feat(audit): add reaction-rate drift gate thresholds
Add optional reaction match-drop and skip-increase drift thresholds, expose CLI flags, and enable conservative defaults in cadence package scripts. Includes tests and docs/state sync.
This commit is contained in:
William Valentin
@@ -23,7 +23,7 @@ The gateway provides:
|
||||
- **HTTP Server**: Serves static dashboard and handles webhook endpoints
|
||||
- **Node Capability Negotiation**: Optional companion-node role/capability registration
|
||||
|
||||
Operational note: onboarding (`flynn setup` / `flynn onboard`) now runs post-save live readiness checks (model/channel/memory/automation) and prints a guided first-success task flow. Companion CLI now also supports bootstrap-manifest export (`flynn companion --export-bootstrap <path|->`), release-bundle export (`--export-release-bundle <dir>` with optional `--signing-key`/`--signing-key-id` signature output), release-bundle verification (`--verify-release-bundle <dir>` with optional `--verify-signing-key`/`--verify-signing-key-id`/`--require-signature`), platform shell-template export (`--export-shell-template <dir>`), plus richer shell bootstrap flags for status/location/push (`--app-version`, `--latitude/--longitude`, `--push-token`, etc.) for desktop/mobile app packaging without changing JSON-RPC method/event shapes. Audit observability now includes live phase-0 baseline capture flows: `pnpm audit:phase0-baseline:live` for channel-origin windows, backend-scoped variants (`pnpm audit:phase0-baseline:live:pi` / `pnpm audit:phase0-baseline:live:native`) via `--backend`, `pnpm audit:phase0-baseline:live:gateway` (auto-detected cancel window) for gateway-origin windows, `pnpm audit:phase0-baseline:live:refresh` for one-shot refresh of all live windows (channel + gateway + backend-scoped), `pnpm audit:phase0-baseline:live:drift` for backend artifact freshness/drift gates (writing `phase0_baseline_live_backend_drift_<tag>.md/.json` reports), `pnpm audit:phase0-baseline:live:refresh:drift:rolling` for cadence runs that stamp each capture with a unique UTC timestamp tag (`YYYY-MM-DD-HHMMSS`) so drift comparisons can immediately use a prior snapshot (or externally supplied `TAG`), `pnpm audit:phase0-baseline:live:prune` / `pnpm audit:phase0-baseline:live:prune:apply` for rolling-tag artifact retention management (writing `phase0_baseline_live_prune_<tag>.md/.json` reports and retaining those prune reports as part of managed rolling families), and `pnpm audit:phase0-baseline:live:refresh:drift:rolling:prune` for one-command cadence refresh+drift+retention apply reusing the same rolling tag (non-negative integer `KEEP_PER_FAMILY` override supported for retention depth). These scripts default to current UTC-date tags unless `--tag` is explicitly provided.
|
||||
Operational note: onboarding (`flynn setup` / `flynn onboard`) now runs post-save live readiness checks (model/channel/memory/automation) and prints a guided first-success task flow. Companion CLI now also supports bootstrap-manifest export (`flynn companion --export-bootstrap <path|->`), release-bundle export (`--export-release-bundle <dir>` with optional `--signing-key`/`--signing-key-id` signature output), release-bundle verification (`--verify-release-bundle <dir>` with optional `--verify-signing-key`/`--verify-signing-key-id`/`--require-signature`), platform shell-template export (`--export-shell-template <dir>`), plus richer shell bootstrap flags for status/location/push (`--app-version`, `--latitude/--longitude`, `--push-token`, etc.) for desktop/mobile app packaging without changing JSON-RPC method/event shapes. Audit observability now includes live phase-0 baseline capture flows: `pnpm audit:phase0-baseline:live` for channel-origin windows, backend-scoped variants (`pnpm audit:phase0-baseline:live:pi` / `pnpm audit:phase0-baseline:live:native`) via `--backend`, `pnpm audit:phase0-baseline:live:gateway` (auto-detected cancel window) for gateway-origin windows, `pnpm audit:phase0-baseline:live:refresh` for one-shot refresh of all live windows (channel + gateway + backend-scoped), `pnpm audit:phase0-baseline:live:drift` for backend artifact freshness/drift gates (including optional reaction-rate thresholds, writing `phase0_baseline_live_backend_drift_<tag>.md/.json` reports), `pnpm audit:phase0-baseline:live:refresh:drift:rolling` for cadence runs that stamp each capture with a unique UTC timestamp tag (`YYYY-MM-DD-HHMMSS`) so drift comparisons can immediately use a prior snapshot (or externally supplied `TAG`), `pnpm audit:phase0-baseline:live:prune` / `pnpm audit:phase0-baseline:live:prune:apply` for rolling-tag artifact retention management (writing `phase0_baseline_live_prune_<tag>.md/.json` reports and retaining those prune reports as part of managed rolling families), and `pnpm audit:phase0-baseline:live:refresh:drift:rolling:prune` for one-command cadence refresh+drift+retention apply reusing the same rolling tag (non-negative integer `KEEP_PER_FAMILY` override supported for retention depth). These scripts default to current UTC-date tags unless `--tag` is explicitly provided.
|
||||
|
||||
### Execution Model (Sessions + Per-Session Queue)
|
||||
|
||||
|
||||
@@ -170,7 +170,7 @@ Gateway streaming UX signals:
|
||||
- `pnpm audit:phase0-baseline:live:pi` and `pnpm audit:phase0-baseline:live:native` capture backend-scoped channel windows using `backend.route` timelines.
|
||||
- `pnpm audit:phase0-baseline:live:gateway` captures gateway-origin baseline windows by auto-selecting the latest cancel/cancelled session window (or use `scripts/capture-phase0-live-baseline.ts --source gateway --since ... --until ...` for explicit windows).
|
||||
- `pnpm audit:phase0-baseline:live:refresh` runs channel + gateway + backend-scoped (`pi_embedded` and `native`) capture commands in one cadence step.
|
||||
- `pnpm audit:phase0-baseline:live:drift` evaluates backend-scoped artifact freshness/drift gates and writes `docs/plans/artifacts/phase0_baseline_live_backend_drift_<UTC-date>.md/.json`; `pnpm audit:phase0-baseline:live:refresh:drift` runs capture + drift checks in one cadence step.
|
||||
- `pnpm audit:phase0-baseline:live:drift` evaluates backend-scoped artifact freshness/drift gates (including optional reaction match/skip rate thresholds) and writes `docs/plans/artifacts/phase0_baseline_live_backend_drift_<UTC-date>.md/.json`; `pnpm audit:phase0-baseline:live:refresh:drift` runs capture + drift checks in one cadence step.
|
||||
- `pnpm audit:phase0-baseline:live:refresh:drift:rolling` runs the same full refresh+drift flow with a shared UTC timestamp tag (`YYYY-MM-DD-HHMMSS`) so each cadence run keeps distinct backend/drift artifacts for immediate baseline-vs-prior comparisons.
|
||||
- `pnpm audit:phase0-baseline:live:prune` provides dry-run retention planning for rolling-tag artifacts; `pnpm audit:phase0-baseline:live:prune:apply` deletes older rolling snapshots while keeping the newest tags per artifact family.
|
||||
- `pnpm audit:phase0-baseline:live:refresh:drift:rolling:prune` now reuses the rolling refresh+drift pipeline via shared `TAG` env wiring, then applies retention (non-negative integer `KEEP_PER_FAMILY`) and writes prune reports tagged to that same rolling run (`phase0_baseline_live_prune_<tag>.md/.json`).
|
||||
|
||||
@@ -35,7 +35,7 @@ If you only want the protocol surface, see `docs/api/PROTOCOL.md`.
|
||||
- Backend-scoped channel snapshots can be regenerated with `pnpm audit:phase0-baseline:live:pi` / `pnpm audit:phase0-baseline:live:native` (`--backend` filtering via `backend.route` timelines).
|
||||
- Gateway-origin phase-0 windows (including cancel-path samples) can be captured with `pnpm audit:phase0-baseline:live:gateway` (auto-detect latest cancel window) or `scripts/capture-phase0-live-baseline.ts --source gateway --since ... --until ...` for explicit bounds.
|
||||
- `pnpm audit:phase0-baseline:live:refresh` runs channel + gateway + backend-scoped (`pi_embedded` and `native`) capture paths in one command.
|
||||
- `pnpm audit:phase0-baseline:live:drift` checks backend-scoped artifact freshness/drift gates and writes `phase0_baseline_live_backend_drift_<UTC-date>.md/.json`; `pnpm audit:phase0-baseline:live:refresh:drift` chains refresh + drift checks for scheduled cadence runs.
|
||||
- `pnpm audit:phase0-baseline:live:drift` checks backend-scoped artifact freshness/drift gates (including optional reaction match/skip rate thresholds) and writes `phase0_baseline_live_backend_drift_<UTC-date>.md/.json`; `pnpm audit:phase0-baseline:live:refresh:drift` chains refresh + drift checks for scheduled cadence runs.
|
||||
- `pnpm audit:phase0-baseline:live:refresh:drift:rolling` performs the same chain using one UTC timestamp tag (`YYYY-MM-DD-HHMMSS`) across channel/gateway/backend/drift outputs so each cadence run preserves a distinct comparison point.
|
||||
- `pnpm audit:phase0-baseline:live:prune` (dry-run) and `pnpm audit:phase0-baseline:live:prune:apply` (delete) manage retention of rolling-tag artifacts to control artifact growth while preserving newest snapshots per family.
|
||||
- `pnpm audit:phase0-baseline:live:refresh:drift:rolling:prune` combines rolling refresh+drift with retention apply for one-command cron scheduling using a shared `TAG`; adjust retention depth with non-negative integer `KEEP_PER_FAMILY` and use generated `phase0_baseline_live_prune_<tag>.md/.json` artifacts for retention audit traceability.
|
||||
|
||||
@@ -203,7 +203,7 @@ Phase 0 is complete when:
|
||||
2. A baseline summary artifact is generated and committed under `docs/plans/artifacts/`.
|
||||
3. No user-visible response behavior changed compared to pre-phase baseline.
|
||||
|
||||
Follow-up status (2026-02-27): live channel-session artifacts exist under `docs/plans/artifacts/phase0_baseline_live_2026-02-27.*` via `pnpm audit:phase0-baseline:live` (anonymized IDs), and a second gateway-origin live window (including `run.cancel` + `cancel_requested`/`cancelled`) exists under `docs/plans/artifacts/phase0_baseline_live_gateway_2026-02-27.*`. Gateway window refreshes can now run via `pnpm audit:phase0-baseline:live:gateway` (auto-selected cancel window), all live windows can be refreshed together with `pnpm audit:phase0-baseline:live:refresh` (channel + gateway + backend-scoped `pi`/`native`; scheduling example included in README), backend artifact freshness/drift checks are now available via `pnpm audit:phase0-baseline:live:drift` (or chained with `pnpm audit:phase0-baseline:live:refresh:drift`) with drift report artifacts written to `docs/plans/artifacts/phase0_baseline_live_backend_drift_<tag>.{md,json}`, cadence runs can preserve distinct timestamped comparison points via `pnpm audit:phase0-baseline:live:refresh:drift:rolling` (supports shared `TAG` override), rolling-tag retention can be managed via `pnpm audit:phase0-baseline:live:prune` (dry-run) / `pnpm audit:phase0-baseline:live:prune:apply` with prune report artifacts written to `phase0_baseline_live_prune_<tag>.{md,json}` (and retained as a managed rolling family), and one-command cadence scheduling is available via `pnpm audit:phase0-baseline:live:refresh:drift:rolling:prune` (non-negative integer `KEEP_PER_FAMILY` optional override).
|
||||
Follow-up status (2026-02-27): live channel-session artifacts exist under `docs/plans/artifacts/phase0_baseline_live_2026-02-27.*` via `pnpm audit:phase0-baseline:live` (anonymized IDs), and a second gateway-origin live window (including `run.cancel` + `cancel_requested`/`cancelled`) exists under `docs/plans/artifacts/phase0_baseline_live_gateway_2026-02-27.*`. Gateway window refreshes can now run via `pnpm audit:phase0-baseline:live:gateway` (auto-selected cancel window), all live windows can be refreshed together with `pnpm audit:phase0-baseline:live:refresh` (channel + gateway + backend-scoped `pi`/`native`; scheduling example included in README), backend artifact freshness/drift checks are now available via `pnpm audit:phase0-baseline:live:drift` (or chained with `pnpm audit:phase0-baseline:live:refresh:drift`) with drift report artifacts written to `docs/plans/artifacts/phase0_baseline_live_backend_drift_<tag>.{md,json}` and optional reaction match/skip drift thresholds, cadence runs can preserve distinct timestamped comparison points via `pnpm audit:phase0-baseline:live:refresh:drift:rolling` (supports shared `TAG` override), rolling-tag retention can be managed via `pnpm audit:phase0-baseline:live:prune` (dry-run) / `pnpm audit:phase0-baseline:live:prune:apply` with prune report artifacts written to `phase0_baseline_live_prune_<tag>.{md,json}` (and retained as a managed rolling family), and one-command cadence scheduling is available via `pnpm audit:phase0-baseline:live:refresh:drift:rolling:prune` (non-negative integer `KEEP_PER_FAMILY` optional override).
|
||||
|
||||
## Subagent Model Assignment Plan
|
||||
|
||||
|
||||
@@ -533,6 +533,25 @@
|
||||
],
|
||||
"test_status": "pnpm test:run src/audit/phase0GatewayWindow.test.ts + pnpm typecheck passing"
|
||||
},
|
||||
"phase0-live-baseline-drift-reaction-threshold-gates": {
|
||||
"status": "completed",
|
||||
"date": "2026-02-27",
|
||||
"updated": "2026-02-27",
|
||||
"summary": "Extended backend drift gating with optional reaction decision thresholds (`maxReactionMatchRateDropPp`, `maxReactionSkipRateIncreasePp`), wired new CLI flags, and enabled conservative defaults in package drift cadence commands.",
|
||||
"files_modified": [
|
||||
"src/audit/phase0BaselineDrift.ts",
|
||||
"src/audit/phase0BaselineDrift.test.ts",
|
||||
"scripts/check-phase0-baseline-backend-drift.ts",
|
||||
"package.json",
|
||||
"README.md",
|
||||
"docs/api/PROTOCOL.md",
|
||||
"docs/architecture/AGENT_DIAGRAM.md",
|
||||
"docs/architecture/GATEWAY_SESSIONS_AND_QUEUE.md",
|
||||
"docs/plans/2026-02-25-phase0-instrumentation-ticket-checklist.md",
|
||||
"docs/plans/state.json"
|
||||
],
|
||||
"test_status": "pnpm test:run src/audit/phase0BaselineDrift.test.ts + pnpm typecheck passing"
|
||||
},
|
||||
"phase0-instrumentation-ticket-checklist": {
|
||||
"status": "completed",
|
||||
"date": "2026-02-25",
|
||||
|
||||
Reference in New Issue
Block a user