Add Experience V1 schema migration and agent-run completion extractor.

Extend h5_experience with structured fields, wire mindspace-agent-runner and agent-run-gateway to persist task_outcome records with provenance, and add local migration and verification scripts.

Co-authored-by: Cursor <cursoragent@cursor.com>

experience-service.mjs

@@ -1,5 +1,12 @@
 import crypto from 'node:crypto';
 
+import {
+  experienceSearchHaystack,
+  mapExperienceRow,
+  normalizeExperienceRecordInput,
+  serializeExperienceJson,
+} from './experience-schema.mjs';
+
 /**
  * Shared experience store (etat C of the goose scale plan).
  *
@@ -16,58 +23,22 @@ import crypto from 'node:crypto';
  */
 export function createExperienceService(pool, options = {}) {
   const now = options.now ?? (() => Date.now());
-  // Recency half-life: a record this old contributes half its keyword score.
   const halfLifeMs = Math.max(
     60_000,
     Number(options.recencyHalfLifeMs ?? 30 * 24 * 60 * 60 * 1000),
   );
 
-  function normalizeTags(tags) {
-    if (!Array.isArray(tags)) return [];
-    return [...new Set(tags.map((t) => String(t).trim()).filter(Boolean))];
-  }
-
-  function rowToExperience(row) {
-    return {
-      id: row.id,
-      scope: row.scope,
-      kind: row.kind,
-      title: row.title,
-      body: row.body,
-      tags: row.tags_json ? safeJsonArray(row.tags_json) : [],
-      sourceSessionId: row.source_session_id ?? null,
-      sourceUserId: row.source_user_id ?? null,
-      useCount: Number(row.use_count ?? 0),
-      createdAt: Number(row.created_at),
-      updatedAt: Number(row.updated_at),
-    };
-  }
-
-  function safeJsonArray(value) {
-    if (Array.isArray(value)) return value;
-    try {
-      const parsed = JSON.parse(value);
-      return Array.isArray(parsed) ? parsed : [];
-    } catch {
-      return [];
-    }
-  }
-
-  // Keyword overlap × recency decay. Kept separate so the pgvector backend can
-  // replace this with cosine similarity without touching search()'s shape.
   function rankRows(rows, queryTerms, nowMs) {
-    const terms = queryTerms;
     return rows
       .map((row) => {
-        const haystack = `${row.title}\n${row.body}`.toLowerCase();
+        const haystack = experienceSearchHaystack(row);
         let keywordScore = 0;
-        for (const term of terms) {
+        for (const term of queryTerms) {
           if (haystack.includes(term)) keywordScore += 1;
         }
         const ageMs = Math.max(0, nowMs - Number(row.updated_at));
         const recency = Math.pow(0.5, ageMs / halfLifeMs);
-        const score = keywordScore * recency;
-        return { row, score };
+        return { row, score: keywordScore * recency };
       })
       .filter((entry) => entry.score > 0)
       .sort((a, b) => b.score - a.score);
@@ -89,43 +60,66 @@ export function createExperienceService(pool, options = {}) {
    * Persist a piece of experience. Returns the stored record.
    */
   async function record(input) {
-    const title = String(input?.title ?? '').trim();
-    const body = String(input?.body ?? '').trim();
-    if (!title) throw experienceError('经验标题不能为空', 'invalid_experience_input');
-    if (!body) throw experienceError('经验内容不能为空', 'invalid_experience_input');
+    const normalized = normalizeExperienceRecordInput(input);
+    if (!normalized.title) throw experienceError('经验标题不能为空', 'invalid_experience_input');
+    if (!normalized.body) throw experienceError('经验内容不能为空', 'invalid_experience_input');
+
     const id = crypto.randomUUID();
     const ts = now();
-    const scope = String(input?.scope ?? 'global').trim() || 'global';
-    const kind = String(input?.kind ?? 'lesson').trim() || 'lesson';
-    const tags = normalizeTags(input?.tags);
+    const environmentJson = serializeExperienceJson(normalized.environment);
+    const actionJson = serializeExperienceJson(normalized.action);
+    const evidenceJson = serializeExperienceJson(normalized.evidence);
+
     await pool.query(
       `INSERT INTO h5_experience
