs
+86 -8
View File
@@ -26,7 +26,7 @@ use std::io::{self, Write};
use std::path::Path;
#[cfg(unix)]
use std::os::unix::fs::OpenOptionsExt;
use std::os::unix::fs::{OpenOptionsExt, PermissionsExt};
/// Creates or opens a file with secure permissions (owner read/write only).
///
@@ -58,17 +58,20 @@ pub fn write_secure_file(path: &Path, contents: &[u8]) -> io::Result<()> {
file.write_all(contents)?;
file.flush()?;
// On Windows, we need to set permissions after file creation
#[cfg(windows)]
{
set_windows_secure_permissions(path)?;
}
// Re-assert owner-only bits: `OpenOptions::mode` only applies on create,
// so an existing world-readable file would otherwise keep open perms.
ensure_owner_only_permissions(path)?;
Ok(())
}
/// Opens a file for writing with secure permissions set during creation (Unix)
/// or prepares it for permission setting after creation (Windows).
///
/// Callers that write secret material should also call
/// [`ensure_owner_only_permissions`] after the write (or use
/// [`write_secure_file`]), because `mode(0o600)` only applies when the file
/// is newly created — not when truncating an existing path.
pub fn open_secure_file(path: &Path) -> io::Result<File> {
let mut options = OpenOptions::new();
options.truncate(true).write(true).create(true);
@@ -82,6 +85,44 @@ pub fn open_secure_file(path: &Path) -> io::Result<File> {
options.open(path)
}
/// Ensure `path` is owner-read/write only (Unix `0o600` / Windows user ACL).
///
/// Best-effort on missing files (`NotFound` is ignored). Other errors
/// propagate so callers can fail closed when tightening a secret store.
///
/// Use on **load** of credential files so a hand-copied or restored
/// world-readable `auth.json` is tightened before the process continues.
pub fn ensure_owner_only_permissions(path: &Path) -> io::Result<()> {
match ensure_owner_only_permissions_inner(path) {
Ok(()) => Ok(()),
Err(e) if e.kind() == io::ErrorKind::NotFound => Ok(()),
Err(e) => Err(e),
}
}
fn ensure_owner_only_permissions_inner(path: &Path) -> io::Result<()> {
#[cfg(unix)]
{
let metadata = std::fs::metadata(path)?;
let mode = metadata.permissions().mode();
if mode & 0o777 != 0o600 {
let mut perms = metadata.permissions();
perms.set_mode(0o600);
std::fs::set_permissions(path, perms)?;
}
Ok(())
}
#[cfg(windows)]
{
set_windows_secure_permissions(path)
}
#[cfg(not(any(unix, windows)))]
{
let _ = path;
Ok(())
}
}
/// Sets Windows-specific secure permissions on a file.
///
/// This function modifies the file's ACL to:
@@ -219,8 +260,6 @@ mod tests {
#[cfg(unix)]
#[test]
fn test_unix_permissions() {
use std::os::unix::fs::PermissionsExt;
let temp_dir = tempfile::tempdir().unwrap();
let file_path = temp_dir.path().join("test_perms.txt");
@@ -231,4 +270,43 @@ mod tests {
// Check that only owner has read/write (0o600), ignoring file type bits
assert_eq!(mode & 0o777, 0o600);
}
#[cfg(unix)]
#[test]
fn ensure_owner_only_tightens_world_readable_file() {
let temp_dir = tempfile::tempdir().unwrap();
let file_path = temp_dir.path().join("loose.txt");
fs::write(&file_path, b"secret").unwrap();
let mut loose = fs::metadata(&file_path).unwrap().permissions();
loose.set_mode(0o644);
fs::set_permissions(&file_path, loose).unwrap();
assert_eq!(
fs::metadata(&file_path).unwrap().permissions().mode() & 0o777,
0o644
);
ensure_owner_only_permissions(&file_path).unwrap();
assert_eq!(
fs::metadata(&file_path).unwrap().permissions().mode() & 0o777,
0o600
);
}
#[cfg(unix)]
#[test]
fn write_secure_file_tightens_existing_world_readable_file() {
let temp_dir = tempfile::tempdir().unwrap();
let file_path = temp_dir.path().join("existing.txt");
fs::write(&file_path, b"old").unwrap();
let mut loose = fs::metadata(&file_path).unwrap().permissions();
loose.set_mode(0o666);
fs::set_permissions(&file_path, loose).unwrap();
write_secure_file(&file_path, b"new secret").unwrap();
assert_eq!(
fs::metadata(&file_path).unwrap().permissions().mode() & 0o777,
0o600
);
assert_eq!(fs::read_to_string(&file_path).unwrap(), "new secret");
}
}
crates/codegen/kigi-shell/Cargo.toml
+1
View File
@@ -150,6 +150,7 @@ fs2 = { workspace = true }
zstd = { workspace = true }
tar = { workspace = true }
url = { workspace = true }
kigi-secrets = { path = "../kigi-secrets" }
walkdir = { workspace = true }
tdigests = "1.0"
semver = { workspace = true }
crates/codegen/kigi-shell/README.md
+4 -2
View File
@@ -1309,6 +1309,7 @@ output_byte_limit = 65536 # max output size (64KB)
[toolset.web_fetch]
proxy_endpoint = "https://proxy.example.com" # egress proxy URL (all requests routed through it)
allowed_domains = ["docs.rs", "x.ai"] # override the built-in ~84-domain allowlist
allow_local = false # true = reach an explicit localhost / 127.0.0.0/8 / ::1 URL
[shortcuts]
send = ["Enter"]
@@ -2100,7 +2101,7 @@ kigi --sandbox strict
| `off` (default) | Unrestricted | Unrestricted | Unrestricted | No sandbox |
| `workspace` | Everywhere | CWD + `/tmp` + `~/.kigi/` | Allowed | Normal development |
| `read-only` | Everywhere | `~/.kigi/` only | Blocked | Exploration, code review |
| `strict` | CWD + system paths | CWD + `/tmp` + `~/.kigi/` | Blocked | Untrusted code |
| `strict` | CWD + system paths | CWD + `/tmp` + `~/.kigi/` state dirs | Blocked | Untrusted code |
Sensitive paths (`~/.ssh/`, `~/.aws/`, `~/.gnupg/`, `~/.kigi/auth/`) are always
write-protected regardless of profile.
@@ -2159,7 +2160,7 @@ model cannot convince the agent to relax restrictions at runtime.
### Event Logging
Sandbox events (profile applied, violations) are logged to `~/.kigi/sandbox-events.jsonl`
Sandbox events (profile applied, violations) are logged to `~/.kigi/sessions/sandbox-events.jsonl`
for telemetry and debugging.
---
@@ -2384,6 +2385,7 @@ The agent persists all session updates automatically. Clients can reconnect and
| `KIGI_AGENT` | Custom agent definition path or name (see [Agent Profiles](#agent-profiles)) |
| `KIGI_WEB_FETCH` | Enable (`1`) or disable (`0`) the `web_fetch` tool |
| `KIGI_WEB_FETCH_PROXY` | Egress proxy URL for `web_fetch` requests (overridden by `[toolset.web_fetch] proxy_endpoint`) |
| `KIGI_WEB_FETCH_ALLOW_LOCAL` | `1` lets `web_fetch` reach an explicit loopback URL; private and metadata ranges stay blocked |
| `KIGI_RESPECT_GITIGNORE` | Disable `.gitignore` filtering in tools when set to `0` |
| `KIGI_FEEDBACK_ENABLED` | Enable (`1`) or disable (`0`) feedback system independently from telemetry |
| `KIGI_DEPLOYMENT_KEY` | Management API key for enterprise deployments |
crates/codegen/kigi-shell/src/agent/config.rs
+65 -2
View File
@@ -2171,6 +2171,19 @@ impl Config {
)
.with_segment_detail(self.resolve_compaction_detail())
}
/// Resolve summarizer `tool_choice`: env `KIGI_COMPACTION_TOOL_CHOICE` >
/// config > remote settings > default (`auto`).
pub(crate) fn resolve_compaction_tool_choice(
&self,
) -> crate::util::config::CompactionToolChoice {
crate::util::config::resolve_compaction_tool_choice_from(
env_string(crate::util::config::ENV_COMPACTION_TOOL_CHOICE).as_deref(),
self.features.compaction_tool_choice.as_deref(),
self.remote_settings
.as_ref()
.and_then(|r| r.compaction_tool_choice.as_deref()),
)
}
/// Resolve verbatim-input flag: env `KIGI_COMPACTION_VERBATIM_INPUT` > config > remote settings > default `true`.
pub(crate) fn resolve_compaction_verbatim_input(&self) -> bool {
BoolFlag::env("KIGI_COMPACTION_VERBATIM_INPUT")
@@ -2588,6 +2601,7 @@ pub(crate) fn resolve_model_list(
}
resolved = prefetched;
}
let mut explicit_api_backend_keys = std::collections::HashSet::new();
for (key, model_override) in &cfg.config_models {
let had_base = resolved.contains_key(key);
let base = resolved.shift_remove(key);
@@ -2604,6 +2618,9 @@ pub(crate) fn resolve_model_list(
);
}
}
if model_override.api_backend.is_some() {
explicit_api_backend_keys.insert(key.as_str());
}
let entry = model_override.apply(key, base, &cfg.endpoints);
tracing::debug!(
model_key = % key, base_url = % entry.info.base_url, has_api_key = entry
@@ -2625,7 +2642,7 @@ pub(crate) fn resolve_model_list(
)
})
.collect();
for entry in resolved.values_mut() {
for (key, entry) in resolved.iter_mut() {
if let Some((donor_cw, donor_backend)) = donors.get(&entry.info.model) {
if entry.info.context_window.get() == default_cw {
tracing::debug!(
@@ -2635,7 +2652,9 @@ pub(crate) fn resolve_model_list(
);
entry.info.context_window = *donor_cw;
}
if entry.info.api_backend == ApiBackend::default()
// An explicit `[model.X] api_backend` wins even when it is the default.
if !explicit_api_backend_keys.contains(key.as_str())
&& entry.info.api_backend == ApiBackend::default()
&& *donor_backend != ApiBackend::default()
{
entry.info.api_backend.clone_from(donor_backend);
@@ -3696,6 +3715,9 @@ pub struct Features {
/// Feed the summarizer the verbatim conversation instead of the lossy rewrite; `None` = defer to env/remote settings/default (true).
#[serde(default, skip_serializing_if = "Option::is_none")]
pub compaction_verbatim_input: Option<bool>,
/// Summarizer `tool_choice` (`auto` | `none`); `None` = defer to env/remote settings/default (`auto`).
#[serde(default, skip_serializing_if = "Option::is_none")]
pub compaction_tool_choice: Option<String>,
/// Snapshot a completed subagent's isolated worktree into a durable git ref
/// and delete its directory (resume rehydrates from the ref). This is the
/// per-deployment rollout lever (set in managed_config.toml `[features]`).
@@ -9051,6 +9073,47 @@ default = "kigi-4.5"
sibling 'kigi-build' (same model slug), not stay at {default_cw}"
);
}
/// An explicit `[model.X] api_backend` survives a same-slug sibling even when it is the default.
#[test]
fn slug_propagation_keeps_an_explicit_default_api_backend() {
// Only kigi-messages has a non-default context window: the slug's single donor.
let default_cw = DEFAULT_CONTEXT_WINDOW;
let raw: toml::Value = toml::from_str(&format!(
r#"
[model.kigi-messages]
model = "kigi-4.5"
context_window = 500000
base_url = "https://test.example.com/v1"
api_backend = "messages"
[model.kigi-chat]
model = "kigi-4.5"
context_window = {default_cw}
base_url = "https://test.example.com/v1"
api_backend = "chat_completions"
[model.kigi-plain]
model = "kigi-4.5"
context_window = {default_cw}
base_url = "https://test.example.com/v1"
"#
))
.unwrap();
let cfg = Config::new_from_toml_cfg(&raw).expect("config should parse");
let resolved = resolve_model_list(&cfg, None, &Default::default());
let backend = |key: &str| resolved.get(key).expect(key).info.api_backend.clone();
assert_eq!(
backend("kigi-chat"),
ApiBackend::ChatCompletions,
"explicit default backend must not be overwritten by the sibling"
);
assert_eq!(
backend("kigi-plain"),
ApiBackend::Messages,
"unset backend still inherits"
);
assert_eq!(backend("kigi-messages"), ApiBackend::Messages);
}
/// Slug propagation should carry over api_backend but NOT agent_type.
#[test]
fn slug_propagation_inherits_api_backend_but_not_agent_type() {
crates/codegen/kigi-shell/src/agent/feedback_client.rs
+1 -1
View File
@@ -3,7 +3,7 @@
//! Port of kimi-cli's `/feedback` slash command (kimi-cli
//! `src/kimi_cli/ui/shell/slash.py`, `feedback()`): subscription (OAuth)
//! sessions POST the user's feedback text to `{coding_api_base_url}/feedback`
//! with a Bearer token; everyone else is pointed at the GitHub issue tracker.
//! with a Bearer token; everyone else is pointed at the project issue tracker.
//! The request body carries exactly the fields kimi-cli sends:
//! `session_id`, `content`, `version`, `os`, `model`.
crates/codegen/kigi-shell/src/agent/handlers/model_switch.rs
+11 -5
View File
@@ -210,9 +210,9 @@ pub(crate) async fn apply(
auto_compact_threshold_percent: new_threshold,
responds_to: tx,
});
let updated_model = rx
let (updated_model, context_window) = rx
.await
.map_err(|_| acp::Error::internal_error().data("failed to set session model"))?;
.map_err(|_| acp::Error::internal_error().data("failed to set session model"))??;
if let Some(handle) = agent.sessions.borrow_mut().get_mut(&session_id) {
handle.model_id = model_id.clone();
handle.reasoning_effort = applied_effort;
@@ -224,6 +224,7 @@ pub(crate) async fn apply(
&session_id,
model_id.0.as_ref(),
applied_effort.map(|eff| eff.to_string()),
context_window.get(),
);
if agent.cfg.borrow().mode != config::AgentMode::Leader {
agent.models_manager.set_current_model_id(model_id);
@@ -232,9 +233,12 @@ pub(crate) async fn apply(
.set_current_reasoning_effort(applied_effort);
}
Ok(acp::SetSessionModelResponse::new().meta(
serde_json::json!({ "model" : updated_model, })
.as_object()
.cloned(),
serde_json::json!({
"model": updated_model,
"totalContextTokens": context_window.get(),
})
.as_object()
.cloned(),
))
}
/// Broadcast a `ModelChanged` to every client subscribed to this session so
@@ -245,12 +249,14 @@ fn broadcast_model_changed(
session_id: &acp::SessionId,
model_id: &str,
reasoning_effort: Option<String>,
context_window: u64,
) {
let notification = crate::extensions::notification::SessionNotification {
session_id: session_id.clone(),
update: crate::extensions::notification::SessionUpdate::ModelChanged {
model_id: model_id.to_owned(),
reasoning_effort,
context_window: Some(context_window),
},
meta: None,
};
crates/codegen/kigi-shell/src/agent/models.rs
+54 -41
View File
@@ -1938,7 +1938,7 @@ fn resolve_prefetch_env(kimi_code_config: Option<KimiCodeConfig>) -> Option<Pref
/// credentials from disk.
pub fn start_early_prefetch_with_auth(auth: Option<KimiAuth>) -> Option<EarlyPrefetchHandle> {
let env = resolve_prefetch_env_with_auth(auth)?;
Some(spawn_prefetch_thread(env))
spawn_prefetch_thread(env)
}
/// Start the model-catalog prefetch on a background thread.
