ZacharyZhang-NY
e37418c489
Land the crash-recoverable managed-file transaction engine (journal, staging, commit, restart recovery, opaque snapshots, private artifacts, Windows DACL hardening) and the Kiro IDE 1.0 workspace hook plugin with legacy migration. Repair five defects that kept the suite red (426 failures to zero): - Anchor the transaction state directory at ~/.elydora/transactions instead of the shared system temp directory. The recovery journal now shares fate with the runtimes it repairs, survives temp cleaners, and isolates naturally under per-test fake homes. - Skip pending-transaction recovery when no state directory exists so rejected installs never create runtime directories. - Create the runtime root's parent before hardening it, matching the Go SDK's expand behavior for fresh homes. - Stop using file timestamps as identity evidence: Windows exposes mtime/ctime with delayed cross-handle visibility. Reads now prove stability through a verification re-read; snapshots compare device, inode, size, and SHA-256. - Stop comparing synthetic Windows permission bits across stat boundaries: path stats derive execute bits from file extensions while handle stats do not, which falsely flagged every managed .cmd wrapper. Snapshots store raw modes for restoration and compare portable bits. Reservation cleanup now verifies content before discarding so a concurrent writer's file at a rollback path is preserved, and the install-failure tests assert the precise contract: no agent runtime residue, with the durable state directory allowed to remain. Gates: pytest 807 passed, mypy clean, pip check clean, wheel builds.
80 lines
20 KiB
Markdown
80 lines
20 KiB
Markdown
# Elydora Python SDK Engineering Contract
|
|
|
|
## Scope
|
|
|
|
This repository owns the published `elydora` package, its CLI, synchronous and asynchronous API clients, local signing behavior, generated hook runtimes, and Python-specific agent adapters.
|
|
|
|
## Integration Sources
|
|
|
|
- Verify every agent hook contract against current official provider documentation before changing an adapter.
|
|
- Treat `../Elydora-Open-Source/integrations/catalog.json` as the cross-product provider inventory.
|
|
- Keep the exported `INTEGRATION_TYPES` tuple aligned with that inventory, and require `integration_type` in every agent registration request.
|
|
- Keep adapter delivery claims aligned with executable tests in this repository.
|
|
- Synchronize completed provider behavior into the Node SDK, Go SDK, Open Source distribution, Console, Docs, and landing page through separate reviewed commits.
|
|
|
|
## Hook Adapter Invariants
|
|
|
|
- Preserve unrelated user configuration and remove only Elydora-owned entries.
|
|
- Parse every affected user configuration before the first write.
|
|
- Surface malformed or unreadable configuration with contextual errors and leave the original file intact.
|
|
- Write configuration through a same-directory temporary file followed by an atomic replace.
|
|
- Quote the current Python executable and generated script paths for the host shell.
|
|
- Forward official hook JSON from STDIN without reshaping provider fields.
|
|
- Use the provider's documented blocking mechanism. Command-hook providers that define exit code `2` must receive exit code `2` from the freeze guard.
|
|
- Report installation as healthy only when a complete hook contract references both generated runtime scripts and both scripts exist.
|
|
- Write Cursor hooks only to `~/.cursor/hooks.json`; keep project and enterprise sources read-only. Preserve user hooks, migrate the prior versionless Elydora contract, and own exact native `preToolUse`, `postToolUse`, and `postToolUseFailure` commands. Forward native success and failure payloads, retain PowerShell exit code `2`, set every handler to `failClosed` with a bounded timeout, and commit the guard, audit script, runtime identity, private key, and hook configuration as one rollback-capable transaction.
|
|
- Model stable, legacy, and early-access hook generations as explicit contracts. Keep their activation requirements visible in CLI output and README guidance.
