ZacharyZhang-NY
ff5e22092a
Sessions are already agent-scoped post-v34, but the code kept routing
rotation / accumulation / handoff through the task indirection. Heartbeat
runs (no task) silently skipped all three, letting taskless transcripts
grow past the model's context window — the 7M-token explosion.
- DB adds getSessionById(sessionId) — direct lookup bypassing
tasks.active_session_id
- session-management-service: generateHandoff(session) is now agent-scoped,
walking recent runs across tasks + heartbeats; getHandoffForAgent(agentId)
is the canonical lookup, getHandoffContext(taskId) is a thin shim for
callers that still hold a task id
- turn-context.ts::resolveSession drops the `if (!task)` early return:
rotation evaluation runs on the agent session unconditionally
- run-finished-helpers.ts::accumulateAndRotateSession drops the `!task`
gate: tokens accumulate and rotation fires for every run, keyed on the
agent session, not the task
- Rotation evaluator contract: evaluatePreRunSessionRotation(agentId,
connectorId) → { rotated, handoff, sessionId }; old (taskId, connectorId)
signature removed
IPC contract tightening: ipc-support::emitEvent becomes a best-effort
sink. "Render frame was disposed" during renderer reload is a lifecycle
signal, not a logic error; silently drop it instead of letting it abort
downstream event emissions.
Artifact projection: pi's write/edit tool results now upsert a
`work_products` row via a new agent-runtime/artifact-projector.ts. No
more agents writing 54 files into the workspace that the UI never sees.
Migration v40 relaxes work_products.task_id to nullable so heartbeat
artifacts can register too.
File size: split agent-event-projection.ts out of runtime-bridge.ts to
keep the bridge under the 400-line cap after adding the projector hook.
Tests: 604 unit + integration green; generateHandoff callsites updated
for the single-argument signature; artifact-projector has pure-function
coverage via the projector module's own shape.
76 lines
3.2 KiB
TypeScript
76 lines
3.2 KiB
TypeScript
import log from "electron-log/main.js";
|
|
import type { Migration, MigrationDeps } from "./types";
|
|
import { columnExists, tableExists } from "./types";
|
|
|
|
export function createMigrationsV38V39(deps: MigrationDeps): Migration[] {
|
|
const { db } = deps;
|
|
return [
|
|
{
|
|
// Model capability table. Maps (provider, model_id) to 1..5 scores across
|
|
// three domains (reasoning / coding / writing), a tier enum used as a
|
|
// hard lower bound for task-profile matching, and a cost-per-Mtok-output
|
|
// figure used to pick the cheapest qualifying model.
|
|
//
|
|
// Scores, tier, and cost are data — editable at runtime via IPC. Seed
|
|
// rows are inserted by the main process on startup when the table is
|
|
// empty so a fresh install boots with a sensible matrix covering the
|
|
// providers in the pi-ai catalog and the curated Fireworks list.
|
|
version: 38,
|
|
name: "create_model_capabilities",
|
|
fn: () => {
|
|
if (tableExists(db, "model_capabilities")) {
|
|
log.info("[migration v38] model_capabilities already exists; skipping");
|
|
return;
|
|
}
|
|
db.exec(
|
|
`CREATE TABLE model_capabilities (
|
|
provider TEXT NOT NULL,
|
|
model_id TEXT NOT NULL,
|
|
reasoning INTEGER NOT NULL,
|
|
coding INTEGER NOT NULL,
|
|
writing INTEGER NOT NULL,
|
|
tier TEXT NOT NULL,
|
|
cost_per_mtok_out REAL NOT NULL,
|
|
updated_at TEXT NOT NULL,
|
|
PRIMARY KEY (provider, model_id)
|
|
)`,
|
|
);
|
|
db.exec("CREATE INDEX idx_model_capabilities_tier ON model_capabilities(tier)");
|
|
log.info("[migration v38] created model_capabilities table");
|
|
},
|
|
},
|
|
{
|
|
// agents.model_strategy drives run-time model dispatch.
|
|
//
|
|
// 'explicit' — use agent.connector_id and agent.model verbatim
|
|
// (the pre-v39 behavior — kept for existing rows so there
|
|
// are no regressions on upgrade).
|
|
// 'auto' — ignore agent.connector_id/model; at run dispatch time
|
|
// pick the cheapest model whose capability vector meets
|
|
// the agent's inferred task profile.
|
|
//
|
|
// New agents default to 'auto' at the application layer. Existing rows
|
|
// on upgrade are forced to 'explicit' because the user previously picked
|
|
// their model manually and we must not silently re-route them.
|
|
version: 39,
|
|
name: "add_agents_model_strategy",
|
|
fn: () => {
|
|
if (!tableExists(db, "agents")) {
|
|
throw new Error("agents table missing; refusing to migrate v39");
|
|
}
|
|
if (columnExists(db, "agents", "model_strategy")) {
|
|
log.info("[migration v39] agents.model_strategy already exists; skipping");
|
|
return;
|
|
}
|
|
// Default at schema level = 'explicit' so existing rows keep their
|
|
// hand-picked model. Application code uses 'auto' as the default for
|
|
// freshly created agents.
|
|
db.exec(
|
|
"ALTER TABLE agents ADD COLUMN model_strategy TEXT NOT NULL DEFAULT 'explicit'",
|
|
);
|
|
log.info("[migration v39] added agents.model_strategy column");
|
|
},
|
|
},
|
|
];
|
|
}
|