@@ -1949,51 +1949,64 @@ pub fn start_early_prefetch(
kimi_code_config: Option<KimiCodeConfig>,
) -> Option<EarlyPrefetchHandle> {
let env = resolve_prefetch_env(kimi_code_config)?;
Some(spawn_prefetch_thread(env))
spawn_prefetch_thread(env)
}
fn spawn_prefetch_thread(env: PrefetchEnv) -> EarlyPrefetchHandle {
std::thread::spawn(move || {
let mut timer = crate::instrumentation_timer!("startup.early_prefetch");
let proxy_endpoint = env.endpoints.proxy_url();
timer.with_field("endpoint", proxy_endpoint.as_str());
// Resolve each stored subscription-OAuth session's bearer (refreshed
// on expiry) so those platforms are part of the STARTUP fetch plan —
// otherwise a claude-pro-max-only user boots onto the bundled Kimi
// table until some later async refresh happens to run, and the cache
// origin (which encodes enabled platforms) never matches the
// async-path's claude-inclusive origin.
let oauth_tokens = tokio::runtime::Builder::new_current_thread()
.enable_all()
.build()
.map(|rt| {
rt.block_on(crate::agent::models_fetch::resolve_generic_oauth_tokens(
&crate::auth::oauth_registry::pool_home(),
))
})
.unwrap_or_default();
let models = prefetch_models_blocking(
&env.endpoints,
env.auth.as_ref(),
&oauth_tokens,
env.model_fetch_auth,
&env.platform_keys,
);
if (env.endpoints.deployment_key.is_some() || crate::managed_config::has_active_team_auth())
&& crate::config::is_managed_config_stale_for(
&crate::managed_config::current_serving_identity(),
)
&& crate::managed_config::is_fetch_enabled()
&& let Ok(rt) = tokio::runtime::Builder::new_current_thread()
/// `None` when the OS refuses a thread: the prefetch is an optimization and
/// runs before the runtime builds, so a spawn panic here would preempt the
/// runtime's own out-of-threads diagnostic.
fn spawn_prefetch_thread(env: PrefetchEnv) -> Option<EarlyPrefetchHandle> {
let spawned = std::thread::Builder::new()
.name("early-prefetch".into())
.spawn(move || {
let mut timer = crate::instrumentation_timer!("startup.early_prefetch");
let proxy_endpoint = env.endpoints.proxy_url();
timer.with_field("endpoint", proxy_endpoint.as_str());
// Resolve each stored subscription-OAuth session's bearer (refreshed
// on expiry) so those platforms are part of the STARTUP fetch plan —
// otherwise a claude-pro-max-only user boots onto the bundled Kimi
// table until some later async refresh happens to run, and the cache
// origin (which encodes enabled platforms) never matches the
// async-path's claude-inclusive origin.
let oauth_tokens = tokio::runtime::Builder::new_current_thread()
.enable_all()
.build()
{
crate::managed_config::clear_orphan();
let _ = rt.block_on(crate::managed_config::sync());
}
.map(|rt| {
rt.block_on(crate::agent::models_fetch::resolve_generic_oauth_tokens(
&crate::auth::oauth_registry::pool_home(),
))
})
.unwrap_or_default();
let models = prefetch_models_blocking(
&env.endpoints,
env.auth.as_ref(),
&oauth_tokens,
env.model_fetch_auth,
&env.platform_keys,
);
if (env.endpoints.deployment_key.is_some()
|| crate::managed_config::has_active_team_auth())
&& crate::config::is_managed_config_stale_for(
&crate::managed_config::current_serving_identity(),
)
&& crate::managed_config::is_fetch_enabled()
&& let Ok(rt) = tokio::runtime::Builder::new_current_thread()
.enable_all()
.build()
{
crate::managed_config::clear_orphan();
let _ = rt.block_on(crate::managed_config::sync());
}
EarlyPrefetchResult { models }
})
EarlyPrefetchResult { models }
});
match spawned {
Ok(handle) => Some(handle),
Err(e) => {
tracing::error!(error = %e, "early prefetch thread spawn failed; skipping prefetch");
None
}
}
}
/// Map a model id (catalog key or routing slug) to its catalog key.
crates/codegen/kigi-shell/src/agent/models_fetch.rs
+4 -4
View File
@@ -1475,7 +1475,7 @@ mod tests {
/// short-circuits BEFORE any HTTP — there is NO mock `/models` server, yet
/// the fetch returns exactly the 4 compiled-in models keyed
/// `openai-codex/<slug>` on the Responses backend, ctx 272000, each exposing
/// its exact reasoning efforts (incl. the codex-only `xhigh`/`max`/`ultra`).
/// its exact reasoning efforts (incl. the codex-only `xhigh`/`max`).
/// A BOGUS base URL confirms no live `/models` request is attempted (it would
/// otherwise fail against an unroutable host).
#[tokio::test(flavor = "multi_thread")]
@@ -1556,10 +1556,10 @@ mod tests {
.iter()
.map(|o| o.id.as_str())
.collect::<Vec<_>>(),
vec!["low", "medium", "high", "xhigh", "max", "ultra"],
"sol exposes the full codex effort menu incl. ultra"
vec!["low", "medium", "high", "xhigh", "max"],
"sol exposes the full codex effort menu"
);
// gpt-5.5 tops out at xhigh (no max/ultra).
// gpt-5.5 tops out at xhigh (no max).
let five_five = result
.models
.iter()
crates/codegen/kigi-shell/src/agent/mvp_agent/acp_agent.rs
+8 -3
View File
@@ -1725,8 +1725,8 @@ impl acp::Agent for MvpAgent {
return Ok(acp::PromptResponse::new(acp::StopReason::EndTurn));
}
}
let intake_lock = self.prompt_intake_lock(&arguments.session_id);
let intake_guard = intake_lock.lock().await;
let dispatch_lock = self.dispatch_lock(&arguments.session_id);
let dispatch_guard = dispatch_lock.lock().await;
let meta_prompt_mode = arguments
.meta
.as_ref()
@@ -1830,7 +1830,7 @@ impl acp::Agent for MvpAgent {
acp::Error::internal_error()
.data(format!("failed to dispatch prompt to session: {e}"))
})?;
drop(intake_guard);
drop(dispatch_guard);
self.push_roster_activity_delta(
&arguments.session_id,
crate::agent::roster::RosterActivity::Working,
@@ -2025,6 +2025,11 @@ impl acp::Agent for MvpAgent {
.and_then(|m| m.get("rewindIfPristine"))
.and_then(|v| v.as_bool())
.unwrap_or(false);
// Ordered behind any in-flight prompt intake: a cancel that
// overtakes the prompt it targets lands on an idle actor, and the
// prompt then starts a turn nothing will ever cancel.
let dispatch_lock = self.dispatch_lock(&args.session_id);
let _dispatch_guard = dispatch_lock.lock().await;
let _ = handle
.cmd_tx
.send(SessionCommand::Cancel {
crates/codegen/kigi-shell/src/agent/mvp_agent/agent_ops.rs
+18 -3
View File
@@ -357,6 +357,11 @@ impl MvpAgent {
// time, so advertise pre-session; the in-session path
// re-checks the live toolset.
graph: goal && self.cfg.borrow().resolve_graph().value,
// Tool-dependent, so fail closed like every other tool gate: the
// builder strips `agent_swarm` whenever subagents are unavailable,
// and advertising it then offers a menu entry that resolves to
// literal prompt text.
swarm: self.cfg.borrow().subagents_enabled,
..crate::session::slash_commands::CommandAvailability::default()
}
}
@@ -973,7 +978,10 @@ impl MvpAgent {
sessions: RefCell::new(HashMap::new()),
activity,
loading_sessions: RefCell::new(HashMap::new()),
prompt_intake_locks: RefCell::new(HashMap::new()),
dispatch_locks: RefCell::new(HashMap::new()),
subagent_limits:
kigi_tools::implementations::kigi::task::admission::SubagentLimits::from_env(),
subagent_spawn_permits: RefCell::new(HashMap::new()),
session_threads: RefCell::new(HashMap::new()),
resident_roster_titles: RefCell::new(HashMap::new()),
initialize_request: OnceLock::new(),
@@ -1357,7 +1365,8 @@ impl MvpAgent {
/// `FlushAndAck` on the persistence actor — a true sync barrier that only
/// resolves after all queued writes (chat messages, updates) hit disk.
///
/// Returns `Ok(())` on success, `Err(reason)` on timeout or channel failure.
/// Returns `Ok(())` once the writes are on disk, `Err(reason)` on a failed
/// flush, timeout or channel failure.
pub(crate) async fn flush_session(
&self,
session_id: &acp::SessionId,
@@ -1376,7 +1385,11 @@ impl MvpAgent {
return Err("send failed");
}
match tokio::time::timeout(std::time::Duration::from_secs(5), rx).await {
Ok(Ok(_)) => Ok(()),
Ok(Ok(Ok(()))) => Ok(()),
Ok(Ok(Err(error))) => {
tracing::error!(session_id = %session_id.0, %error, "persistence flush failed");
Err("persistence flush failed")
}
Ok(Err(_)) => Err("channel closed"),
Err(_) => Err("timeout"),
}
@@ -2178,6 +2191,7 @@ impl MvpAgent {
.cfg
.borrow()
.resolve_compaction_verbatim_input();
let compaction_tool_choice = self.cfg.borrow().resolve_compaction_tool_choice();
let two_pass_enabled = self.cfg.borrow().is_two_pass_compaction_enabled();
let auto_update = self.cfg.borrow().cli.auto_update;
let client_type = *self.client_type.borrow();
@@ -2563,6 +2577,7 @@ impl MvpAgent {
system_prompt_label,
compaction_mode,
compaction_verbatim_input,
compaction_tool_choice,
two_pass_enabled,
buffering_settings,
origin_client.clone(),
crates/codegen/kigi-shell/src/agent/mvp_agent/mod.rs
+34 -17
View File
@@ -462,21 +462,30 @@ pub struct MvpAgent {
loading_sessions: RefCell<
HashMap<acp::SessionId, tokio::sync::watch::Receiver<bool>>,
>,
/// Per-session prompt-intake serialization lock. LEADER-SAFE(per-session):
/// keyed by SessionId, mirrors `sessions` lifecycle.
///
/// Each incoming `session/prompt` RPC is dispatched as its own task by the
/// ACP message loop, and [`Self::prompt`] runs an async preamble (prompt-mode
/// query, trace context, model lookup) BEFORE it enqueues
/// `SessionCommand::Prompt` onto the actor's FIFO mailbox. Without
/// serialization those preambles interleave across tasks, so the mailbox —
/// and therefore the authoritative prompt queue — receives prompts out of
/// submission order. `prompt()` holds this lock across the preamble and
/// releases it immediately after the enqueue (the turn itself runs unlocked),
/// which makes intake order match arrival order.
prompt_intake_locks: RefCell<
/// Per-session lock ordering dispatch onto the actor's mailbox:
/// [`Self::prompt`] holds it across its intake preamble (prompt-mode
/// query, trace context, model lookup) and releases it right after the
/// enqueue, so prompts land in submission order; `cancel` holds it
/// around its `Cancel` send, so a cancel cannot overtake the prompt it
/// targets and wedge the turn slot (see
/// `cancel_never_overtakes_in_flight_prompt_intake`). Cancels wait out
/// preambles held ahead of them — keep preambles lean.
/// LEADER-SAFE(per-session): mirrors `sessions` lifecycle.
dispatch_locks: RefCell<
HashMap<acp::SessionId, std::rc::Rc<tokio::sync::Mutex<()>>>,
>,
/// Session-scoped subagent spawn limits, read once from env at
/// construction (composition root).
subagent_limits: kigi_tools::implementations::kigi::task::admission::SubagentLimits,
/// Per-parent-session spawn semaphores sized to `subagent_limits`; a
/// wide fan-out queues on the semaphore instead of exhausting file
/// descriptors. Keyed by session id string (the coordinator's
/// `parent_session_id`). Entries deliberately live for the whole
/// process: a removed-then-resumed session keeps its id, so sharing one
/// semaphore across generations keeps the limit correct while old
/// children still run. Growth is bounded by distinct sessions per
/// process and each entry is tiny. LEADER-SAFE(per-session).
subagent_spawn_permits: RefCell<HashMap<String, std::sync::Arc<tokio::sync::Semaphore>>>,
/// LEADER-SAFE(per-session): keyed by SessionId. Mirrors `sessions` lifecycle.
session_threads: RefCell<HashMap<acp::SessionId, SessionThread>>,
/// Title per resident session id, refreshed each `build_roster`. Lets the
@@ -717,10 +726,18 @@ pub struct MvpAgent {
/// sampling client is discarded, but the TLS root certificates it loads
/// are cached at the process level by `rustls-native-certs`.
pub fn warm_async_http_client() {
std::thread::spawn(|| {
let _timer = crate::instrumentation_timer!("startup.async_http_warmup");
let _ = crate::http::shared_client();
});
// Recoverable spawn: the warmup is an optimization that runs before the
// runtime builds — an out-of-threads host must reach the runtime's own
// diagnostic, not panic here.
let spawned = std::thread::Builder::new()
.name("http-warmup".into())
.spawn(|| {
let _timer = crate::instrumentation_timer!("startup.async_http_warmup");
let _ = crate::http::shared_client();
});
if let Err(e) = spawned {
tracing::warn!(error = %e, "async HTTP warmup thread spawn failed; first request warms instead");
}
}
pub(crate) fn resolve_required_agent_type(
model_agent_type: Option<&str>,
crates/codegen/kigi-shell/src/agent/mvp_agent/session_lifecycle.rs
+5 -8
View File
@@ -38,19 +38,16 @@ impl MvpAgent {
/// drops the `session_live_state` entry so that map stays bounded.
pub(crate) fn remove_session(&self, id: &acp::SessionId) {
self.sessions.borrow_mut().remove(id);
self.prompt_intake_locks.borrow_mut().remove(id);
self.dispatch_locks.borrow_mut().remove(id);
self.session_threads.borrow_mut().remove(id);
self.session_index_claims.borrow_mut().remove(id);
self.require_gateway_sessions.borrow_mut().remove(id);
self.session_live_state.borrow_mut().remove(id);
}
/// Get-or-create the per-session prompt-intake lock (see
/// [`Self::prompt_intake_locks`]).
pub(super) fn prompt_intake_lock(
&self,
id: &acp::SessionId,
) -> std::rc::Rc<tokio::sync::Mutex<()>> {
self.prompt_intake_locks
/// Get-or-create the per-session dispatch lock (see
/// [`Self::dispatch_locks`]). Cheap clone of the shared `Rc`.
pub(super) fn dispatch_lock(&self, id: &acp::SessionId) -> std::rc::Rc<tokio::sync::Mutex<()>> {
self.dispatch_locks
.borrow_mut()
.entry(id.clone())
.or_default()
crates/codegen/kigi-shell/src/agent/mvp_agent/subagent_coordinator.rs
+15
View File
@@ -272,6 +272,19 @@ impl MvpAgent {
self.try_build_subagent_spawn_context(parent_session_id)
.expect("parent session must exist when spawning subagents")
}
/// Get-or-create the parent session's spawn semaphore (see
/// [`Self::subagent_spawn_permits`]). Cheap clone of the shared `Arc`.
fn subagent_spawn_permit(&self, parent_session_id: &str) -> std::sync::Arc<tokio::sync::Semaphore> {
self.subagent_spawn_permits
.borrow_mut()
.entry(parent_session_id.to_string())
.or_insert_with(|| {
std::sync::Arc::new(tokio::sync::Semaphore::new(
self.subagent_limits.effective_max_concurrent(),
))
})
.clone()
}
/// Fallible variant of [`Self::build_subagent_spawn_context`]: returns
/// `None` when the parent `SessionHandle` is absent (evicted / torn down)
/// instead of panicking, so read-only paths that can race a teardown can
@@ -387,6 +400,8 @@ impl MvpAgent {
};
Some(crate::agent::subagent::SubagentSpawnContext {
lsp: parent_lsp,
spawn_limits: self.subagent_limits,
spawn_permits: self.subagent_spawn_permit(parent_session_id),
gateway: self.gateway.clone(),
client_hooks: Default::default(),
sampling_config: self.sampling_config.borrow().clone(),
crates/codegen/kigi-shell/src/agent/mvp_agent/tests.rs
+53
View File
@@ -2152,6 +2152,59 @@ fn cancel_does_not_forward_to_bridge_in_local_mode() {
);
});
}
/// Regression (post-cancel slot hang): a cancel racing the prompt it targets
/// must land on the actor mailbox AFTER that prompt, or it hits an idle
/// actor as a no-op and the prompt then starts a turn nothing cancels.