          (id, scope, kind, title, body, tags_json, source_session_id, source_user_id,
-          use_count, created_at, updated_at)
-       VALUES (?, ?, ?, ?, ?, ?, ?, ?, 0, ?, ?)`,
+          use_count, problem, environment_json, hypothesis, action_json, result, confidence,
+          evidence_json, parent_experience_id, supersedes_id, status, created_at, updated_at)
+       VALUES (?, ?, ?, ?, ?, ?, ?, ?, 0, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
       [
         id,
-        scope,
-        kind,
-        title,
-        body,
-        JSON.stringify(tags),
-        input?.sourceSessionId ?? null,
-        input?.sourceUserId ?? null,
+        normalized.scope,
+        normalized.kind,
+        normalized.title,
+        normalized.body,
+        JSON.stringify(normalized.tags),
+        normalized.sourceSessionId,
+        normalized.sourceUserId,
+        normalized.problem,
+        environmentJson,
+        normalized.hypothesis,
+        actionJson,
+        normalized.result,
+        normalized.confidence,
+        evidenceJson,
+        normalized.parentExperienceId,
+        normalized.supersedesId,
+        normalized.status,
         ts,
         ts,
       ],
     );
-    return rowToExperience({
+
+    return mapExperienceRow({
       id,
-      scope,
-      kind,
-      title,
-      body,
-      tags_json: tags,
-      source_session_id: input?.sourceSessionId ?? null,
-      source_user_id: input?.sourceUserId ?? null,
+      scope: normalized.scope,
+      kind: normalized.kind,
+      title: normalized.title,
+      body: normalized.body,
+      tags_json: JSON.stringify(normalized.tags),
+      source_session_id: normalized.sourceSessionId,
+      source_user_id: normalized.sourceUserId,
       use_count: 0,
+      problem: normalized.problem,
+      environment_json: environmentJson,
+      hypothesis: normalized.hypothesis,
+      action_json: actionJson,
+      result: normalized.result,
+      confidence: normalized.confidence,
+      evidence_json: evidenceJson,
+      parent_experience_id: normalized.parentExperienceId,
+      supersedes_id: normalized.supersedesId,
+      status: normalized.status,
       created_at: ts,
       updated_at: ts,
     });
@@ -138,20 +132,19 @@ export function createExperienceService(pool, options = {}) {
   async function search(query, { scope = 'global', limit = 5 } = {}) {
     const terms = tokenize(query);
     if (terms.length === 0) return [];
-    // Pull a bounded candidate set by scope+recency, then rank in JS. The
-    // candidate cap keeps this cheap on MySQL; pgvector would push ranking into SQL.
     const [rows] = await pool.query(
       `SELECT id, scope, kind, title, body, tags_json, source_session_id,
-              source_user_id, use_count, created_at, updated_at
+              source_user_id, use_count, problem, environment_json, hypothesis,
+              action_json, result, confidence, evidence_json, parent_experience_id,
+              supersedes_id, status, created_at, updated_at
        FROM h5_experience
-       WHERE scope = ?
+       WHERE scope = ? AND status = 'active'
        ORDER BY updated_at DESC
        LIMIT 500`,
       [scope],
     );
     const ranked = rankRows(rows, terms, now()).slice(0, Math.max(1, limit));
     if (ranked.length > 0) {
-      // Best-effort usage bump so popular experience can be surfaced/weighted later.
       const ids = ranked.map((entry) => entry.row.id);
       const placeholders = ids.map(() => '?').join(',');
       await pool
@@ -161,7 +154,7 @@ export function createExperienceService(pool, options = {}) {
         )
         .catch(() => {});
     }
-    return ranked.map((entry) => rowToExperience(entry.row));
+    return ranked.map((entry) => mapExperienceRow(entry.row));
   }
 
   /**
@@ -179,3 +172,5 @@ export function createExperienceService(pool, options = {}) {
 function experienceError(message, code) {
   return Object.assign(new Error(message), { code });
 }
+
+export { formatExperienceInjectionBlock, formatExperienceInjectionLine } from './experience-schema.mjs';