|
|
- Resolve stable Kimi hooks through `$KIMI_CODE_HOME/config.toml` with `~/.kimi-code/config.toml` as the default, and treat an empty override as the default. Activate `~/.kimi/config.toml` only when the legacy home exists. Register exact `PreToolUse`, `PostToolUse`, and `PostToolUseFailure` rules with ten-second commands; stable Kimi supports its current sixteen-event schema and legacy kimi-cli supports its thirteen-event schema. Preserve TOML comments and array style, then parse the complete rendered document before staging it. Preserve the native snake_case payload, propagate exit code `2`, and keep audit delivery fail-open with owner-only error logs. Use encoded PowerShell commands on Windows and exact two-argument shell commands on Unix. Commit every detected config, generated runtime, runtime config, and private key through one rollback-capable transaction.
|
|
- Resolve Grok Build user hooks through `$GROK_HOME/hooks/*.json` with `~/.grok/hooks/*.json` as the default, and treat an empty override as the default. Keep Claude Code, Cursor, project `.grok/hooks`, plugin, and `hooks-paths` sources read-only. Register exact `PreToolUse`, `PostToolUse`, and `PostToolUseFailure` groups with ten-second command handlers, preserve the complete native camelCase payload, and emit Grok's documented deny JSON plus exit code `2` for frozen or revoked agents. Use encoded PowerShell commands on Windows and exact two-argument shell commands on Unix. Commit Grok's hook file, generated runtimes, runtime config, and private key through one rollback-capable transaction.
|
|
- Resolve Claude Code user hooks through `$CLAUDE_CONFIG_DIR/settings.json` with `~/.claude/settings.json` when the variable is absent. Match Claude Code's path resolution exactly: relative and empty values resolve from the current working directory, and literal tildes remain path segments. Keep project, local, managed, plugin, skill, and agent hook sources read-only. Validate the complete shipped handler schema, including command exec form, async rewake metadata, HTTP, MCP tool, prompt, and agent handlers. Register exact matchless `PreToolUse`, `PostToolUse`, and `PostToolUseFailure` groups with ten-second exec-form command handlers. Preserve the complete native snake_case payload and propagate freeze and revocation through exit code `2`. Commit Claude Code's user settings, generated runtimes, runtime config, and private key through one rollback-capable transaction. Reject installation while the user source sets `disableAllHooks`, then require `/hooks` and `claude doctor` verification for higher-scope policy effects.
|
|
- Resolve Gemini CLI user hooks through `$GEMINI_CLI_HOME/.gemini/settings.json` with `~/.gemini/settings.json` when the variable is absent or empty. Preserve relative and literal-tilde home values as Gemini CLI does. Keep workspace, system defaults, system overrides, and extension hooks read-only. Preserve JSON comments while rejecting trailing commas and duplicate keys. Register exact matchless `BeforeTool` and `AfterTool` groups named `elydora-guard` and `elydora-audit` with ten-second command handlers. Preserve the complete native snake_case payload, emit valid JSON on exit code `0`, and propagate freeze and revocation through exit code `2`. Use encoded PowerShell commands on Windows. Commit Gemini CLI user settings, generated runtimes, runtime config, and private key through one rollback-capable transaction. Respect `hooksConfig.enabled` and `hooksConfig.disabled`, then require `/hooks list` verification for higher-scope policy effects.
|
|
- Resolve Auggie user hooks through `~/.augment/settings.json`; keep system, workspace, local workspace, and alternate `--augment-cache-dir` settings read-only. Validate the shipped Auggie 0.33 hook schema: `PreToolUse`, `PostToolUse`, `Stop`, `SessionStart`, `SessionEnd`, `Notification`, and `PromptSubmit`; regex matchers belong to tool events, command handlers use string-array arguments and positive millisecond timeouts, and metadata flags are booleans. Generate `.cmd` wrappers on Windows and `.sh` wrappers on Unix, preserve the complete native snake_case payload, and propagate freeze and revocation through exit code `2`.
|
|
- Validate Auggie matcher syntax during installation with Node.js `new RegExp`; keep status and uninstall independent from the JavaScript validator so recovery remains available offline. Commit Auggie settings, both wrappers, generated runtimes, runtime config, and private key through one rollback-capable transaction. Require exact runtime identity, canonical private keys, physical files, and exact wrapper sources during status checks, then verify the effective configuration with `auggie tools list`.