/// The dispatch lock (`dispatch_locks`) provides that ordering.
#[test]
fn cancel_never_overtakes_in_flight_prompt_intake() {
use crate::session::SessionCommand;
use acp::Agent as _;
run_local_for_bridge_test(|| async {
let agent = build_minimal_agent_for_tests();
let sid = acp::SessionId::new("sess-cancel-intake-race");
let (handle, _tx, mut cmd_rx) = make_live_session_handle(&sid, None);
agent.sessions.borrow_mut().insert(sid.clone(), handle);
let order: std::rc::Rc<std::cell::RefCell<Vec<&'static str>>> =
std::rc::Rc::new(std::cell::RefCell::new(Vec::new()));
let (intake_parked_tx, intake_parked_rx) = tokio::sync::oneshot::channel::<()>();
let driver_order = order.clone();
tokio::task::spawn_local(async move {
let mut intake_parked_tx = Some(intake_parked_tx);
while let Some(cmd) = cmd_rx.recv().await {
match cmd {
SessionCommand::GetCurrentPromptMode { .. } => {
if let Some(tx) = intake_parked_tx.take() {
let _ = tx.send(());
tokio::time::sleep(std::time::Duration::from_millis(10)).await;
}
}
SessionCommand::Prompt { .. } => driver_order.borrow_mut().push("prompt"),
SessionCommand::Cancel { .. } => driver_order.borrow_mut().push("cancel"),
_ => {}
}
}
});
let prompt_fut = agent.prompt(acp::PromptRequest::new(
sid.clone(),
vec![acp::ContentBlock::from("hi")],
));
let cancel_fut = async {
intake_parked_rx
.await
.expect("prompt intake reaches the fake actor");
let _ = agent
.cancel(acp::CancelNotification::new(sid.clone()))
.await;
};
let _ = futures::join!(prompt_fut, cancel_fut);
assert_eq!(
order.borrow().as_slice(),
["prompt", "cancel"],
"cancel must land on the actor mailbox after the prompt it targets"
);
});
}
use crate::session::SessionCommand as TestSessionCommand;
/// Build a session handle wired to a *live* command channel. Returns the
/// handle (move into `sessions`) plus a probe `cmd_tx`/`cmd_rx` so a test
crates/codegen/kigi-shell/src/agent/mvp_agent/tests/subagent_spawn_context_tests.rs
+3 -1
View File
@@ -27,7 +27,7 @@ async fn subagent_spawn_context_inherits_parent_permission_handle() {
kigi_workspace::permission::spawn_permission_manager(
sid.clone(),
gateway,
cwd,
cwd.clone(),
kigi_workspace::permission::types::ClientType::Generic,
Some(PermissionConfig::new(vec![PermissionRule {
action: RuleAction::Deny,
@@ -64,6 +64,8 @@ async fn subagent_spawn_context_inherits_parent_permission_handle() {
Some("child-session".to_owned()),
Some("general-purpose".to_owned()),
Some("permission inheritance regression".to_owned()),
cwd.clone(),
None,
)
.await;
assert!(
crates/codegen/kigi-shell/src/agent/session_config.rs
-1
View File
@@ -60,7 +60,6 @@ fn effort_label(effort: ReasoningEffort) -> String {
ReasoningEffort::High => "High",
ReasoningEffort::Xhigh => "X-High",
ReasoningEffort::Max => "Max",
ReasoningEffort::Ultra => "Ultra",
}
.to_string()
}
crates/codegen/kigi-shell/src/agent/subagent/handle_request.rs
+276 -101
View File
@@ -19,6 +19,99 @@ use kigi_tools::implementations::kigi::task::types::*;
use kigi_workspace::file_system::AsyncFileSystem;
use kigi_hunk_tracker::HunkTrackerHandle;
use super::*;
/// Bounds each parent-side await in the child completion path so a starved parent `cmd_rx` cannot park a completed child forever.
pub(super) const PARENT_ACK_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(30);
/// Bounds reads of the child's own actors, which freeze with a tool that blocks the child thread; teardown must still reach `Shutdown`.
pub(super) const CHILD_ACTOR_ACK_TIMEOUT: std::time::Duration =
std::time::Duration::from_secs(10);
/// One bounded child-actor query, degrading to `fallback` when the actor never answers.
pub(super) async fn child_actor_query<T>(
what: &'static str,
query: impl std::future::Future<Output = T>,
fallback: T,
) -> T {
match tokio::time::timeout(CHILD_ACTOR_ACK_TIMEOUT, query).await {
Ok(value) => value,
Err(_) => {
tracing::warn!(query = what, "child actor did not answer in time; using fallback");
fallback
}
}
}
/// Bounds the post-Shutdown worktree snapshot (a `git add` behind a stuck clean filter must not park completion).
const WORKTREE_SNAPSHOT_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(120);
/// Snapshot the child worktree to `ref_name`, persist the ref, then remove the worktree; returns `(snapshot_ref, removed)`.
async fn dispose_subagent_worktree(
worktree: &Path,
source_repo: &Path,
ref_name: &str,
meta_dir: &Path,
final_status: &str,
subagent_id: &str,
) -> (Option<String>, bool) {
let snapshot = tokio::time::timeout(
WORKTREE_SNAPSHOT_TIMEOUT,
crate::session::worktree::snapshot_subagent_worktree(worktree, source_repo, ref_name),
)
.await;
let snapshot_ref = match snapshot {
Ok(Ok(snapshot_ref)) => snapshot_ref,
Ok(Err(e)) => {
tracing::warn!(
subagent_id = % subagent_id, worktree_path = % worktree.display(), error = % e,
"Failed to snapshot subagent worktree; preserving for review"
);
return (None, false);
}
Err(_) => {
tracing::warn!(
subagent_id = % subagent_id, worktree_path = % worktree.display(),
"worktree snapshot did not finish in time; preserving worktree for resume"
);
return (None, false);
}
};
if !update_subagent_meta_snapshot_ref(meta_dir, &snapshot_ref, final_status) {
tracing::warn!(
subagent_id = % subagent_id, worktree_path = % worktree.display(),
"snapshot_ref not persisted; preserving worktree for resume"
);
return (None, false);
}
// Removal is awaited in full: a detached deletion could race a resume that rehydrates into this path.
match crate::session::worktree::remove_subagent_worktree(worktree).await {
Ok(()) => {
tracing::info!(
subagent_id = % subagent_id, worktree_path = % worktree.display(),
"snapshotted and removed subagent worktree"
);
(Some(snapshot_ref), true)
}
Err(e) => {
tracing::warn!(
subagent_id = % subagent_id, worktree_path = % worktree.display(), error = % e,
"snapshotted subagent worktree but removal failed; ref persisted for resume"
);
(Some(snapshot_ref), false)
}
}
}
/// Await a parent ack for at most `PARENT_ACK_TIMEOUT`; `false` when it never answers or drops the sender.
pub(super) async fn parent_ack(what: &'static str, ack: oneshot::Receiver<()>) -> bool {
match tokio::time::timeout(PARENT_ACK_TIMEOUT, ack).await {
Ok(Ok(())) => true,
Ok(Err(_)) => false,
Err(_) => {
tracing::warn!(ack = what, "parent did not ack in time; proceeding to teardown");
false
}
}
}
pub(super) fn task_model_override_error(
requested: Option<&str>,
provenance: ModelOverrideProvenance,
@@ -81,6 +174,64 @@ pub(crate) async fn handle_subagent_request(
}
_ => {}
}
// Session-scoped concurrency admission: hold one permit for the whole
// child run so a wide fan-out queues (default) or fails (configured)
// instead of exhausting file descriptors. Queued spawns are not yet
// visible in the coordinator; they admit in FIFO order.
use kigi_tools::implementations::kigi::task::admission::LimitBehavior;
let _spawn_permit = match ctx.spawn_limits.behavior {
LimitBehavior::Fail => match ctx.spawn_permits.clone().try_acquire_owned() {
Ok(permit) => permit,
Err(_) => {
let limit = ctx.spawn_limits.effective_max_concurrent();
kigi_log::unified_log::warn(
"subagent.spawn.rejected_at_concurrent_limit",
Some(&ctx.parent_session_id),
Some(serde_json::json!({ "limit": limit })),
);
let msg = format!(
"Session subagent limit reached ({limit} running); \
retry after one finishes, or raise KIGI_MAX_CONCURRENT_SUBAGENTS"
);
send_pre_spawn_failure(request, &msg, coordinator, &ctx, gateway);
return;
}
},
LimitBehavior::Queue => {
if ctx.spawn_permits.available_permits() == 0 {
kigi_log::unified_log::warn(
"subagent.spawn.queued_at_concurrent_limit",
Some(&ctx.parent_session_id),
Some(serde_json::json!({
"limit": ctx.spawn_limits.effective_max_concurrent(),
"subagent_id": request.id,
})),
);
}
ctx.spawn_permits
.clone()
.acquire_owned()
.await
.expect("subagent spawn semaphore is never closed")
}
};
// The parent can end while a spawn is queued on the semaphore (a queued
// request is not yet in the coordinator, so the teardown cancel sweep
// cannot see it). A closed parent command channel is the teardown
// signal. The channel closes when the actor finishes shutting down, so a
// permit granted during that brief drain can still slip through — the
// same in-flight window an ordinary spawn racing teardown already has;
// closing it fully would require queueing inside the coordinator.
if ctx.parent_cmd_tx.as_ref().is_some_and(|tx| tx.is_closed()) {
send_pre_spawn_failure(
request,
"Parent session ended while the spawn was queued at the subagent limit.",
coordinator,
&ctx,
gateway,
);
return;
}
let run_in_background = request.run_in_background
|| definition.background.unwrap_or(false);
let cancel_token = CancellationToken::new();
@@ -396,11 +547,14 @@ pub(crate) async fn handle_subagent_request(
let child_depth = ctx.parent_depth + 1;
if child_depth >= MAX_SUBAGENT_DEPTH {
let before = definition.tool_config.tools.len();
definition.tool_config.tools.retain(|tc| tc.kind != Some(ToolKind::Task));
definition
.tool_config
.tools
.retain(|tc| !matches!(tc.kind, Some(ToolKind::Task | ToolKind::AgentSwarm)));
if definition.tool_config.tools.len() < before {
tracing::info!(
subagent_id = % request.id, child_depth, max_depth =
MAX_SUBAGENT_DEPTH, "Stripped task tool from child at max depth"
MAX_SUBAGENT_DEPTH, "Stripped subagent-spawning tools from child at max depth"
);
}
prune_orphaned_background_task_tools(&mut definition.tool_config);
@@ -834,13 +988,22 @@ pub(crate) async fn handle_subagent_request(
}
}
}
let agent_mcp_servers: Vec<_> = if is_plugin_agent {
if !definition.mcp_servers.is_empty() {
tracing::warn!(
agent = % definition.name, plugin = ? definition.plugin_name,
"ignoring mcpServers on plugin agent (not supported for security)"
);
}
let agent_mcp_servers: Vec<_> = if definition.mcp_servers.is_empty() {
vec![]
} else if is_plugin_agent {
tracing::warn!(
agent = % definition.name, plugin = ? definition.plugin_name,
"ignoring mcpServers on plugin agent (not supported for security)"
);
vec![]
} else if !crate::agent::folder_trust::agent_inline_hooks_allowed(definition.scope, || {
crate::agent::folder_trust::project_scope_allowed(&ctx.parent_cwd)
}) {
// An untrusted repo's inline stdio MCP command would run like an inline hook.
tracing::warn!(
agent = % definition.name,
"ignoring mcpServers on untrusted project agent (folder not trusted; re-run with --trust)"
);
vec![]
} else {
definition
@@ -1003,6 +1166,7 @@ pub(crate) async fn handle_subagent_request(
kigi_agent::DEFAULT_SYSTEM_PROMPT_LABEL.to_string(),
kigi_chat_state::CompactionMode::Summary,
ctx.resolve_compaction_verbatim_input(),
ctx.resolve_compaction_tool_choice(),
false,
None,
None,
@@ -1271,7 +1435,12 @@ pub(crate) async fn handle_subagent_request(
let duration_ms = start.elapsed().as_millis() as u64;
let mut result = match wait_outcome {
SubagentWaitOutcome::Cancelled => {
let (tool_calls, turns) = signals_snapshot_counts(&child_handle).await;
let (tool_calls, turns) = child_actor_query(
"signals_snapshot_counts",
signals_snapshot_counts(&child_handle),
(0, 0),
)
.await;
SubagentResult {
success: false,
cancelled: true,
@@ -1298,14 +1467,28 @@ pub(crate) async fn handle_subagent_request(
},
),
) => (snapshot.current.tool_call_count, snapshot.current.turn_count),
_ => signals_snapshot_counts(&child_handle).await,
_ => {
child_actor_query(
"signals_snapshot_counts",
signals_snapshot_counts(&child_handle),
(0, 0),
)
.await
}
};
let final_text = child_handle
.chat_state_handle
.get_last_assistant_text()
.await
.unwrap_or_default();
let result_tokens = child_handle.chat_state_handle.get_total_tokens().await;
let final_text = child_actor_query(
"get_last_assistant_text",
child_handle.chat_state_handle.get_last_assistant_text(),
None,
)
.await
.unwrap_or_default();
let result_tokens = child_actor_query(
"get_total_tokens",
child_handle.chat_state_handle.get_total_tokens(),
0,
)
.await;
match *turn_result {
Ok(
Ok(
@@ -1322,6 +1505,7 @@ pub(crate) async fn handle_subagent_request(
SubagentResult {
success: false,
cancelled: true,
rate_limited: false,
error: Some(reason),
output: if final_text.is_empty() {
std::sync::Arc::from(
@@ -1359,6 +1543,7 @@ pub(crate) async fn handle_subagent_request(
SubagentResult {
success: false,
cancelled: true,
rate_limited: false,
error: Some(format!("max turns reached (limit: {limit})")),
output: if final_text.is_empty() {
std::sync::Arc::from(
@@ -1413,6 +1598,9 @@ pub(crate) async fn handle_subagent_request(
SubagentResult {
success: false,
cancelled: was_cancelled,
rate_limited: !was_cancelled
&& i32::from(e.code)
== crate::sampling::error::RATE_LIMITED_ERROR_CODE,
error: Some(
if was_cancelled {
"Subagent was cancelled".to_string()
@@ -1460,14 +1648,21 @@ pub(crate) async fn handle_subagent_request(
let final_status = result.status().to_string();
let snapshot_dispose_enabled = ctx.resolve_subagent_worktree_snapshot_enabled();
let telemetry_tokens = if result.tool_calls > 0 || result.success {
child_handle.chat_state_handle.get_total_tokens().await
child_actor_query(
"get_total_tokens",
child_handle.chat_state_handle.get_total_tokens(),
0,
)
.await
} else {
0
};
let (subagent_usage_by_model, subagent_usage_incomplete) = match child_handle
.chat_state_handle
.try_get_session_usage()
.await
let (subagent_usage_by_model, subagent_usage_incomplete) = match child_actor_query(
"try_get_session_usage",
child_handle.chat_state_handle.try_get_session_usage(),
Err(()),
)
.await
{
Ok(u) => (Some(u.by_model.into_iter().collect::<Vec<_>>()), u.incomplete),
Err(()) => (None, true),
@@ -1486,7 +1681,7 @@ pub(crate) async fn handle_subagent_request(
respond_to,
})
{
Ok(()) => ack.await.is_ok(),
Ok(()) => parent_ack("RecordSubagentUsage", ack).await,
Err(_) => false,
}
} else {
@@ -1503,53 +1698,71 @@ pub(crate) async fn handle_subagent_request(
.parent_prompt_id
.clone()
.or_else(|| coordinator.borrow().parent_prompt_id_for(&request.id));
if let Some(cmd_tx) = ctx.parent_cmd_tx.as_ref() {
// A slow parent may apply both the fold and this mark; both are idempotent.
