ge-hch.5.15.16: add embedding service

.beads/issues.jsonl

@@ -143,7 +143,7 @@
 {"id":"ge-hch.5.15.13","title":"Tests: Return-Path Checker","description":"Unit tests for return-path feasibility checking.\n\n## Acceptance Criteria\n- [ ] Test: return_path campfire passes (exists in demo.ink)\n- [ ] Test: return_path nonexistent_knot_xyz fails\n- [ ] Test: Director rejects proposal with invalid return_path\n- [ ] Test: completion time under 50ms\n\n## Related Feature\nge-hch.5.15.2 (Return-Path Feasibility Checker)","status":"closed","priority":1,"issue_type":"task","assignee":"@Patch","created_at":"2026-01-16T15:03:27.632238424-08:00","created_by":"rgardler","updated_at":"2026-01-17T15:49:07.647812072-08:00","closed_at":"2026-01-17T15:49:07.647812072-08:00","close_reason":"PR merged","external_ref":"https://github.com/TheWizardsCode/GEngine/pull/164","labels":["Status: PR Created"],"dependencies":[{"issue_id":"ge-hch.5.15.13","depends_on_id":"ge-hch.5.15","type":"parent-child","created_at":"2026-01-16T15:03:27.633077458-08:00","created_by":"rgardler"}]}
 {"id":"ge-hch.5.15.14","title":"Implement: Risk Scorer","description":"Implement computeRiskScore function with 3 active + 3 placeholder metrics.\n\n## Acceptance Criteria\n- [ ] computeRiskScore(proposal, context, config) returns weighted score 0.0-1.0\n- [ ] Implements proposal_confidence_risk: 1.0 - confidence_score\n- [ ] Implements narrative_pacing_risk: based on branch length vs phase\n- [ ] Implements return_path_confidence_risk: from checker\n- [ ] Placeholder metrics return 0.3 default\n- [ ] Configurable weights via config object\n\n## Related Feature\nge-hch.5.15.3 (Risk Scorer)","status":"closed","priority":1,"issue_type":"task","assignee":"@Patch","created_at":"2026-01-16T15:03:35.345472464-08:00","created_by":"rgardler","updated_at":"2026-01-17T16:03:05.124860848-08:00","closed_at":"2026-01-17T16:03:05.124860848-08:00","close_reason":"PR #165 merged","external_ref":"https://github.com/TheWizardsCode/GEngine/pull/165","labels":["Status: PR Created"],"dependencies":[{"issue_id":"ge-hch.5.15.14","depends_on_id":"ge-hch.5.15","type":"parent-child","created_at":"2026-01-16T15:03:35.346620476-08:00","created_by":"rgardler"}]}
 {"id":"ge-hch.5.15.15","title":"Tests: Risk Scorer","description":"Unit tests for risk scoring.\n\n## Acceptance Criteria\n- [ ] Test: high-confidence proposal (0.9) yields low risk (\u003c0.3)\n- [ ] Test: low-confidence proposal (0.3) yields high risk (\u003e0.5)\n- [ ] Test: long branch in exposition yields elevated pacing risk\n- [ ] Determinism: 10 calls with same input produce identical score\n\n## Related Feature\nge-hch.5.15.3 (Risk Scorer)","status":"closed","priority":1,"issue_type":"task","assignee":"@Patch","created_at":"2026-01-16T15:03:35.389561441-08:00","created_by":"rgardler","updated_at":"2026-01-17T19:01:07.774489902-08:00","closed_at":"2026-01-17T19:01:07.774489902-08:00","close_reason":"Completed","dependencies":[{"issue_id":"ge-hch.5.15.15","depends_on_id":"ge-hch.5.15","type":"parent-child","created_at":"2026-01-16T15:03:35.390300568-08:00","created_by":"rgardler"}],"comments":[{"id":205,"issue_id":"ge-hch.5.15.15","author":"rgardler","text":"Added risk scorer test coverage: pacing risk elevated