|
|
- Treat Cline 3 CLI file hooks as the adapter contract. Resolve hooks additively from `~/Documents/Cline/Hooks`, `$CLINE_DIR/hooks` with `~/.cline/hooks` as the default, `.clinerules/hooks`, and `.cline/hooks`; write only the `$CLINE_DIR` source. Preserve the complete hybrid CLI payload, resolve generated Python runtimes from absolute shebangs, and emit one-line JSON stdout with `cancel: true` for guard exit code `2`. Commit both wrappers, generated runtimes, runtime config, and private key through one rollback-capable transaction. Status requires physical files, canonical private keys, exact runtime identity, exact generated sources, and exact wrapper sources. Hook errors, timeouts, and invalid controls fail open in Cline while remaining observable.
|
|
- Resolve Codex user hooks through `$CODEX_HOME/hooks.json` with `~/.codex/hooks.json` as the default, matching Codex's existing-directory canonicalization rule. Preserve additive user TOML, project, plugin, and managed sources. Register exact `PreToolUse` and `PostToolUse` match-all command groups with ten-second handlers, preserve the complete native payload, propagate freeze and revocation through exit code `2`, keep guard lookup and audit delivery fail-open with observable errors, and commit user hooks plus all four runtime artifacts through one rollback-capable transaction. Require `/hooks` approval for both definition hashes.
|
|
- Resolve GitHub Copilot CLI user hooks through `$COPILOT_HOME/hooks/elydora-audit.json` with `~/.copilot/hooks/elydora-audit.json` as the default, and treat an empty override as the default. Keep policy, repository, inline settings, cross-tool Claude, and plugin hook sources read-only. Evaluate `disableAllHooks` through legacy user config, user settings, Claude repository and local settings, then GitHub repository and local settings; the managed hook file flag remains independently authoritative. Register exact `preToolUse`, `postToolUse`, and `postToolUseFailure` command handlers with ten-second timeouts, preserve the complete native camelCase payload, and propagate frozen or revoked state through exit code `2`. Commit the user hook file, exact legacy migration, generated runtimes, runtime config, and private key through one rollback-capable transaction. Require physical source paths, strict hook schemas, exact runtime sources, canonical private keys, and matching runtime identity during status checks, then verify discovery with the current official Copilot native hook loader and restart active CLI sessions.
|
|
- Resolve Factory Droid 0.175.0 user hooks through `~/.factory/hooks.json`, the legacy `~/.factory/hooks/hooks.json`, then the user `settings.local.json` and `settings.json` hook containers. Current hook files use a top-level `hooks` object; retain direct event maps as an explicit legacy read contract until Factory migrates them. Register exact `PreToolUse` and `PostToolUse` groups with ten-second absolute command handlers in one active source, preserve the native snake_case payload, and propagate freeze or revocation through exit code `2`. Use PowerShell's call operator with single-quoted absolute arguments and explicit `$LASTEXITCODE` propagation on Windows, and migrate the legacy double-quoted command form through exact ownership checks. Remove exact Elydora ownership from inactive user sources while preserving JSONC comments and unknown event arrays. Keep project, folder, plugin, and organization hook definitions read-only. Resolve local settings over their matching base file, then apply system, project, folder, and user `hooksDisabled` through Factory's extension-only scalar precedence; reject system-managed `allowManagedHooksOnly`. Commit user hook sources, generated runtimes, runtime config, and private key through one rollback-capable transaction, precondition every parsed source and policy file, require exact generated runtime sources during status checks, and require `/hooks` review for the session snapshot and remote organization policy.