let marked_by_parent = if let Some(cmd_tx) = ctx.parent_cmd_tx.as_ref() {
let (respond_to, ack) = tokio::sync::oneshot::channel();
if cmd_tx
cmd_tx
.send(crate::session::commands::SessionCommand::MarkSubagentUsageNotApplied {
parent_prompt_id: sticky_prompt,
parent_prompt_id: sticky_prompt.clone(),
respond_to,
})
.is_ok()
{
let _ = ack.await;
}
} else if let Some(ref pid) = sticky_prompt {
&& parent_ack("MarkSubagentUsageNotApplied", ack).await
} else {
false
};
if !marked_by_parent && let Some(ref pid) = sticky_prompt {
coordinator.borrow_mut().mark_subagent_usage_not_applied(pid);
}
}
match (&ctx.parent_terminal_backend, &ctx.parent_notification_handle) {
(Some(parent_tb), Some(parent_notif_handle)) => {
if !request.surface_completion {
let reparented_task_ids: Vec<String> = parent_tb
.list_tasks()
.await
.into_iter()
.filter(|t| {
!t.completed
&& t.owner_session_id.as_deref()
== Some(&*child_session_id.0)
})
.map(|t| t.task_id)
.collect();
if !reparented_task_ids.is_empty()
&& let Some(cmd_tx) = ctx.parent_cmd_tx.as_ref()
{
let _ = cmd_tx
.send(SessionCommand::RecordGoalTurnTaskIds {
task_ids: reparented_task_ids,
});
match tokio::time::timeout(PARENT_ACK_TIMEOUT, parent_tb.list_tasks()).await {
Ok(tasks) => {
let reparented_task_ids: Vec<String> = tasks
.into_iter()
.filter(|t| {
!t.completed
&& t.owner_session_id.as_deref()
== Some(&*child_session_id.0)
})
.map(|t| t.task_id)
.collect();
if !reparented_task_ids.is_empty()
&& let Some(cmd_tx) = ctx.parent_cmd_tx.as_ref()
{
let _ = cmd_tx.send(SessionCommand::RecordGoalTurnTaskIds {
task_ids: reparented_task_ids,
});
}
}
Err(_) => tracing::warn!(
child_session_id = % child_session_id.0,
parent_session_id = % ctx.parent_session_id,
"list_tasks not answered in time; skipping goal-turn task-id recording"
),
}
}
let parent_backend_weak = std::sync::Arc::downgrade(parent_tb);
parent_tb
.reparent_notifications(
if tokio::time::timeout(
PARENT_ACK_TIMEOUT,
parent_tb.reparent_notifications(
&child_session_id.0,
&ctx.parent_session_id,
parent_notif_handle.clone(),
parent_backend_weak,
)
.await;
),
)
.await
.is_err()
{
tracing::warn!(
child_session_id = % child_session_id.0,
parent_session_id = % ctx.parent_session_id,
"reparent_notifications not acked in time; proceeding to child shutdown"
);
}
}
(Some(_), None) | (None, Some(_)) => {
tracing::warn!(
@@ -1570,55 +1783,17 @@ pub(crate) async fn handle_subagent_request(
if snapshot_dispose_enabled {
let ref_name = format!("refs/kigi/subagents/{}", request.id);
let source_repo = resolve_subagent_source_repo(&ctx);
match crate::session::worktree::snapshot_subagent_worktree(
wt_path,
&source_repo,
&ref_name,
)
.await
{
Ok(snapshot_ref) => {
let persisted = update_subagent_meta_snapshot_ref(
&subagent_meta_dir,
&snapshot_ref,
&final_status,
);
if persisted {
disposed_snapshot_ref = Some(snapshot_ref);
match crate::session::worktree::remove_subagent_worktree(wt_path)
.await
{
Ok(()) => {
worktree_removed = true;
tracing::info!(
subagent_id = % request.id, worktree_path = % wt_path
.display(), "snapshotted and removed subagent worktree"
);
}
Err(e) => {
tracing::warn!(
subagent_id = % request.id, worktree_path = % wt_path
.display(), error = % e,
"snapshotted subagent worktree but removal failed; ref persisted for resume"
)
}
}
} else {
tracing::warn!(
subagent_id = % request.id, worktree_path = % wt_path
.display(),
"snapshot_ref not persisted; preserving worktree for resume"
);
}
}
Err(e) => {
tracing::warn!(
subagent_id = % request.id, worktree_path = % wt_path.display(),
error = % e,
"Failed to snapshot subagent worktree; preserving for review"
);
}
}
let (snapshot_ref, removed) = dispose_subagent_worktree(
wt_path,
&source_repo,
&ref_name,
&subagent_meta_dir,
&final_status,
&request.id,
)
.await;
disposed_snapshot_ref = snapshot_ref;
worktree_removed = removed;
} else {
tracing::info!(
subagent_id = % request.id, worktree_path = % wt_path.display(),
crates/codegen/kigi-shell/src/agent/subagent/mod.rs
+22
View File
@@ -133,6 +133,13 @@ impl AutoCompactThresholdTiers {
pub(crate) struct SubagentSpawnContext {
/// Parent's LSP runtime — inherited via ToolContext, same as fs/terminal.
pub lsp: Option<std::sync::Arc<dyn kigi_tools::implementations::lsp::LspBackend>>,
/// Session-scoped spawn limits (env-resolved once at the composition root).
pub spawn_limits: kigi_tools::implementations::kigi::task::admission::SubagentLimits,
/// Parent session's spawn semaphore, sized to
/// `spawn_limits.effective_max_concurrent()`. A permit is held for the
/// whole child run, so a wide fan-out queues (or fails, per behavior)
/// instead of exhausting file descriptors.
pub spawn_permits: std::sync::Arc<tokio::sync::Semaphore>,
#[expect(
dead_code,
reason = "unused in production; remove expect when wired or delete the item"
@@ -397,6 +404,21 @@ impl SubagentSpawnContext {
.resolve()
.value
}
/// Subagent summarizer `tool_choice`, mirroring
/// `Config::resolve_compaction_tool_choice` (env > config > remote
/// settings > default `auto`).
pub fn resolve_compaction_tool_choice(&self) -> crate::util::config::CompactionToolChoice {
crate::util::config::resolve_compaction_tool_choice_from(
crate::agent::config::env_string(crate::util::config::ENV_COMPACTION_TOOL_CHOICE)
.as_deref(),
self.agent_config
.as_ref()
.and_then(|c| c.features.compaction_tool_choice.as_deref()),
self.remote_settings
.as_ref()
.and_then(|r| r.compaction_tool_choice.as_deref()),
)
}
/// Whether a completed subagent's worktree is snapshotted into a durable ref
/// and its directory deleted. Resolution mirrors the other subagent gates
/// (env > config > remote settings > default). Default `false` so it ships dark;
crates/codegen/kigi-shell/src/agent/subagent/tests/mod.rs
+24
View File
@@ -3298,3 +3298,27 @@ fn spawn_test_parent_chat_state(model_slug: &str) -> kigi_chat_state::ChatStateH
)
}
mod rest;
/// A blocked child actor must not park teardown: the read degrades to the fallback in bounded time.
#[tokio::test(start_paused = true)]
async fn child_actor_query_is_bounded_when_the_actor_never_answers() {
use super::handle_request::{CHILD_ACTOR_ACK_TIMEOUT, child_actor_query};
let tokens = tokio::time::timeout(
20 * CHILD_ACTOR_ACK_TIMEOUT,
child_actor_query("test_query", std::future::pending::<u64>(), 7),
)
.await
.expect("child-actor queries must complete in bounded time");
assert_eq!(tokens, 7, "a starved query must degrade to the fallback");
}
/// A parent that never services `cmd_rx` must not park a completed child.
#[tokio::test(start_paused = true)]
async fn parent_ack_is_bounded_when_the_parent_never_answers() {
use super::handle_request::{PARENT_ACK_TIMEOUT, parent_ack};
let (_never_answered, ack) = tokio::sync::oneshot::channel::<()>();
let acked = tokio::time::timeout(20 * PARENT_ACK_TIMEOUT, parent_ack("test_ack", ack))
.await
.expect("parent acks must complete in bounded time");
assert!(!acked, "an unanswered ack must report a miss");
}
crates/codegen/kigi-shell/src/auth/credential_provider.rs
+61
View File
@@ -83,6 +83,35 @@ impl AuthCredentialProvider for ShellAuthCredentialProvider {
self.auth_manager.try_recover_unauthorized().await
}
}
/// The memory-embedding credentials for `embed_base_url`, decided by the ONE
/// authority rather than re-derived here (C1).
///
/// `embed_base_url` is the session model's own endpoint, so on a BYOK or
/// subscription-OAuth model it points at that provider's host — and
/// [`kigi_auth::AuthRetryMiddleware`] stamps `Authorization` on every request
/// it wraps. Asking [`CredentialAuthority::manager_for`] answers both halves at
/// once: whether a session credential may ride there at all, and WHICH manager
/// governs it (a subscription-OAuth platform's pooled manager at its own host,
/// the primary at the session's coding endpoint). No manager means no session
/// credential at all; the platform's own `embed_api_key` is untouched and keeps
/// serving its own endpoint.
///
/// The session's `SharedApiKeyProvider` is deliberately NOT forwarded: it is
/// hard-wired to the PRIMARY manager, so at a pooled platform's host — where a
/// credential may ride, but only that platform's own — it would resolve the
/// wrong bearer.
pub(crate) fn embedding_session_credentials(
embed_base_url: &str,
platform: Option<kigi_models::PlatformId>,
authority: &crate::auth::credential_authority::CredentialAuthority,
) -> kigi_memory::EndpointScopedCredentials {
let auth_credentials = authority.manager_for(platform, embed_base_url).map(|am| {
Arc::new(ShellAuthCredentialProvider::new(am, None, None))
as Arc<dyn AuthCredentialProvider>
});
let may_ride = auth_credentials.is_some();
kigi_memory::EndpointScopedCredentials::for_endpoint(embed_base_url, may_ride, auth_credentials)
}
#[cfg(test)]
mod tests {
use super::*;
@@ -253,6 +282,38 @@ mod tests {
"snapshot must reflect refreshed token for subsequent apply() calls"
);
}
/// C1: memory embeddings follow the credential authority, not the session.
///
/// A session whose model is a BYOK platform aims `embed_base_url` at that
/// provider's host; the authority answers "no manager governs a credential
/// there", so nothing rides. The session's own coding endpoint still does.
#[test]
fn embedding_credentials_follow_the_credential_authority() {
let _guard = EarlyInvalidationGuard::pin_to_default();
let dir = tempfile::tempdir().unwrap();
let mgr = make_manager(
&dir,
Some(make_auth("session-bearer", ChronoDuration::hours(1))),
);
let endpoints = crate::agent::config::EndpointsConfig::default();
let coding_endpoint = endpoints.proxy_url();
let authority =
crate::auth::credential_authority::CredentialAuthority::new(endpoints, Some(mgr));
assert!(
!embedding_session_credentials(&coding_endpoint, None, &authority).is_empty(),
"the session's own coding endpoint keeps its credential"
);
assert!(
embedding_session_credentials(
"https://api.anthropic.com/v1",
kigi_models::PlatformId::parse("anthropic"),
&authority,
)
.is_empty(),
"an API-key platform's host must receive no session credential"
);
}
/// Deployment-key path has no recovery (operator owns the bearer).
#[tokio::test]
async fn refresh_after_unauthorized_is_noop_for_deployment_key() {
crates/codegen/kigi-shell/src/auth/storage.rs
+154 -12
View File
@@ -224,6 +224,18 @@ impl AuthFileLock {
pub fn read_auth_json(auth_file: &Path) -> std::io::Result<AuthStore> {
let mut file = File::open(auth_file)?;
// Tighten world-readable copies (hand-restored, umask edge cases, etc.)
// right after open, so even an unreadable/invalid file gets fixed.
// Best-effort: a chmod failure must not block login/read paths.
if let Err(e) = crate::util::secure_file::ensure_owner_only_permissions(auth_file) {
tracing::warn!(
path = %auth_file.display(),
error = %e,
"auth: failed to enforce owner-only permissions on auth.json"
);
}
let mut contents = String::new();
file.read_to_string(&mut contents)?;
@@ -289,6 +301,14 @@ pub(crate) fn backup_corrupt_auth_file(path: &Path) -> Option<PathBuf> {
match std::fs::rename(path, &backup) {
Ok(()) => {
// Corrupt backups still hold token material — keep them owner-only.
if let Err(e) = crate::util::secure_file::ensure_owner_only_permissions(&backup) {
tracing::warn!(
path = %backup.display(),
error = %e,
"auth: failed to tighten permissions on corrupt-auth backup"
);
}
tracing::warn!(
original = %path.display(),
backup = %backup.display(),
@@ -421,18 +441,59 @@ fn write_store_to(path: &Path, auth_store: &AuthStore) -> std::io::Result<()> {
.into_inner()
.map_err(|e| e.into_error())?
.sync_all()?;
#[cfg(windows)]
{
crate::util::secure_file::set_windows_secure_permissions(path)?;
// `open_secure_file` mode bits apply only on create; tighten existing
// paths. Best-effort after durable content: a chmod-only failure must not
// look like a failed write — the in-place fallback restores the prior
// snapshot on any `write_store_to` Err, which would discard freshly
// written tokens. The load path re-tightens on the next read.
if let Err(e) = crate::util::secure_file::ensure_owner_only_permissions(path) {
tracing::warn!(
error = %e,
path = %path.display(),
"auth: failed to ensure owner-only permissions after write"
);
}
Ok(())
}
/// Atomic write: tmp + Windows-safe replace (see `util::fs::replace_file`).
fn write_auth_json_atomic(auth_file: &Path, auth_store: &AuthStore) -> std::io::Result<()> {
let tmp = auth_file.with_extension(format!("json.{}.tmp", std::process::id()));
write_store_to(&tmp, auth_store)?;
crate::util::fs::replace_file(&tmp, auth_file)
write_auth_json_atomic_with(auth_file, auth_store, write_store_to)
}
/// Inner of [`write_auth_json_atomic`] with the tmp-file writer injectable so
/// the torn-tmp-write path is unit-testable without real I/O faults.
fn write_auth_json_atomic_with(
auth_file: &Path,
auth_store: &AuthStore,
write: fn(&Path, &AuthStore) -> std::io::Result<()>,
) -> std::io::Result<()> {
// Unique per write (pid + monotonic seq): two concurrent in-process
// writers (e.g. background mint + proactive refresher) must not share
// one tmp path, or each can tear the other's half-written bytes.
static TMP_SEQ: std::sync::atomic::AtomicU64 = std::sync::atomic::AtomicU64::new(0);
let tmp = auth_file.with_extension(format!(
"json.{}.{}.tmp",
std::process::id(),
TMP_SEQ.fetch_add(1, std::sync::atomic::Ordering::Relaxed)
));
if let Err(e) = write(&tmp, auth_store) {
// Reclaim the partial tmp; the unique name otherwise accumulates
// one orphan per failed write. `replace_file` cleans its own leg.
let _ = std::fs::remove_file(&tmp);
return Err(e);
}
crate::util::fs::replace_file(&tmp, auth_file)?;
// Re-assert on the final path (covers rename edge cases / FS quirks).
// Best-effort: the replace already published the new tokens.
if let Err(e) = crate::util::secure_file::ensure_owner_only_permissions(auth_file) {
tracing::warn!(
error = %e,
path = %auth_file.display(),
"auth: failed to ensure owner-only permissions after replace"
);
}
Ok(())
}
/// Non-atomic fallback: truncate and rewrite `auth.json` in place.
@@ -486,10 +547,7 @@ fn restore_prior_bytes(auth_file: &Path, bytes: &[u8]) -> std::io::Result<()> {
let mut file = open_secure_file(auth_file)?;
file.write_all(bytes)?;
file.sync_all()?;
#[cfg(windows)]
{
crate::util::secure_file::set_windows_secure_permissions(auth_file)?;
}
crate::util::secure_file::ensure_owner_only_permissions(auth_file)?;
Ok(())
}
@@ -804,7 +862,87 @@ mod write_fallback_tests {
);
}
/// Rollback after a failed write must keep the file owner-only (0o600).
/// An in-place rewrite over a pre-existing world-readable auth.json must
/// tighten it to 0o600 — `open_secure_file` mode bits apply only on
/// create, so truncating an existing loose file keeps its old mode. The
/// in-place path is the one that reuses the inode (the atomic path
/// replaces it with a fresh 0o600 temp file).
#[cfg(unix)]
#[test]
fn in_place_write_tightens_preexisting_world_readable_auth_json() {
use std::os::unix::fs::PermissionsExt;
let dir = tempfile::tempdir().unwrap();
let path = dir.path().join("auth.json");
std::fs::write(&path, b"{}").unwrap();
let mut loose = std::fs::metadata(&path).unwrap().permissions();
loose.set_mode(0o644);
std::fs::set_permissions(&path, loose).unwrap();
write_auth_json_in_place(&path, &sample_store()).unwrap();
let mode = std::fs::metadata(&path).unwrap().permissions().mode();
assert_eq!(
mode & 0o777,
0o600,
"in-place rewrite must tighten preexisting open perms"
);
}
/// Simulates a torn tmp write: partial bytes land at the target path,
/// then the write fails — the temp file must not be left behind.
fn fake_partial_write_then_fail(path: &Path, _: &AuthStore) -> std::io::Result<()> {
std::fs::write(path, b"{ torn")?;
Err(std::io::Error::from(std::io::ErrorKind::StorageFull))
}
/// A tmp write that fails partway must reclaim the temp file; the unique
/// tmp name otherwise accumulates one orphan per failed write.
#[test]
fn atomic_write_reclaims_tmp_when_store_write_fails() {
let dir = tempfile::tempdir().unwrap();
let path = dir.path().join("auth.json");
let err = write_auth_json_atomic_with(&path, &sample_store(), fake_partial_write_then_fail)
.unwrap_err();
assert_eq!(err.kind(), std::io::ErrorKind::StorageFull);
let orphans: Vec<_> = std::fs::read_dir(dir.path())
.unwrap()
.filter_map(Result::ok)
.filter(|e| e.file_name().to_string_lossy().contains(".tmp"))
.collect();
assert!(
orphans.is_empty(),
"failed tmp write must not leave temp files behind: {orphans:?}"
);
}
/// The replace step failing (target path occupied by a directory) must
/// also leave no temp file — `replace_file` reclaims its tmp on failure.
#[test]
fn atomic_write_reclaims_tmp_when_replace_fails() {
let dir = tempfile::tempdir().unwrap();
let path = dir.path().join("auth.json");
std::fs::create_dir(&path).unwrap();
assert!(
write_auth_json_atomic(&path, &sample_store()).is_err(),
"replacing a directory must fail"
);
let orphans: Vec<_> = std::fs::read_dir(dir.path())
.unwrap()
.filter_map(Result::ok)
.filter(|e| e.file_name().to_string_lossy().contains(".tmp"))
.collect();
assert!(
orphans.is_empty(),
"failed replace must not leave temp files behind: {orphans:?}"
);
}
/// Rollback after a failed write must leave the file owner-only (0o600)
/// even when the file had loose permissions going in — the restore path
/// truncates the existing inode, which keeps its old mode unless
/// explicitly tightened.
#[cfg(unix)]
#[test]
fn in_place_restore_is_owner_only() {
@@ -812,9 +950,13 @@ mod write_fallback_tests {
let dir = tempfile::tempdir().unwrap();
let path = dir.path().join("auth.json");
write_auth_json_in_place(&path, &sample_store()).unwrap();
let mut loose = std::fs::metadata(&path).unwrap().permissions();
loose.set_mode(0o644);
std::fs::set_permissions(&path, loose).unwrap();
let _ = write_auth_json_in_place_with(&path, &sample_store(), fake_truncate_then_fail);
let mode = std::fs::metadata(&path).unwrap().permissions().mode();
assert_eq!(mode & 0o777, 0o600, "restored file must stay 0o600");
assert_eq!(mode & 0o777, 0o600, "restored file must be 0o600");
}
}
crates/codegen/kigi-shell/src/config/mod.rs
+18 -7
View File
@@ -1034,16 +1034,27 @@ pub fn apply_sandbox(
.and_then(|p| dunce::canonicalize(p).ok())
.or_else(|| std::env::current_dir().ok())
.unwrap_or_else(|| std::path::PathBuf::from("."));
if kigi_sandbox::restricts_home_writes(&sandbox_profile, &workspace) {
// The profile leaves ~/.kigi read-only outside its state dirs: mint
// what first use would otherwise write after enforcement.
if let Err(e) = crate::auth::device::device_id() {
tracing::warn!(error = %e, "sandbox: device id not created before enforcement");
}
kigi_tools::implementations::kigi::grep::ripgrep::rg_path();
crate::builtin::extract_bundled_files(&crate::util::kigi_home::kigi_home());
}
// The devbox `/data` write-deny is a bwrap mount; Landlock would grant the workspace under `/data`.
#[cfg(target_os = "linux")]
let requires_read_deny = kigi_sandbox::requires_read_deny(&sandbox_profile, &workspace);
let bwrap_required = kigi_sandbox::requires_read_deny(&sandbox_profile, &workspace)
|| kigi_sandbox::requires_data_write_deny(&sandbox_profile, &workspace);
#[cfg(target_os = "linux")]
{
let refuse_unprotected = |detail: &str| {
eprintln!(
"error: this sandbox could not enforce its read-deny set on Linux \
(bubblewrap missing/unusable, or a deny glob exceeded its expansion \
limit — see any message above). Install bubblewrap with \
`apt install -y bubblewrap` if needed. Refusing to start with denied \
"error: this sandbox could not enforce its read-deny set or the devbox \
`/data` write-deny on Linux (bubblewrap missing/unusable, or a deny glob \
exceeded its expansion limit — see any message above). Install bubblewrap \
with `apt install -y bubblewrap` if needed. Refusing to start with denied \
paths unprotected.{detail}"
);
};
@@ -1051,7 +1062,7 @@ pub fn apply_sandbox(
Some(mut cmd) => {
use std::os::unix::process::CommandExt;
let err = cmd.exec();
if requires_read_deny {
if bwrap_required {
refuse_unprotected(&format!(" (bwrap exec failed: {err})"));
std::process::exit(1);
}
@@ -1061,7 +1072,7 @@ pub fn apply_sandbox(
Install bubblewrap: apt install -y bubblewrap"
);
}
None if requires_read_deny && !kigi_sandbox::is_inside_bwrap() => {
None if bwrap_required && !kigi_sandbox::is_inside_bwrap() => {
refuse_unprotected("");
std::process::exit(1);
}
crates/codegen/kigi-shell/src/extensions/memory.rs
+2 -2
View File
@@ -37,9 +37,9 @@ async fn handle_compact(agent: &MvpAgent, args: &acp::ExtRequest) -> ExtResult {
respond_to: tx,
});
}
// The session's own ACP error goes through unchanged: a Debug wrapper would escape its detail.
rx.await
.map_err(|_| acp::Error::internal_error().data("session failed to respond"))?
.map_err(|e| acp::Error::internal_error().data(format!("Internal error: {:?}", e)))?;
.map_err(|_| acp::Error::internal_error().data("session failed to respond"))??;
to_raw_response(&CompactConversationResponse {})
}
crates/codegen/kigi-shell/src/extensions/notification.rs
+11
View File
@@ -691,6 +691,11 @@ pub enum SessionUpdate {
/// does not support reasoning effort or no effort override was applied.
#[serde(default, skip_serializing_if = "Option::is_none")]
reasoning_effort: Option<String>,
/// The session's effective context window after the switch, so a
/// follower repaints its header without waiting for the next turn.
/// `None` from an agent that predates the field.
#[serde(default, skip_serializing_if = "Option::is_none")]
context_window: Option<u64>,
},
/// Streaming chunk of a tool call's arguments.
///
@@ -1970,15 +1975,18 @@ mod tests {
let with_effort = SessionUpdate::ModelChanged {
model_id: "kigi-4".into(),
reasoning_effort: Some("high".into()),
context_window: Some(262_144),
};
let json = serde_json::to_value(&with_effort).unwrap();
assert_eq!(json["sessionUpdate"], "model_changed");
assert_eq!(json["model_id"], "kigi-4");
assert_eq!(json["reasoning_effort"], "high");
assert_eq!(json["context_window"], 262_144);
let without_effort = SessionUpdate::ModelChanged {
model_id: "kigi-3".into(),
reasoning_effort: None,
context_window: None,
};
let json = serde_json::to_value(&without_effort).unwrap();
assert_eq!(json["sessionUpdate"], "model_changed");
@@ -1988,6 +1996,7 @@ mod tests {
"reasoning_effort: None must be skipped on the wire so old pagers \
and third-party ACP clients see a smaller, no-extra-keys payload"
);
assert!(json.get("context_window").is_none());
}
/// `ModelChanged` round-trips through JSON: a follower client deserializes
@@ -2000,6 +2009,7 @@ mod tests {
let original = SessionUpdate::ModelChanged {
model_id: "kigi-4".into(),
reasoning_effort: Some("medium".into()),
context_window: Some(1_000_000),
};
let json_str = serde_json::to_string(&original).unwrap();
let parsed: SessionUpdate = serde_json::from_str(&json_str).unwrap();
@@ -2021,6 +2031,7 @@ mod tests {
update: SessionUpdate::ModelChanged {
model_id: "kigi-4".into(),
reasoning_effort: None,
context_window: None,
},
meta: None,
};
crates/codegen/kigi-shell/src/leader/lock.rs
+13 -4
View File
@@ -52,10 +52,16 @@ pub fn default_lock_path_in(root: &Path) -> PathBuf {
root.join("leader.lock")
}
/// Runtime files live under `~/.kigi/run`: the one home subtree a strict
/// sandbox leaves writable for them.
pub(super) fn runtime_root() -> PathBuf {
kigi_home().join("run")
}
/// Effective leader lock path: the [`LEADER_SOCKET_ENV`] override's sibling
/// `.lock` when set, else the default under kigi home.
/// `.lock` when set, else the default under `~/.kigi/run`.
pub fn default_lock_path() -> PathBuf {
resolve_lock_path(leader_socket_override(), &kigi_home())
resolve_lock_path(leader_socket_override(), &runtime_root())
}
pub fn default_socket_path_in(root: &Path) -> PathBuf {
@@ -63,9 +69,9 @@ pub fn default_socket_path_in(root: &Path) -> PathBuf {
}
/// Effective leader socket path: the [`LEADER_SOCKET_ENV`] override when set,
/// else the default under kigi home.
/// else the default under `~/.kigi/run`.
pub fn default_socket_path() -> PathBuf {
resolve_socket_path(leader_socket_override(), &kigi_home())
resolve_socket_path(leader_socket_override(), &runtime_root())
}
/// The instance suffix encoded in a lock/socket file-name pair
@@ -145,6 +151,9 @@ impl LeaderLock {
}
fn open_lock_file(&self) -> Result<File, LockError> {
if let Some(parent) = self.lock_path.parent() {
std::fs::create_dir_all(parent)?;
}
Ok(OpenOptions::new()
.read(true)
.write(true)
crates/codegen/kigi-shell/src/leader/mod.rs
+68 -5
View File
@@ -23,7 +23,7 @@
//! │ │ - Tracks session ownership for routing ││
//! │ └────────────────────────┬────────────────────────────────┘│
//! └───────────────────────────┼──────────────────────────────────┘
//! │ IPC (Unix socket at ~/.kigi/leader.sock)
//! │ IPC (Unix socket at ~/.kigi/run/leader.sock)
//! ┌───────────────────┼───────────────────┐
//! ▼ ▼ ▼
//! ┌───────────────┐ ┌───────────────┐ ┌───────────────┐
@@ -481,7 +481,10 @@ async fn discover_leaders_in(root: &Path) -> Vec<LeaderDescriptor> {
entries
}
pub async fn discover_leaders() -> Vec<LeaderDescriptor> {
discover_leaders_in(&crate::util::kigi_home::kigi_home()).await
let mut entries = discover_leaders_in(&lock::runtime_root()).await;
// Pre-0.1.15 leaders and socket overrides keep their files at the home root.
entries.extend(discover_leaders_in(&crate::util::kigi_home::kigi_home()).await);
entries
}
/// (pid, leader_binary_version) of socket-verified (Reachable) leaders; a
/// stale-lock-only descriptor is skipped (its `pid_from_lock` may be recycled).
