Skip to content

Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS

fix: scope series autopilot outcome corrections to owned runs (#13094)

Adam Eivydfa42af3c768
diff --git a/server/routes/pipeline/autopilot.js b/server/routes/pipeline/autopilot.js
index 563b12f3af..f5f4d70ccc 100644
--- a/server/routes/pipeline/autopilot.js
+++ b/server/routes/pipeline/autopilot.js
@@ -280,7 +280,7 @@ router.get('/series/:id/autopilot/model-metrics', asyncHandler(async (req, res)
router.post('/series/:id/autopilot/model-outcomes', asyncHandler(async (req, res) => {
await seriesSvc.getSeries(req.params.id).catch((err) => { throw mapServiceError(err); });
const body = validateRequest(modelOutcomeSchema, req.body ?? {});
- const recorded = await recordModelOutcome(body.runId, body);
+ const recorded = await recordModelOutcome(body.runId, body, req.params.id);
if (!recorded) throw new ServerError('AI run not found', { status: 404, code: 'RUN_NOT_FOUND' });
res.json({ success: true, runId: body.runId });
}));
diff --git a/server/routes/pipeline/autopilot.test.js b/server/routes/pipeline/autopilot.test.js
new file mode 100644
index 0000000000..79e5a58818
--- /dev/null
+++ b/server/routes/pipeline/autopilot.test.js
@@ -0,0 +1,83 @@
+import { beforeEach, describe, expect, it, vi } from 'vitest';
+import express from 'express';
+import { request } from '../../lib/testHelper.js';
+import { errorMiddleware } from '../../lib/errorHandler.js';
+
+const { runs, patchRunMetadata } = vi.hoisted(() => {
+ const runs = new Map();
+ return {
+ runs,
+ patchRunMetadata: vi.fn(async (id, patch) => Object.assign(runs.get(id), patch)),
+ };
+});
+
+vi.mock('../../services/runner.js', () => ({
+ getRun: vi.fn(async (id) => runs.get(id) || null),
+ listRuns: vi.fn(),
+ patchRunMetadata,
+}));
+vi.mock('../../services/providers.js', () => ({ getAllProviders: vi.fn() }));
+vi.mock('../../services/pipeline/series.js', () => ({
+ AUTOPILOT_LLM_STAGE_KINDS: ['foundationGate'],
+ getSeries: vi.fn(async (id) => ({ id })),
+}));
+vi.mock('../../services/pipeline/seriesAutopilot.js', () => ({}));
+vi.mock('../../services/pipeline/editorialScore.js', () => ({ READINESS_GATES: ['story'] }));
+vi.mock('./shared.js', () => ({
+ mapServiceError: (err) => err,
+ providerOverrideShape: {},
+}));
+
+const autopilotRoutes = (await import('./autopilot.js')).default;
+const app = express();
+app.use(express.json());
+app.use('/api/pipeline', autopilotRoutes);
+app.use(errorMiddleware);
+
+const runId = '11111111-1111-4111-8111-111111111111';
+const outcome = {
+ runId, role: 'creative', stage: 'foundationGate', outcome: 'accepted',
+ effort: 'max', scoreBefore: 0.4, scoreAfter: 0.8,
+};
+
+beforeEach(() => {
+ runs.clear();
+ patchRunMetadata.mockClear();
+ runs.set(runId, { id: runId, seriesId: 'series-a', qualityOutcome: 'rejected' });
+});
+
+describe('Series autopilot model outcomes', () => {
+ it('refuses another series run without changing its evidence', async () => {
+ const original = { ...runs.get(runId), seriesId: 'series-b' };
+ runs.set(runId, { ...original });
+ const res = await request(app)
+ .post('/api/pipeline/series/series-a/autopilot/model-outcomes').send(outcome);
+ expect(res.status).toBe(404);
+ expect(res.body.code).toBe('RUN_NOT_FOUND');
+ expect(patchRunMetadata).not.toHaveBeenCalled();
+ expect(runs.get(runId)).toEqual(original);
+ });
+
+ it('records an outcome on the series own run', async () => {
+ const res = await request(app)
+ .post('/api/pipeline/series/series-a/autopilot/model-outcomes').send(outcome);
+ expect(res.status).toBe(200);
+ expect(res.body).toEqual({ success: true, runId });
+ expect(runs.get(runId)).toMatchObject({
+ seriesId: 'series-a', qualityOutcome: 'accepted', effort: 'max',
+ qualityScoreBefore: 0.4, qualityScoreAfter: 0.8, qualityScoreDelta: 0.4,
+ pipelineRole: 'creative', pipelineStage: 'foundationGate',
+ qualityEvaluatedAt: expect.any(String),
+ });
+ });
+
+ it.each(['missing', 'unowned'])('refuses a %s run without writing evidence', async (kind) => {
+ if (kind === 'missing') runs.delete(runId);
+ else delete runs.get(runId).seriesId;
+ const res = await request(app)
+ .post('/api/pipeline/series/series-a/autopilot/model-outcomes').send(outcome);
+ expect(res.status).toBe(404);
+ expect(res.body.code).toBe('RUN_NOT_FOUND');
+ expect(patchRunMetadata).not.toHaveBeenCalled();
+ });
+});
diff --git a/server/services/pipeline/seriesAutopilot.test.js b/server/services/pipeline/seriesAutopilot.test.js
index 9ab4f3ce9b..eb3943fc1e 100644
--- a/server/services/pipeline/seriesAutopilot.test.js
+++ b/server/services/pipeline/seriesAutopilot.test.js
@@ -5,6 +5,11 @@ import { SERIES_AUTOPILOT_DEFAULTS } from '../../lib/seriesAutopilotDefaults.js'
// ---- File-backed store (like autoRunner.test.js) so the REAL series/issues
// services run against an in-memory map instead of Postgres. ------------------
const fileStore = new Map();
+const patchRunMetadata = vi.fn(async () => {});
+vi.mock('../runner.js', async (importOriginal) => ({
+ ...(await importOriginal()),
+ patchRunMetadata,
+}));
vi.mock('../../lib/fileUtils.js', () => ({
sleep: vi.fn().mockResolvedValue(undefined),
tryReadFile: vi.fn().mockResolvedValue(null),
@@ -2724,6 +2729,9 @@ describe('autopilot conductor', () => {
const last = autopilot.__testing.runs.get(seriesId)?.lastPayload;
expect(last?.pauseKind).toBe('inapplicable');
expect(last?.discardedFindings).toEqual(blockers);
+ expect(patchRunMetadata).toHaveBeenCalledWith('structure-reverted', expect.objectContaining({
+ seriesId, autopilotSystem: 'series', pipelineRole: 'creative',
+ }));
expect(recordModelOutcome).toHaveBeenCalledWith('structure-reverted', expect.objectContaining({
role: 'creative', stage: 'foundationGate', outcome: 'rejected', target: 'structure',
}));
diff --git a/server/services/pipeline/seriesAutopilot/modelPerformance.js b/server/services/pipeline/seriesAutopilot/modelPerformance.js
index a189572426..eec73ceb97 100644
--- a/server/services/pipeline/seriesAutopilot/modelPerformance.js
+++ b/server/services/pipeline/seriesAutopilot/modelPerformance.js
@@ -171,8 +171,11 @@ export async function getModelPerformanceReport() {
};
}
-export async function recordModelOutcome(runId, patch = {}) {
- if (!runId || !await getRun(runId)) return false;
+// HTTP corrections require recorded series ownership; internal lifecycle callers
+// already own their run ids and may annotate legacy evidence without this scope.
+export async function recordModelOutcome(runId, patch = {}, seriesId) {
+ const run = runId ? await getRun(runId) : null;
+ if (!run || (seriesId !== undefined && run.seriesId !== seriesId)) return false;
const scoreBefore = finiteOrNull(patch.scoreBefore);
const scoreAfter = finiteOrNull(patch.scoreAfter);
await patchRunMetadata(runId, {
diff --git a/server/services/pipeline/seriesAutopilot/orchestrator.js b/server/services/pipeline/seriesAutopilot/orchestrator.js
index e73e6fd650..c0874a7651 100644
--- a/server/services/pipeline/seriesAutopilot/orchestrator.js
+++ b/server/services/pipeline/seriesAutopilot/orchestrator.js
@@ -154,6 +154,7 @@ export async function startSeriesAutopilot(sId, options = {}) {
const runId = randomUUID();
const record = {
runId,
+ seriesId: sId,
clients: [],
lastPayload: null,
cancelRequested: false,
diff --git a/server/services/pipeline/seriesAutopilot/session.js b/server/services/pipeline/seriesAutopilot/session.js
index 0772b90138..fa34d02468 100644
--- a/server/services/pipeline/seriesAutopilot/session.js
+++ b/server/services/pipeline/seriesAutopilot/session.js
@@ -418,6 +418,7 @@ const runLifecycle = (record, role) => ({
record.activeLlmRunId = runId;
patchRunMetadata(runId, {
autopilotSystem: 'series',
+ seriesId: record.seriesId,
autopilotRunId: record.runId || null,
pipelineStage: record.currentStep || 'unknown',
pipelineRole: role,