|
|
- Resolve Qwen Code 0.20.0 user settings through explicit `QWEN_HOME`, then the initial Qwen home `.env`, then the OS home `.env`; when bootstrap discovers a redirected home, read that home's `.env` as well. Treat empty overrides as unset while preserving their process-level ownership semantics. Read system defaults, user settings, the active trusted workspace, system overrides, and trusted-folder rules with Qwen's documented precedence; write only the user source. Validate the current twenty-one event names and command, HTTP, and prompt handler schemas while preserving future event payloads. Register exact matchless `PreToolUse`, `PostToolUse`, and `PostToolUseFailure` groups named `elydora-guard` and `elydora-audit` with ten-second millisecond timeouts. Preserve the complete native snake_case payload, propagate freeze and revocation through exit code `2`, and use PowerShell's call operator with explicit `$LASTEXITCODE` propagation on Windows.
|
|
- Commit Qwen Code user settings, generated runtimes, runtime config, and private key through one rollback-capable transaction, with snapshot preconditions for every consulted routing, policy, trust, and settings source. Status requires physical files, exact runtime sources, a canonical private key, matching runtime identity, and an effective enabled state. Verify discovery with the official 0.20.0 loader, review `/hooks`, and restart active Qwen Code sessions.
|
|
- Resolve Kiro IDE 1.0 hooks through the active workspace `.kiro/hooks/elydora-audit.json`. Register one exact `PreToolUse` and `PostToolUse` command pair in the v1 JSON schema with `.*` matchers, ten-second timeouts, and enabled state. Bind healthy current commands to the active Python executable and the expected system PowerShell launcher. Bind each Kiro IDE `agent_id` to one absolute workspace root in runtime config; require the bound workspace before install, status, or uninstall can use that runtime. Recognize recoverable v1 ownership per reserved managed name through the exact safe command grammar, an absolute Python-like executable, the Elydora runtime root, and the matching managed script name; this permits repair after interpreter relocation or hook-field drift. Preserve Kiro's complete native payload and propagate frozen or revoked state through exit code `2`. Treat `~/.kiro/hooks/elydora-audit.kiro.hook` solely as exact Elydora legacy migration input. Accept an ownerless legacy runtime config with an empty optional `token` only when that exact global legacy contract proves the same agent. Claim the absolute active workspace and remove the exact legacy hook in one installation transaction; let CLI uninstall use the same proof before runtime removal. Pin the physical identities of the workspace, `.kiro`, hooks, runtime parent, `~/.elydora`, and agent runtime directories throughout prepare, stage, commit, and recovery. Keep every read provider source, runtime file, file mode, and directory identity as a transaction precondition, including no-op installs and runtime-only updates. Commit the workspace hook file, exact legacy migration, generated runtimes, runtime config, and private key through one rollback-capable transaction, then remove only transaction-created empty directories after a failed commit. Status requires one unique complete managed pair with every managed reference accounted for, exact current commands and runtime sources, canonical private keys, matching runtime and workspace identity, and owner-only private keys plus token-bearing runtime configs on POSIX. Verify both entries in the Agent Hooks panel.
|
|
- Resolve Letta Code 0.28.13 global hooks through `${HOME:-~}/.letta/settings.json`, matching Letta's `HOME || homedir()` behavior. Keep Elydora runtimes under the shared CLI root from `os.path.expanduser("~")`, independent from Letta's configurable settings home. Keep the active workspace `.letta/settings.json` and `.letta/settings.local.json` sources read-only, deduplicate the project source when the workspace is the home directory, and apply the official `hooks.disabled` precedence: an explicit global `false` enables hooks, global `true` disables them, then project and project-local `true` values disable them. Validate all eleven current events plus command and prompt handler shapes while preserving future event payloads. Register exact `PreToolUse`, `PostToolUse`, and `PostToolUseFailure` groups with `*` matchers and ten-second millisecond timeouts. Preserve the complete native snake_case payload, propagate frozen or revoked state through exit code `2`, and surface audit delivery failures through stderr, owner-only `error.log`, and exit code `1`; Letta post events retain their documented nonblocking semantics. Use PowerShell's call operator with explicit `$LASTEXITCODE` propagation on Windows and exact two-argument shell commands on Unix. Commit global settings, generated runtimes, runtime config, and private key through one rollback-capable transaction guarded by every consulted source snapshot. Migrate only exact legacy Python command ownership, require physical sources, canonical private keys, exact runtime sources, executable identity, matching runtime identity, and an effective enabled state during status checks, then run `/hooks` and restart active sessions.