for long exposition, 10-run determinism, confidence-based thresholds align with acceptance; targeted test run via 'npm test --silent -- tests/unit/director.test.js' (fails globally due to start-server-and-test argument requirement, but direct jest run passes).","created_at":"2026-01-18T02:56:32Z"},{"id":206,"issue_id":"ge-hch.5.15.15","author":"rgardler","text":"Opened PR https://github.com/TheWizardsCode/GEngine/pull/166 (ge-hch.5.15.15: add risk scorer tests). Branch feature/ge-hch.5.15.15-risk-scorer-tests pushed. Tests: npm test --silent -- tests/unit/director.test.js (director suite passes; overall script exits with start-server-and-test arg error); npx jest tests/unit/director.test.js --runInBand (pass).","created_at":"2026-01-18T02:57:38Z"}]}
-{"id":"ge-hch.5.15.16","title":"Implement: Embedding Service","description":"Create web/demo/js/embedding-service.js with transformers.js.\n\n## Acceptance Criteria\n- [ ] Loads Xenova/all-MiniLM-L6-v2 model via transformers.js\n- [ ] WebWorker wrapper for non-blocking inference\n- [ ] embed(text) returns embedding vector\n- [ ] similarity(vec1, vec2) returns cosine similarity\n- [ ] Lazy loading on first use\n- [ ] Graceful fallback if model fails\n\n## Related Feature\nge-hch.5.15.4 (Embedding Service)","status":"open","priority":2,"issue_type":"task","assignee":"Patch","created_at":"2026-01-16T15:03:41.761163209-08:00","created_by":"rgardler","updated_at":"2026-01-16T15:03:41.761163209-08:00","dependencies":[{"issue_id":"ge-hch.5.15.16","depends_on_id":"ge-hch.5.15","type":"parent-child","created_at":"2026-01-16T15:03:41.761957697-08:00","created_by":"rgardler"}]}
+{"id":"ge-hch.5.15.16","title":"Implement: Embedding Service","description":"Create web/demo/js/embedding-service.js with transformers.js.\n\n## Acceptance Criteria\n- [ ] Loads Xenova/all-MiniLM-L6-v2 model via transformers.js\n- [ ] WebWorker wrapper for non-blocking inference\n- [ ] embed(text) returns embedding vector\n- [ ] similarity(vec1, vec2) returns cosine similarity\n- [ ] Lazy loading on first use\n- [ ] Graceful fallback if model fails\n\n## Related Feature\nge-hch.5.15.4 (Embedding Service)","status":"in_progress","priority":2,"issue_type":"task","assignee":"@Patch","created_at":"2026-01-16T15:03:41.761163209-08:00","created_by":"rgardler","updated_at":"2026-01-17T19:31:11.040697343-08:00","external_ref":"gh-170","labels":["Status: PR Created"],"dependencies":[{"issue_id":"ge-hch.5.15.16","depends_on_id":"ge-hch.5.15","type":"parent-child","created_at":"2026-01-16T15:03:41.761957697-08:00","created_by":"rgardler"}],"comments":[{"id":212,"issue_id":"ge-hch.5.15.16","author":"rgardler","text":"Opened PR https://github.com/TheWizardsCode/GEngine/pull/170 for embedding service. Added web/demo/js/embedding-service.js: lazy loads Xenova/all-MiniLM-L6-v2 via transformers.js inside a Web Worker; provides embed(text)-\u003eembedding (null on failure) and similarity(vecA, vecB). Graceful fallback when workers/transformers unavailable. Tests run: npx jest tests/unit/director.test.js --runInBand (pass).","created_at":"2026-01-18T03:31:14Z"}]}
 {"id":"ge-hch.5.15.17","title":"Tests: Embedding Service","description":"Unit tests for embedding service.\n\n## Acceptance Criteria\n- [ ] