feat(profile): /api/profile + eligibility filter + inference framework (ADR-0014 steps 4-6)

Step 4 — /api/profile read-through API:
  GET  /api/profile          → { user, prefs, consents, contexts }
  PATCH /api/profile/prefs/:scope  upsert user_preferences (source='user')
  PATCH /api/profile/consents      grant / revoke consent keys
  PATCH /api/profile/contexts      create / activate / deactivate contexts
  Legacy consentGiven bit folded in as data:core fallback.

Step 5 — registry-driven eligibility filter:
  fetchRegistry() exported from agent-registry.ts.
  profile/eligibility.ts: getEligibleAgentIds(userId) — filters by required
  consents, silenced_in_contexts, and user_preferences[enabled=false].
  fetchOrchestratorTip filters agent_outputs to eligible set before calling
  ml/serving /recommend. Fail-closed: registry unavailable → empty set.

Step 6 — shared context-inference framework (#111) + time-of-day proof (#112):
  ml/agents/inference/: UserHistory, FeedbackEvent, run_inference().
  Framework: cold-start, min_history gating, error fallback, structured logs.
  TimeOfDayAgent v1.1.0: inferred_params=[preferred_hour]; also reads
  quiet_start/quiet_end from agent_prefs. agent_prefs injected by TS caller.
  AgentInput gains agent_prefs field.
  ml/serving: POST /agents/{agent_id}/infer endpoint.
  agent-outputs.ts computeAndStore: loads prefs before compute, calls /infer
  after, persists results (source='inferred'); user overrides never touched.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

services/api/src/routes/agent-outputs.ts

@@ -1,7 +1,7 @@
 import { Router, type Request, type Response, type IRouter } from 'express';
 import { nanoid } from 'nanoid';
 import { db } from '../db/index.js';
-import { agentOutputs, tipFeedback, tipViews } from '../db/schema.js';
+import { agentOutputs, tipFeedback, tipViews, userPreferences } from '../db/schema.js';
 import { eq, and, gt, lt } from 'drizzle-orm';
 import { config } from '../config.js';
 import { getProfile, type Profile } from '../profile/builder.js';
@@ -78,6 +78,54 @@ router.get('/active-users', async (req: Request, res: Response) => {
 
 // ── Core compute logic (used by route + scheduler) ───────────────────────────
 
+/** Load agent prefs for a user from user_preferences, merging user+inferred.
+ * User source wins: if both exist, the 'user' row is returned. */
+async function loadAgentPrefs(userId: string, agentId: string): Promise<Record<string, unknown>> {
+  const scope = `agent:${agentId}`;
+  const rows = await db
+    .select({ key: userPreferences.key, valueJson: userPreferences.valueJson, source: userPreferences.source })
+    .from(userPreferences)
+    .where(and(eq(userPreferences.userId, userId), eq(userPreferences.scope, scope)));
+
+  // Build merged dict: 'user' source takes precedence over 'inferred'
+  const merged: Record<string, { value: unknown; source: string }> = {};
+  for (const row of rows) {
+    try {
+      const value = JSON.parse(row.valueJson);
+      const existing = merged[row.key];
+      if (!existing || row.source === 'user') {
+        merged[row.key] = { value, source: row.source };
+      }
+    } catch {
+      // skip malformed
+    }
+  }
+  return Object.fromEntries(Object.entries(merged).map(([k, v]) => [k, v.value]));
+}
+
+/** Persist inferred prefs to user_preferences, skipping keys the user has explicitly set. */
+async function persistInferredPrefs(
+  userId: string,
+  agentId: string,
+  inferredPrefs: Record<string, unknown>,
+): Promise<void> {
+  if (!Object.keys(inferredPrefs).length) return;
+  const scope = `agent:${agentId}`;
+  const now = new Date().toISOString();
+  for (const [key, value] of Object.entries(inferredPrefs)) {
+    const valueJson = JSON.stringify(value);
+    await db
+      .insert(userPreferences)
+      .values({ userId, scope, key, valueJson, source: 'inferred', updatedAt: now })
+      .onConflictDoUpdate({
+        target: [userPreferences.userId, userPreferences.scope, userPreferences.key],
+        set: { valueJson, updatedAt: now },
+        // Only overwrite rows already marked inferred; user overrides are untouched.
+        setWhere: eq(userPreferences.source, 'inferred'),
+      });
+  }
+}
+
 export async function computeAndStore(userId: string, agentId: string): Promise<void> {
   let tasks: object[] = [];
   try {
@@ -111,10 +159,13 @@ export async function computeAndStore(userId: string, agentId: string): Promise<
     created_at: f.createdAt,
   }));
 
+  // Load agent prefs (user overrides + previous inferences) to inject into the compute call.
+  const agentPrefs = await loadAgentPrefs(userId, agentId);
+
   const mlResp = await fetch(`${config.ML_SERVING_URL}/agents/${agentId}/compute`, {
     method: 'POST',
     headers: { 'Content-Type': 'application/json' },
-    body: JSON.stringify({ user_id: userId, tasks, profile, feedback_history: feedbackHistory }),
+    body: JSON.stringify({ user_id: userId, tasks, profile, feedback_history: feedbackHistory, agent_prefs: agentPrefs }),
     signal: AbortSignal.timeout(15_000),
   });
 
@@ -129,6 +180,23 @@ export async function computeAndStore(userId: string, agentId: string): Promise<
   };
 
   await storeAgentOutput(output);
+
+  // Run inference framework for this agent and persist results.
+  // Failures are non-fatal — the compute result is already stored.
+  try {
+    const inferResp = await fetch(`${config.ML_SERVING_URL}/agents/${agentId}/infer`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({ user_id: userId, feedback_history: feedbackHistory }),
+      signal: AbortSignal.timeout(10_000),
+    });
+    if (inferResp.ok) {
+      const inferResult = await inferResp.json() as { inferred_prefs: Record<string, unknown> };
+      await persistInferredPrefs(userId, agentId, inferResult.inferred_prefs);
+    }
+  } catch {
+    // inference failure is non-fatal
+  }
 }
 
 // ── POST /api/agents/:agentId/compute ─────────────────────────────────────────