|
|
|
|
## Code Quality
|
|
|
|
- Keep the package, distribution, and CLI version in `elydora/_version.py`; build metadata must read that literal through Setuptools dynamic metadata.
|
|
- Ship `elydora/py.typed` in every wheel and verify public annotations from an installed-wheel consumer.
|
|
- Support every Python version declared in `pyproject.toml`.
|
|
- Keep production source files at or below 500 lines.
|
|
- Keep functions focused on one ownership boundary.
|
|
- Propagate unexpected errors to the CLI boundary.
|
|
- Use documented defaults only for genuinely optional configuration.
|
|
- Avoid compatibility shims without a named public or user configuration contract.
|
|
- Resolve every agent runtime directory as one physical child of `~/.elydora`; reject separators, traversal segments, cross-platform reserved names, symbolic-link directories, and linked identity configs before writes or recursive removal. Validate stored directory identity before changing host CLI configuration, and require an explicit agent ID when discovery is ambiguous.
|
|
- Accept CLI install credentials through hidden terminal input or physical owner-only credential files. Keep credentials out of process arguments, reject legacy secret options with redacted errors, and require one UTF-8 line of at most 64 KiB.
|
|
- Persist the signing key once at `~/.elydora/<agent-id>/private.key` with mode `0600`. Generated audit runtimes read and validate that physical file at execution time.
|
|
- Read runtime config, private keys, status cache, chain state, and error logs through physical-file descriptors with identity checks. Write cache and chain state atomically, and append error logs through no-follow owner-only descriptors.
|
|
- Write runtime secrets, generated scripts, and provider configuration through same-directory temporary files followed by atomic replacement. Surface permission and replacement failures.
|
|
- Serialize provider transactions across processes, flush staged and rollback files before mutation, keep existing targets continuously visible through atomic replacement, and persist an owner-only journal under `~/.elydora/transactions` for deterministic restart recovery. The state directory shares fate with the runtimes it repairs and survives failed installs; recovery never creates it for commands that may still reject their input. Commit markers retain completed changes; earlier markers restore every original target.
|
|
- Compare file identity through device, inode, size, and content hashes. Exclude timestamps everywhere and synthetic permission bits on Windows: both are stat-boundary artifacts there, not evidence. Captured snapshots still store raw modes because restoration re-applies them.
|
|
- Use capability-checked parent-dirfd removal on POSIX and write-through atomic replacement on Windows. Delete verified empty POSIX quarantine entries and containers after their identity boundary is proven.
|
|
- Interpret `max_retries` as retries after the initial attempt and reject negative or non-integer values.
|
|
- Automatically retry RFC-idempotent methods; retry non-idempotent methods only when the transport proves the request was never sent.
|
|
- Honor valid `Retry-After` delay-seconds and HTTP-date values, and release retryable responses before sleeping.
|
|
|
|
## Verification
|
|
|
|
Run the focused adapter test during development, then execute all release gates before commit:
|
|
|
|
```powershell
|
|
py -3 -m pytest tests/test_<provider>_plugin.py -q
|
|
py -3 -m pytest -q
|
|
py -3 -m mypy elydora
|
|
py -3 -m pip check
|
|
py -3 -m pip wheel . --no-deps --wheel-dir <temporary-directory>
|
|
git diff --check
|
|
```
|
|
|
|
Provider adapter tests must cover installation, idempotency, official event forwarding, blocking behavior, status, missing runtime files, uninstall ownership, and malformed configuration preservation.
|
|
|
|
Commit and push one root issue before starting the next one.
|