@@ -610,13 +613,28 @@ fn resolve_matching_descriptors(
fn resolve_target_from_descriptors(
target: LeaderTarget,
leaders: Vec<LeaderDescriptor>,
current_default_socket: &Path,
) -> Result<LeaderTargetSelection, LeaderTargetError> {
match target {
LeaderTarget::Default => {
let matching: Vec<_> = leaders
let defaults: Vec<_> = leaders
.into_iter()
.filter(|descriptor| descriptor.is_default)
.collect();
// An upgrade can leave a pre-0.1.15 default at the home root; the
// current default socket wins while it exists.
let current: Vec<_> = defaults
.iter()
.filter(|descriptor| {
descriptor.socket_path.as_deref() == Some(current_default_socket)
})
.cloned()
.collect();
let matching = if current.is_empty() {
defaults
} else {
current
};
resolve_matching_descriptors("default", matching)
}
LeaderTarget::Socket(socket_path) => {
@@ -685,7 +703,7 @@ pub async fn resolve_leader_target(
target: LeaderTarget,
) -> Result<LeaderTargetSelection, LeaderTargetError> {
let leaders = discover_leaders().await;
resolve_target_from_descriptors(target, leaders)
resolve_target_from_descriptors(target, leaders, &lock::default_socket_path())
}
#[derive(Debug, thiserror::Error)]
pub enum ConnectionError {
@@ -1315,7 +1333,12 @@ fn spawn_leader_subprocess() -> Result<u32, ConnectionError> {
}
cmd.stdin(std::process::Stdio::null())
.stdout(std::process::Stdio::null());
let log_path = crate::util::kigi_home::kigi_home().join("leader.log");
let log_path = crate::util::kigi_home::kigi_home()
.join("logs")
.join("leader.log");
if let Some(parent) = log_path.parent() {
let _ = std::fs::create_dir_all(parent);
}
match std::fs::File::create(&log_path) {
Ok(log_file) => {
info!("Leader stderr → log file");
@@ -1444,6 +1467,46 @@ mod tests {
vec![(222, "0.2.52".to_string())]
);
}
fn reachable_default(socket: &str) -> LeaderDescriptor {
let socket_path = PathBuf::from(socket);
LeaderDescriptor {
pid_from_lock: Some(1),
lock_path: Some(socket_path.with_extension("lock")),
socket_path: Some(socket_path.clone()),
socket_suffix: String::new(),
classification: LeaderDiscoveryState::Reachable,
is_default: true,
live_info: Some(LiveLeaderInfo {
pid: 1,
socket_path,
lock_path: PathBuf::new(),
socket_suffix: String::new(),
leader_protocol_version: 1,
leader_binary_version: "0.1.15".to_string(),
}),
target_error: None,
}
}
#[test]
fn default_target_prefers_the_current_socket_over_a_legacy_default() {
let legacy = reachable_default("/h/.kigi/leader.sock");
let current = reachable_default("/h/.kigi/run/leader.sock");
let current_socket = PathBuf::from("/h/.kigi/run/leader.sock");
let picked = resolve_target_from_descriptors(
LeaderTarget::Default,
vec![current.clone(), legacy.clone()],
&current_socket,
)
.expect("current default resolves");
assert_eq!(picked.descriptor.socket_path, current.socket_path);
let fallback = resolve_target_from_descriptors(
LeaderTarget::Default,
vec![legacy.clone()],
&current_socket,
)
.expect("legacy default still resolves alone");
assert_eq!(fallback.descriptor.socket_path, legacy.socket_path);
}
#[test]
fn leader_is_older_than_directional() {
assert!(leader_is_older_than("0.1.0", "0.2.0"));
crates/codegen/kigi-shell/src/leader/server.rs
+3
View File
@@ -1146,6 +1146,9 @@ pub async fn run_leader_server(
control_state: LeaderServerControlState,
) -> Result<(), ServerError> {
let _ = std::fs::remove_file(&socket_path);
if let Some(parent) = socket_path.parent() {
std::fs::create_dir_all(parent)?;
}
let shutdown_reason_rx = shutdown_tx.subscribe();
let listener = LeaderListener::bind(&socket_path)?;
info!("Leader server listening");
crates/codegen/kigi-shell/src/sampling/error.rs
+135
View File
@@ -171,6 +171,70 @@ pub fn error_detail_from_data(data: &serde_json::Value) -> Option<String> {
.map(str::to_owned)
}
/// The human-readable text of an ACP error: its `data` detail when present (Display would JSON-escape it), else the message.
pub fn acp_error_message(err: &acp::Error) -> String {
err.data
.as_ref()
.and_then(error_detail_from_data)
.unwrap_or_else(|| err.message.clone())
}
/// First line of a compaction error without internal `compact failed:` prefixes, capped for a banner; diagnostic lines (request URLs) are dropped.
pub fn user_facing_compact_error(raw: &str) -> String {
const INTERNAL_PREFIX: &str = "compact failed:";
const MAX_CHARS: usize = 200;
// Redact before line selection: a line break inside a URL must not hide its `@`.
let raw = redact_urls_in_text(raw);
let mut rest = raw.trim();
while rest
.get(..INTERNAL_PREFIX.len())
.is_some_and(|head| head.eq_ignore_ascii_case(INTERNAL_PREFIX))
{
rest = rest[INTERNAL_PREFIX.len()..].trim_start();
}
rest.lines()
.next()
.unwrap_or_default()
.trim()
.chars()
.take(MAX_CHARS)
.collect()
}
/// Every canonical `scheme://` URL in `text` loses userinfo, sensitive query values and fragment; one that fails to parse is not echoed.
fn redact_urls_in_text(text: &str) -> String {
let mut out = String::with_capacity(text.len());
let mut rest = text;
while let Some(sep) = rest.find("://") {
// The delimiter before the scheme may be multi-byte (`“`, `：`): step past its UTF-8 length.
let scheme_start = rest[..sep]
.char_indices()
.rev()
.find(|(_, c)| !(c.is_ascii_alphanumeric() || matches!(c, '+' | '-' | '.')))
.map_or(0, |(at, c)| at + c.len_utf8());
out.push_str(&rest[..scheme_start]);
// The authority may hold whitespace (a configured password); the token ends at whitespace after it.
let authority = &rest[sep + 3..];
let authority_end = authority.find(['/', '?', '#']).unwrap_or(authority.len());
let token_end = authority[authority_end..]
.find(char::is_whitespace)
.map_or(rest.len(), |at| sep + 3 + authority_end + at);
let token = &rest[scheme_start..token_end];
let url_text = token.trim_end_matches([')', ']', '}', ',', '.', ';', '\'', '"', '>']);
match url::Url::parse(url_text) {
Ok(mut url) => {
kigi_secrets::redact_url(&mut url);
out.push_str(url.as_str());
}
Err(_) => out.push_str("<unparseable url>"),
}
out.push_str(&token[url_text.len()..]);
rest = &rest[token_end..];
}
out.push_str(rest);
out
}
pub fn http_status_from_error(err: &acp::Error) -> Option<u16> {
err.data
.as_ref()?
@@ -254,6 +318,77 @@ pub fn prompt_complete_fields(
#[cfg(test)]
mod tests {
#[test]
fn compact_error_text_keeps_the_first_decoded_line_only() {
let err = super::acp::Error::internal_error().data(
"compact failed: API error (status 500): overloaded\n\nRequest URL: http://u:TEST_PASSWORD@127.0.0.1/v1",
);
let detail = super::user_facing_compact_error(&super::acp_error_message(&err));
assert_eq!(detail, "API error (status 500): overloaded");
assert!(!detail.contains("TEST_PASSWORD"));
assert_eq!(
super::user_facing_compact_error(&"x".repeat(500))
.chars()
.count(),
200
);
assert_eq!(
super::acp_error_message(&super::acp::Error::internal_error()),
"Internal error"
);
}
/// A sampler 401 names the endpoint on the retained line; its userinfo must not survive.
#[test]
fn compact_error_text_redacts_url_credentials_on_the_first_line() {
for password in [
"TEST_PASSWORD",
"TEST PASSWORD",
"TEST%20PASSWORD",
"TEST_PASSWORD\n",
"TEST_PASSWORD\r\n",
] {
let raw = format!(
"compact failed: Unauthorized (401) from http://u:{password}@127.0.0.1:33263/401/v1/chat/completions: bad auth"
);
assert_eq!(
super::user_facing_compact_error(&raw),
"Unauthorized (401) from http://127.0.0.1:33263/401/v1/chat/completions: bad auth",
"{password:?}"
);
}
assert_eq!(
super::redact_urls_in_text("see https://a:b@h.example/x and http://h2/"),
"see https://h.example/x and http://h2/"
);
}
/// A credential carried as a query value survives no failure shape: 401 message or transport error.
#[test]
fn compact_error_text_redacts_query_credentials_in_every_failure_shape() {
let unauthorized = "compact failed: Unauthorized (401) from http://127.0.0.1:1/?access_token=TEST_QUERY_TOKEN&route=/v1/chat/completions: bad auth";
let transport = "compact failed: request error: error sending request for url (http://127.0.0.1:1/?access_token=TEST_QUERY_TOKEN&route=/500/v1/chat/completions)";
for raw in [unauthorized, transport] {
let detail = super::user_facing_compact_error(raw);
assert!(!detail.contains("TEST_QUERY_TOKEN"), "{detail}");
assert!(detail.contains("route="), "{detail}");
}
assert!(super::user_facing_compact_error("see http://:://").contains("<unparseable url>"));
}
/// A multi-byte delimiter right before the URL must not break the scan.
#[test]
fn compact_error_text_handles_unicode_before_a_url() {
for raw in [
"compact failed: Service unavailable. See “https://u:TEST_PASSWORD@status.example/”.",
"compact failed: 请参阅：https://u:TEST_PASSWORD@status.example/",
] {
let detail = super::user_facing_compact_error(raw);
assert!(!detail.contains("TEST_PASSWORD"), "{detail}");
assert!(detail.contains("https://status.example/"), "{detail}");
}
}
use super::*;
use reqwest::StatusCode;
crates/codegen/kigi-shell/src/session/acp_session.rs
+13
View File
@@ -121,6 +121,10 @@ mod model_switch;
mod prompt_queue;
#[path = "acp_session_impl/slash_exec.rs"]
mod slash_exec;
#[path = "acp_session_impl/stationarity_seam.rs"]
mod stationarity_seam;
#[path = "acp_session_impl/swarm.rs"]
mod swarm;
use super::PromptOrigin;
use super::acp_types;
use super::chat_persistence;
@@ -616,6 +620,9 @@ pub(crate) struct SessionActor {
/// layered over the goal engine. Modeled after `goal_tracker` above;
/// all graph state logic lives in `graph_tracker.rs`.
pub(crate) graph_tracker: Arc<parking_lot::Mutex<crate::session::graph_tracker::GraphTracker>>,
/// Swarm mode: a standing instruction to fan work out, independent of the
/// goal/graph engines — it steers tool choice, it does not drive turns.
pub(crate) swarm_mode: std::cell::Cell<crate::session::swarm_mode::SwarmMode>,
/// Max graph nodes running concurrently (1 = serial G0 behavior).
/// Cached at actor construction from `resolve_graph_concurrency`.
pub(crate) graph_concurrency: u32,
@@ -1033,6 +1040,9 @@ impl SessionActor {
// Graph rides the goal harness: nodes execute as goals, so
// `/graph` is only real when `/goal` is.
graph: self.graph_enabled && goal,
swarm: tool_names
.iter()
.any(|n| n == kigi_tools::implementations::kigi::AGENT_SWARM_TOOL_NAME),
}
}
/// Names of every tool registered with the session's tool bridge.
@@ -1508,6 +1518,9 @@ mod load_user_prompts_tests;
#[path = "acp_session_tests/memory_config_tests.rs"]
mod memory_config_tests;
#[cfg(test)]
#[path = "acp_session_tests/model_switch_window_tests.rs"]
mod model_switch_window_tests;
#[cfg(test)]
#[path = "acp_session_tests/parallel_dispatch_tests.rs"]
mod parallel_dispatch_tests;
#[cfg(test)]
crates/codegen/kigi-shell/src/session/acp_session_impl/hook_dispatch.rs
+6 -1
View File
@@ -8,7 +8,12 @@ pub(super) fn turn_result_to_hook_outcome(
) -> kigi_tool_protocol::turn_hook::TurnHookOutcome {
use kigi_tool_protocol::turn_hook::TurnHookOutcome;
match result {
Ok(TurnOutcome::Completed { .. }) => TurnHookOutcome::Completed,
// A stationarity halt is a completed turn for hook purposes: nobody
// cancelled it, and a `Cancelled` outcome would tell every Stop hook
// the user interrupted the model.
Ok(TurnOutcome::Completed { .. }) | Ok(TurnOutcome::StationarityHalted { .. }) => {
TurnHookOutcome::Completed
}
Ok(TurnOutcome::Cancelled { .. }) | Ok(TurnOutcome::MaxTurnsReached { .. }) => {
TurnHookOutcome::Cancelled
}
crates/codegen/kigi-shell/src/session/acp_session_impl/model_switch.rs
+2 -2
View File
@@ -10,7 +10,7 @@ impl SessionActor {
apply_prompt_override: bool,
skip_prompt_rewrite: bool,
auto_compact_threshold_percent: u8,
) -> Result<acp::ModelId, acp::Error> {
) -> Result<(acp::ModelId, std::num::NonZeroU64), acp::Error> {
let model_id = acp::ModelId::new(sampling_config.model.clone());
// H4: record the picker's catalog KEY as this SESSION's own selection.
// `sampling_config.model` is the ambiguous bare slug; the key is what
@@ -120,7 +120,7 @@ impl SessionActor {
agent_name: Some(agent_name),
reasoning_effort: Some(sampling_config.reasoning_effort),
});
Ok(model_id)
Ok((model_id, new_context_window))
}
/// Handle [`SessionCommand::RebuildAgentForDefinition`].
///
crates/codegen/kigi-shell/src/session/acp_session_impl/rewind.rs
+4 -11
View File
@@ -548,17 +548,10 @@ impl SessionActor {
if report.changed() && !dry_run {
// Flush barrier: success must mean the rewrite is on disk.
let (flush_tx, flush_rx) = tokio::sync::oneshot::channel();
if self
.notifications
.persistence_tx
.send(PersistenceMsg::FlushAndAck {
respond_to: flush_tx,
})
.is_err()
|| flush_rx.await.is_err()
{
anyhow::bail!("history repaired in memory but the persistence flush failed");
if let Err(error) = self.persistence_barrier().await {
anyhow::bail!(
"history repaired in memory but the persistence flush failed: {error}"
);
}
tracing::warn!(
session_id = %self.session_info.id.0,
crates/codegen/kigi-shell/src/session/acp_session_impl/slash_exec.rs
+30
View File
@@ -709,6 +709,36 @@ impl SessionActor {
BuiltinAction::GoalSet { .. } => {
unreachable!("GoalSet is intercepted in handle_prompt")
}
BuiltinAction::SwarmSet { enabled } => {
let msg = self.apply_swarm_mode(
enabled.then_some(crate::session::swarm_mode::SwarmTrigger::Manual),
);
self.send_slash_command_output(&msg).await;
ok_end_turn(0, None)
}
BuiltinAction::SwarmToggle => {
let turning_on = !self.swarm_mode.get().is_active();
let msg = self.apply_swarm_mode(
turning_on.then_some(crate::session::swarm_mode::SwarmTrigger::Manual),
);
self.send_slash_command_output(&msg).await;
ok_end_turn(0, None)
}
// `/swarm <task>` is handled before dispatch (it has to seed the
// turn with the task text); reaching here means the interception
// was bypassed, so report rather than silently dropping the task.
BuiltinAction::SwarmTask { prompt } => {
tracing::warn!(
prompt_len = prompt.len(),
"/swarm <task> reached the builtin executor; the turn seam did not intercept it"
);
self.send_slash_command_output(
"Could not start a swarm turn for that task. Run `/swarm on` and send the \
task as a normal message.",
)
.await;
ok_end_turn(0, None)
}
BuiltinAction::GoalStatus => {
let current_tokens = self.chat_state_handle.get_total_tokens().await as i64;
let goal_tokens = self.goal_tokens_used(current_tokens);
crates/codegen/kigi-shell/src/session/acp_session_impl/spawn.rs
+57 -23
View File
@@ -112,6 +112,7 @@ pub(crate) async fn spawn_session_actor(
system_prompt_label: String,
compaction_mode: kigi_chat_state::CompactionMode,
compaction_verbatim_input: bool,
compaction_tool_choice: crate::util::config::CompactionToolChoice,
two_pass_enabled: bool,
buffering_settings: Option<BufferingSettings>,
origin_client: Option<crate::http::OriginClientInfo>,
@@ -311,6 +312,18 @@ pub(crate) async fn spawn_session_actor(
};
let embed_base_url = sampling_config.base_url.clone();
let embed_api_key = sampling_config.api_key.clone();
// The platform behind the endpoint memory embeddings will call, resolved
// through the SAME session-key disambiguation the actor is seeded with, so
// a slug that collides across platforms cannot resolve to the twin (H-b).
let embed_platform = {
let models = models_manager.models();
crate::agent::models::platform_for_slug(
&models,
crate::agent::models::selected_catalog_key_for_spawn(&models, &session_model_id)
.as_deref(),
&sampling_config.model,
)
};
let session_pruning_config: crate::config::PruningConfig = memory_config.as_ref().map_or_else(
|| crate::config::PruningConfig {
enabled: false,
@@ -618,16 +631,11 @@ pub(crate) async fn spawn_session_actor(
watcher,
stale_claim_secs: watcher_config.stale_claim_secs,
search_source: "tool",
api_key_provider: api_key_provider.clone(),
auth_credentials: auth_manager.as_ref().map(|am| {
std::sync::Arc::new(
crate::auth::credential_provider::ShellAuthCredentialProvider::new(
am.clone(),
None,
None,
),
) as std::sync::Arc<dyn kigi_auth::AuthCredentialProvider>
}),
embedding_credentials: crate::auth::credential_provider::embedding_session_credentials(
&embed_base_url,
embed_platform,
&models_manager.credential_authority(),
),
};
let backend = crate::session::memory::MemoryBackendImpl::from_session_params(
storage.clone(),
@@ -1025,6 +1033,7 @@ pub(crate) async fn spawn_session_actor(
previous_model: std::cell::Cell::new(None),
compaction_mode,
verbatim_input: compaction_verbatim_input,
tool_choice: compaction_tool_choice,
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
@@ -1108,6 +1117,7 @@ pub(crate) async fn spawn_session_actor(
goal_tracker,
graph_enabled,
graph_tracker,
swarm_mode: std::cell::Cell::new(Default::default()),
graph_concurrency: effective_config.resolve_graph_concurrency(),
graph_node_rounds: effective_config.resolve_graph_node_rounds(),
graph_replan_cap: effective_config.resolve_graph_replan_cap(),
@@ -1287,8 +1297,11 @@ pub(crate) async fn spawn_session_actor(
.map(|mc| mc.embedding.clone())
.unwrap_or_default();
let embed_dims = embed_config.dimensions;
let sampling_base_url = embed_base_url.clone();
let sampling_api_key = embed_api_key.clone();
// The session's own params, so the background reindex embeds through
// the same endpoint-scoped credential as every foreground path — a
// second locally-built provider would re-derive credentials outside
// the chokepoint and would carry a static key with no 401 refresh.
let reindex_params = session.memory.backend_params.clone();
let session_id_for_reindex = session_info.id.to_string();
let chunks_added_counter = session.memory.chunks_added.clone();
tokio::task::spawn_local(async move {
@@ -1311,13 +1324,8 @@ pub(crate) async fn spawn_session_actor(
target : kigi_log::memory_log::TARGET, files = files.len(),
"MEMORY_REINDEX: background reindex complete"
);
if let Some(api_key) = sampling_api_key
&& let Some(provider) =
crate::session::memory::embedding::ApiEmbeddingProvider::from_session(
&embed_config,
sampling_base_url,
api_key,
)
if let Some(ref params) = reindex_params
&& let Some(provider) = params.make_embedding_provider().await
{
crate::session::memory::embed_missing_chunks(&index, &provider).await;
}
@@ -1529,6 +1537,7 @@ pub(crate) async fn spawn_session_on_thread(
system_prompt_label: String,
compaction_mode: kigi_chat_state::CompactionMode,
compaction_verbatim_input: bool,
compaction_tool_choice: crate::util::config::CompactionToolChoice,
two_pass_enabled: bool,
buffering_settings: Option<BufferingSettings>,
origin_client: Option<crate::http::OriginClientInfo>,
@@ -1630,10 +1639,23 @@ pub(crate) async fn spawn_session_on_thread(
};
(initial_last_compaction, initial_prompt_texts)
};
let rt = tokio::runtime::Builder::new_current_thread()
// Runtime construction acquires fds (epoll/kqueue, waker) and
// fails with EMFILE/EAGAIN under resource pressure — report
// through the init channel instead of panicking the thread.
let rt = match tokio::runtime::Builder::new_current_thread()
.enable_all()
.build()
.expect("session runtime");
{
Ok(rt) => rt,
Err(e) => {
tracing::error!(
error = %e,
"failed to build session runtime (resource exhaustion?)"
);
let _ = init_tx.send(Err(kigi_agent::AgentBuildError::RuntimeBuild(e)));
return;
}
};
let local = tokio::task::LocalSet::new();
local.block_on(&rt, async move {
let _trace_span = parent_traceparent.as_ref().map(|tp| {
@@ -1673,6 +1695,7 @@ pub(crate) async fn spawn_session_on_thread(
system_prompt_label,
compaction_mode,
compaction_verbatim_input,
compaction_tool_choice,
two_pass_enabled,
buffering_settings,
origin_client,
@@ -1753,8 +1776,19 @@ pub(crate) async fn spawn_session_on_thread(
}));
let _ = session_done_rx.await;
});
})
.expect("spawn session thread");
});
let join_handle = match join_handle {
Ok(handle) => handle,
Err(e) => {
tracing::error!(
error = %e,
"failed to spawn session thread (thread/PID limit or memory pressure?)"
);
return Err(
acp::Error::internal_error().data(format!("failed to spawn session thread: {e}"))
);
}
};
let init = init_rx
.await
.map_err(|_| {
crates/codegen/kigi-shell/src/session/acp_session_impl/stationarity_seam.rs
+120
View File
@@ -0,0 +1,120 @@
//! Turn-loop seam for the stationarity detector.
use super::*;
use crate::session::stationarity::{
IdenticalToolCallRun, NUDGE_AFTER_IDENTICAL_TOOL_CALLS, command_is_true, hash_batch,
};
/// Sent once, naming the blocking wait a poller should use.
const STATIONARITY_NUDGE: &str = "\
You have called the same tool with the same arguments repeatedly and are in a \
polling loop. Stop repeating that call. If you are waiting on a background \
task, block on it with the wait-tasks tool instead of re-checking its output; \
if you are waiting on anything else, sleep once and check once. If you cannot \
make progress, stop and tell the user what you are waiting for. This turn will \
be halted automatically if the identical call keeps repeating.";
impl SessionActor {
/// Enforces the ceilings; `Some` halts the turn. Silent but logged.
pub(crate) async fn observe_tool_call_stationarity(
self: &Arc<Self>,
run: &mut IdenticalToolCallRun,
tool_calls: &[kigi_sampling_types::conversation::ToolCall],
loop_index: usize,
) -> Option<StationarityHalt> {
let batch_hash = hash_batch(
tool_calls
.iter()
.map(|tc| (tc.name.as_str(), tc.arguments.as_ref())),
);
let tool_name = tool_calls
.first()
.map(|tc| tc.name.clone())
.unwrap_or_default();
let is_true_noop = self.is_run_true_step(tool_calls).await;
let run_len = run.observe(batch_hash, &tool_name, is_true_noop);
if run_len == NUDGE_AFTER_IDENTICAL_TOOL_CALLS {
tracing::warn!(
tool_name = %run.tool_name(),
run_len,
loop_index,
"action stationarity: nudging a repeating tool call"
);
kigi_log::unified_log::warn(
"shell.turn.action_stationarity_nudge",
Some(self.session_info.id.0.as_ref()),
Some(serde_json::json!({
"tool_name": run.tool_name(),
"run_len": run_len,
"loop_index": loop_index,
})),
);
self.push_system_reminder(STATIONARITY_NUDGE);
}
if run_len < run.hard_stop_threshold() {
return None;
}
tracing::warn!(
tool_name = %run.tool_name(),
run_len,
loop_index,
true_noop = run.is_true_noop_run(),
"action stationarity: halting the turn"
);
kigi_log::unified_log::warn(
"shell.turn.action_stationarity_stop",
Some(self.session_info.id.0.as_ref()),
Some(serde_json::json!({
"tool_name": run.tool_name(),
"run_len": run_len,
"loop_index": loop_index,
"true_noop": run.is_true_noop_run(),
})),
);
Some(StationarityHalt {
tool_name: run.tool_name().to_string(),
run_len,
true_noop: run.is_true_noop_run(),
})
}
/// Whether this batch is a single shell call that does nothing.
///
/// Size-gated, not name-gated: the shell tool is renamed
/// `run_terminal_command`, so a name gate would silently disable this.
/// Parse failure fails open; multi-call batches use the ordinary ceiling.
async fn is_run_true_step(
&self,
tool_calls: &[kigi_sampling_types::conversation::ToolCall],
) -> bool {
/// Bound on `{"command":"true"}` plus sibling fields.
const MAX_NOOP_ARGS_BYTES: usize = 512;
let [tc] = tool_calls else {
return false;
};
if tc.arguments.as_ref().len() > MAX_NOOP_ARGS_BYTES {
return false;
}
let Ok(args) = serde_json::from_str::<serde_json::Value>(tc.arguments.as_ref()) else {
return false;
};
let Ok(input) = self.tool_bridge_handle().try_parse(&tc.name, args).await else {
return false;
};
matches!(
input,
kigi_tools::types::tool_io::ToolInput::Bash(ref b) if command_is_true(&b.command)
)
}
}
/// Why a turn was halted, carried to its outcome.
pub(crate) struct StationarityHalt {
pub(crate) tool_name: String,
pub(crate) run_len: u32,
pub(crate) true_noop: bool,
}
crates/codegen/kigi-shell/src/session/acp_session_impl/swarm.rs
+80
View File
@@ -0,0 +1,80 @@
//! Swarm-mode seam: arming/disarming the mode and the reminder it injects.
use super::*;
use crate::session::swarm_mode::{SWARM_ENTER_REMINDER, SWARM_EXIT_REMINDER, SwarmTrigger};
impl SessionActor {
/// Arms (`Some(trigger)`) or disarms (`None`) swarm mode, injecting or
/// retracting the doctrine exactly once, and returns the line to show.
pub(crate) fn apply_swarm_mode(self: &Arc<Self>, trigger: Option<SwarmTrigger>) -> String {
let mut mode = self.swarm_mode.get();
let message = match trigger {
Some(trigger) => {
if mode.enter(trigger) {
self.inject_swarm_reminder(SWARM_ENTER_REMINDER);
}
"Swarm mode on: work will be split across a fleet of subagents. \
`/swarm off` to stop."
}
None => {
// Unconditional, unlike the automatic expiry: the doctrine
// rides the conversation and therefore survives a resume, a
// fork and a compaction that the in-memory flag does not. If
// the user explicitly asks for it off, the retraction has to
// reach the model even when this session never saw it armed.
mode.exit();
self.inject_swarm_reminder(SWARM_EXIT_REMINDER);
"Swarm mode off."
}
};
self.swarm_mode.set(mode);
message.to_string()
}
/// A turn-scoped guard that disarms a per-turn swarm mode however the turn
/// ends.
///
/// The post-loop call site is not enough: a user interrupt ABORTS the turn
/// future (`cancel_running_task` → `JoinHandle::abort`), dropping it at its
/// current await point, and several `?` paths return before the loop's end.
/// Each of those leaks a `/swarm <task>` mode into the user's next,
/// unrelated prompt. The goal engine hit the same class and compensates
/// inside the cancel path; a guard is the version that cannot be forgotten
/// at a new exit.
pub(crate) fn swarm_turn_guard(self: &Arc<Self>) -> SwarmTurnGuard {
SwarmTurnGuard {
session: self.clone(),
}
}
/// Disarms at a turn boundary when the trigger was per-turn.
pub(crate) fn expire_swarm_mode_at_turn_end(self: &Arc<Self>) {
if !self.swarm_mode.get().expires_at_turn_end() {
return;
}
let mut mode = self.swarm_mode.get();
if mode.exit() {
self.inject_swarm_reminder(SWARM_EXIT_REMINDER);
}
self.swarm_mode.set(mode);
}
/// The doctrine rides the session's existing `<system-reminder>` channel,
/// so it is tagged the same way every other reminder is and needs no
/// second injection path of its own.
fn inject_swarm_reminder(self: &Arc<Self>, text: &str) {
self.push_system_reminder(text);
}
}
/// Runs [`SessionActor::expire_swarm_mode_at_turn_end`] on every turn exit,
/// including an aborted future.
pub(crate) struct SwarmTurnGuard {
session: Arc<SessionActor>,
}
impl Drop for SwarmTurnGuard {
fn drop(&mut self) {
self.session.expire_swarm_mode_at_turn_end();
}
}
crates/codegen/kigi-shell/src/session/acp_session_impl/tool_calls.rs
+14
View File
@@ -957,6 +957,8 @@ impl SessionActor {
Some(self.session_info.id.0.to_string()),
None,
None,
self.tool_context.cwd.clone(),
self.display_cwd.get().map(std::path::PathBuf::from),
)
.await
};
@@ -1546,6 +1548,18 @@ impl SessionActor {
vec![],
vec![],
),
// Without an explicit arm the catch-all below titles this "Tool
// call" — for an entry that can hold the turn for the whole swarm.
ToolInput::AgentSwarm(swarm) => (
format!(
"{} ({} members)",
swarm.description,
swarm.items.len() + swarm.resume_agent_ids.len()
),
acp::ToolKind::Other,
vec![],
vec![],
),
ToolInput::EnterPlanMode(_) => (
"Plan: Enter".to_string(),
acp::ToolKind::Other,
crates/codegen/kigi-shell/src/session/acp_session_impl/turn.rs
+83 -20
View File
@@ -217,6 +217,10 @@ impl SessionActor {
persist_ack: Option<oneshot::Sender<()>>,
) -> PromptTurnResult {
let handle_prompt_start = std::time::Instant::now();
// Armed before anything can arm swarm mode, so a `/swarm <task>` mode
// is disarmed on EVERY exit from this turn — including the abort a
// user interrupt performs, which never reaches post-loop code.
let _swarm_turn_guard = self.swarm_turn_guard();
let prompt_length: usize = prompt_blocks
.iter()
.map(|b| match b {
@@ -301,6 +305,13 @@ impl SessionActor {
span.record("command_source", "builtin");
}
match action {
// `/swarm <task>` arms the mode for this turn only and
// sends the task as the prompt, so the doctrine is in the
// conversation before the model reads the work.
BuiltinAction::SwarmTask { prompt } => {
self.apply_swarm_mode(Some(crate::session::swarm_mode::SwarmTrigger::Task));
vec![text_block(prompt)]
}
BuiltinAction::GoalSet {
objective,
token_budget,
@@ -670,22 +681,14 @@ impl SessionActor {
.await
.is_some()
{
let (flush_tx, flush_rx) = oneshot::channel();
if self
.notifications
.persistence_tx
.send(PersistenceMsg::FlushAndAck {
respond_to: flush_tx,
})
.is_ok()
&& flush_rx.await.is_ok()
{
let _ = ack.send(());
} else {
tracing::error!(
session_id = % self.session_info.id.0, prompt_id = %
prompt_id, "persist_ack flush barrier failed"
);
match self.persistence_barrier().await {
Ok(()) => {
let _ = ack.send(());
}
Err(error) => tracing::error!(
session_id = % self.session_info.id.0, prompt_id = % prompt_id,
%error, "persist_ack flush barrier failed"
),
}
} else {
tracing::error!(
@@ -779,7 +782,7 @@ impl SessionActor {
let turn_tool_count = self.events.tool_count_this_turn();
let bridge_outcome = turn_result_to_hook_outcome(&result);
match &result {
Ok(TurnOutcome::Completed { .. }) => {
Ok(TurnOutcome::Completed { .. }) | Ok(TurnOutcome::StationarityHalted { .. }) => {
self.emit_turn_ended(
crate::session::events::TurnOutcomeLabel::Completed,
None,
@@ -869,7 +872,9 @@ impl SessionActor {
let doom_tally = std::mem::take(&mut *self.doom_loop_turn_tally.lock());
doom_tally.fired();
let stop_reason_str = match &result {
Ok(TurnOutcome::Completed { .. }) => "end_turn",
Ok(TurnOutcome::Completed { .. }) | Ok(TurnOutcome::StationarityHalted { .. }) => {
"end_turn"
}
Ok(TurnOutcome::Cancelled { .. }) | Ok(TurnOutcome::MaxTurnsReached { .. }) => {
"cancelled"
}
@@ -885,7 +890,7 @@ impl SessionActor {
)
.await;
match &result {
Ok(TurnOutcome::Completed { .. }) => {
Ok(TurnOutcome::Completed { .. }) | Ok(TurnOutcome::StationarityHalted { .. }) => {
for contributor in self.extension_registry.turn_lifecycle_contributors() {
contributor
.on_turn_done(&kigi_agent_lifecycle::TurnDoneInput)
@@ -970,6 +975,26 @@ impl SessionActor {
PromptCompletionKind::MaxTurnsReached { limit },
None,
),
// `EndTurn`, not `Cancelled`: the client must not render
// an interrupted turn for something nobody interrupted.
// The halt's detail rides `completion_kind`, which is
// where a bug report can still read it.
TurnOutcome::StationarityHalted {
snapshot,
tool_name,
run_len,
true_noop,
..
} => (
acp::StopReason::EndTurn,
*snapshot,
PromptCompletionKind::StationarityHalted {
tool_name,
run_len,
true_noop,
},
None,
),
};
if let Some(snapshot) = snapshot.as_mut() {
self.apply_prompt_modes_to_snapshot(snapshot);
@@ -1271,7 +1296,11 @@ impl SessionActor {
let mut result = self
.process_conversation_turn(req_id, json_schema.clone())
.await;
if matches!(result, Ok(TurnOutcome::MaxTurnsReached { .. })) {
// Harness stopped the turn; retrying re-enters the same wall.
if matches!(
result,
Ok(TurnOutcome::MaxTurnsReached { .. }) | Ok(TurnOutcome::StationarityHalted { .. })
) {
return result;
}
if let Ok(TurnOutcome::Completed {
@@ -1588,6 +1617,10 @@ impl SessionActor {
self.record_turn_model().await;
let mut metrics_drop_guard = TurnMetrics::new();
let mut turn_tools_called: Vec<String> = Vec::new();
let mut identical_tool_calls =
crate::session::stationarity::IdenticalToolCallRun::default();
// Retained across execute: observed only after results land.
let mut last_batch: Vec<kigi_sampling_types::conversation::ToolCall> = Vec::new();
let mut tool_turn_count: usize = 1;
let mut loop_index: u32 = 0;
let mut todo_gate_fires: u32 = 0;
@@ -2004,6 +2037,8 @@ impl SessionActor {
}
turn_tools_called.push(tc.name.clone());
}
last_batch.clear();
last_batch.extend(tool_calls.iter().cloned());
let tool_call_responses: Vec<ToolCallResponse> = tool_calls
.into_iter()
.map(|tc| ToolCallResponse {
@@ -2046,6 +2081,34 @@ impl SessionActor {
}
_ => {}
}
// After execute: every call has a result, so nothing dangles
// and the nudge cannot trigger the "cancelled" repair.
// Ok-gated: one `?` inside can leave a call resultless.
if execute_tool_calls_result.is_ok()
&& let Some(halt) = self
.observe_tool_call_stationarity(
&mut identical_tool_calls,
&last_batch,
tool_turn_count,
)
.await
{
let snapshot = self
.finalize_turn_bookkeeping(
req_id,
conv_turn_start,
&turn_span_totals,
model_fingerprint.clone(),
)
.await;
return Ok(TurnOutcome::StationarityHalted {
snapshot: Box::new(snapshot),
tools_called: std::mem::take(&mut turn_tools_called),
tool_name: halt.tool_name,
run_len: halt.run_len,
true_noop: halt.true_noop,
});
}
let next_turn = tool_turn_count + 1;
if let Some(limit) = self.max_turns
&& next_turn > limit
crates/codegen/kigi-shell/src/session/acp_session_impl/types.rs
+13
View File
@@ -62,6 +62,19 @@ pub(crate) enum TurnOutcome {
},
/// The `--max-turns` limit was reached after a tool-execution cycle.
MaxTurnsReached { limit: usize },
/// One tool call repeated past its ceiling; the turn was halted.
///
/// Groups with [`Self::Completed`], NOT [`Self::Cancelled`]: nobody
/// cancelled anything. As a cancellation it would report
/// `StopReason::Cancelled`, fire the abort lifecycle, kill the turn's
/// subagents, grow the goal back-off streak, and let recovery re-run it.
StationarityHalted {
snapshot: Box<Option<TurnDeltaSnapshot>>,
tools_called: Vec<String>,
tool_name: String,
run_len: u32,
true_noop: bool,
},
}
#[derive(Debug)]
crates/codegen/kigi-shell/src/session/acp_session_impl/updates.rs
+11 -2
View File
@@ -333,15 +333,24 @@ impl SessionActor {
if let Err(e) = crate::session::replay_events::flush_replay_actor(&self.event_tx).await {
tracing::warn!(?e, "flush_replay_actor failed");
}
if let Err(error) = self.persistence_barrier().await {
tracing::error!(%error, "persistence flush failed");
}
}
/// The `FlushAndAck` durability barrier: `Ok` only once every prior
/// persistence write is on stable media.
pub(super) async fn persistence_barrier(&self) -> std::io::Result<()> {
let (tx, rx) = tokio::sync::oneshot::channel();
if self
.notifications
.persistence_tx
.send(PersistenceMsg::FlushAndAck { respond_to: tx })
.is_ok()
.is_err()
{
let _ = rx.await;
return Err(std::io::Error::other("persistence actor gone"));
}
rx.await
.unwrap_or_else(|_| Err(std::io::Error::other("persistence actor gone")))
}
/// Extracts the update type name and relevant parameters for logging
fn extract_update_info(
crates/codegen/kigi-shell/src/session/acp_session_tests/cancel_running_task_tests.rs
+12 -5
View File
@@ -106,6 +106,7 @@ async fn persist_ack_waits_for_disk_flush_before_success() {
tokio_util::sync::CancellationToken::new(),
);
let actor = Arc::new(SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info,
auth_method_id: test_auth_method_id("test-auth"),
model_auth_facts: std::cell::RefCell::new(None),
@@ -156,6 +157,7 @@ async fn persist_ack_waits_for_disk_flush_before_success() {
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
@@ -433,7 +435,7 @@ async fn first_turn_memory_injection_persists_to_chat_history() {
respond_to: flush_tx,
})
.unwrap();
flush_rx.await.unwrap();
flush_rx.await.unwrap().unwrap();
let loaded = storage
.load_session_without_updates(&session_info)
.await
@@ -555,11 +557,11 @@ async fn first_turn_memory_injection_disabled_does_not_persist_to_chat_history()
watcher: None,
stale_claim_secs: 60,
search_source: "tool",
api_key_provider: None,
auth_credentials: None,
embedding_credentials: kigi_memory::EndpointScopedCredentials::none(),
};
let (event_tx, _event_rx) = tokio::sync::mpsc::unbounded_channel::<SessionEvent>();
let actor = Arc::new(SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: session_info.clone(),
auth_method_id: test_auth_method_id("test-auth"),
model_auth_facts: std::cell::RefCell::new(None),
@@ -610,6 +612,7 @@ async fn first_turn_memory_injection_disabled_does_not_persist_to_chat_history()
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
@@ -754,7 +757,7 @@ async fn first_turn_memory_injection_disabled_does_not_persist_to_chat_history()
respond_to: flush_tx,
})
.unwrap();
flush_rx.await.unwrap();
flush_rx.await.unwrap().unwrap();
let storage = crate::session::storage::JsonlStorageAdapter::with_explicit_session_dir(
session_dir.path().to_path_buf(),
);
@@ -821,6 +824,7 @@ async fn cancel_running_task_teardown_clears_running_and_pending_work() {
)
.await;
let actor = SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-cancel"),
cwd: cwd.as_str().to_string(),
@@ -869,6 +873,7 @@ async fn cancel_running_task_teardown_clears_running_and_pending_work() {
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
@@ -1305,7 +1310,7 @@ async fn actor_with_persistence_drain() -> std::sync::Arc<SessionActor> {
tokio::task::spawn_local(async move {
while let Some(msg) = persistence_rx.recv().await {
if let PersistenceMsg::FlushAndAck { respond_to } = msg {
let _ = respond_to.send(());
let _ = respond_to.send(Ok(()));
}
}
});
@@ -1814,6 +1819,7 @@ async fn cancel_propagates_to_sampler_handle_so_no_further_emission() {
)
.await;
let actor = SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-cancel-sampler"),
cwd: cwd.as_str().to_string(),
@@ -1862,6 +1868,7 @@ async fn cancel_propagates_to_sampler_handle_so_no_further_emission() {
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
crates/codegen/kigi-shell/src/session/acp_session_tests/idle_resume_tests.rs
+2
View File
@@ -115,6 +115,7 @@ async fn test_e2e_idle_resume_refreshes_model_metadata() {
});
tokio::time::sleep(std::time::Duration::from_millis(50)).await;
let actor = SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-idle-resume"),
cwd: cwd.as_str().to_string(),
@@ -175,6 +176,7 @@ async fn test_e2e_idle_resume_refreshes_model_metadata() {
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
crates/codegen/kigi-shell/src/session/acp_session_tests/inline_auto_compact_flow_tests.rs
+8 -4
View File
@@ -64,6 +64,7 @@ async fn create_test_actor(
);
chat_state_handle.record_token_usage(total_tokens);
SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-auto-compact"),
cwd: cwd.as_str().to_string(),
@@ -111,6 +112,7 @@ async fn create_test_actor(
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
@@ -393,8 +395,7 @@ fn initial_injection_backend_params_use_override_min_score() {
watcher: None,
stale_claim_secs: 60,
search_source: "tool",
api_key_provider: None,
auth_credentials: None,
embedding_credentials: kigi_memory::EndpointScopedCredentials::none(),
};
let initial_injection = crate::config::MemoryInitialInjectionConfig {
enabled: true,
@@ -422,8 +423,7 @@ fn initial_injection_backend_params_preserve_default_zero_min_score() {
watcher: None,
stale_claim_secs: 60,
search_source: "tool",
api_key_provider: None,
auth_credentials: None,
embedding_credentials: kigi_memory::EndpointScopedCredentials::none(),
};
let (adjusted, effective_min_score) = build_initial_injection_backend_params(
&params,
@@ -495,6 +495,7 @@ async fn create_test_actor_with_memory(
.as_ref()
.map_or_else(Default::default, |mc| mc.initial_injection.clone());
SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-memory"),
cwd: cwd.as_str().to_string(),
@@ -539,6 +540,7 @@ async fn create_test_actor_with_memory(
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
@@ -1237,6 +1239,7 @@ async fn test_e2e_idle_resume_refreshes_model_metadata() {
});
tokio::time::sleep(std::time::Duration::from_millis(50)).await;
let actor = SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-idle-resume"),
cwd: cwd.as_str().to_string(),
@@ -1297,6 +1300,7 @@ async fn test_e2e_idle_resume_refreshes_model_metadata() {
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
crates/codegen/kigi-shell/src/session/acp_session_tests/memory_config_tests.rs
+5 -6
View File
@@ -18,8 +18,7 @@ fn initial_injection_backend_params_use_override_min_score() {
watcher: None,
stale_claim_secs: 60,
search_source: "tool",
api_key_provider: None,
auth_credentials: None,
embedding_credentials: kigi_memory::EndpointScopedCredentials::none(),
};
let initial_injection = crate::config::MemoryInitialInjectionConfig {
enabled: true,
@@ -47,8 +46,7 @@ fn initial_injection_backend_params_preserve_default_zero_min_score() {
watcher: None,
stale_claim_secs: 60,
search_source: "tool",
api_key_provider: None,
auth_credentials: None,
embedding_credentials: kigi_memory::EndpointScopedCredentials::none(),
};
let (adjusted, effective_min_score) = build_initial_injection_backend_params(
&params,
@@ -121,6 +119,7 @@ async fn create_test_actor_with_memory(
.as_ref()
.map_or_else(Default::default, |mc| mc.initial_injection.clone());
SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-memory"),
cwd: cwd.as_str().to_string(),
@@ -164,6 +163,7 @@ async fn create_test_actor_with_memory(
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
@@ -524,8 +524,7 @@ async fn create_injection_ready_actor(
watcher: None,
stale_claim_secs: 60,
search_source: "tool",
api_key_provider: None,
auth_credentials: None,
embedding_credentials: kigi_memory::EndpointScopedCredentials::none(),
});
actor
.chat_state_handle
crates/codegen/kigi-shell/src/session/acp_session_tests/model_switch_window_tests.rs
+24
View File
@@ -0,0 +1,24 @@
//! `handle_set_session_model` reports the session's effective context
//! window so clients repaint their header without waiting for a turn.
use super::support::*;
#[tokio::test(flavor = "current_thread")]
async fn set_session_model_returns_the_effective_context_window() {
let local = tokio::task::LocalSet::new();
local
.run_until(async {
let (actor, _rx) = build_actor().await;
let cfg = kigi_sampler::SamplerConfig {
model: "narrow".to_string(),
context_window: 128_000,
..Default::default()
};
let (model_id, window) = actor
.handle_set_session_model(cfg, None, false, false, true, 85)
.await
.expect("model switch");
assert_eq!(model_id.0.as_ref(), "narrow");
assert_eq!(window.get(), 128_000);
})
.await;
}
crates/codegen/kigi-shell/src/session/acp_session_tests/permission_auto_mode_tests.rs
+4
View File
@@ -82,6 +82,8 @@ async fn set_auto_mode_path_wires_live_side_query_via_session_actor() {
None,
None,
None,
session.tool_context.cwd.clone(),
None,
)
.await;
// cargo is heuristic-allow when sampling fails; must not be Prompt-only
@@ -99,6 +101,8 @@ async fn set_auto_mode_path_wires_live_side_query_via_session_actor() {
None,
None,
None,
session.tool_context.cwd.clone(),
None,
)
.await;
assert!(
crates/codegen/kigi-shell/src/session/acp_session_tests/replay_buffer_send_update_tests.rs
+2
View File
@@ -72,6 +72,7 @@ pub(super) async fn make_replay_send_update_fixture() -> ReplaySendUpdateFixture
});
let (event_tx, event_rx) = mpsc::unbounded_channel::<SessionEvent>();
let actor = SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-session"),
cwd: cwd.as_str().to_string(),
@@ -115,6 +116,7 @@ pub(super) async fn make_replay_send_update_fixture() -> ReplaySendUpdateFixture
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
crates/codegen/kigi-shell/src/session/acp_session_tests/support.rs
+2
View File
@@ -182,6 +182,7 @@ pub(crate) async fn create_test_actor_ex(
chat_state_handle.record_token_usage(total_tokens);
let (goal_update_tx, goal_update_rx) = tokio::sync::mpsc::unbounded_channel();
let actor = SessionActor {
swarm_mode: std::cell::Cell::new(Default::default()),
session_info: SessionInfo {
id: acp::SessionId::new("test-actor"),
cwd: cwd.as_str().to_string(),
@@ -225,6 +226,7 @@ pub(crate) async fn create_test_actor_ex(
previous_model: std::cell::Cell::new(None),
compaction_mode: kigi_chat_state::CompactionMode::Transcript,
verbatim_input: true,
tool_choice: Default::default(),
prefire: crate::session::compaction_config::PrefireState::default(),
prefix_released: std::sync::atomic::AtomicBool::new(false),
},
Loaded 100 of 268 files, more files were not shown because too many files have changed in this diff. Show more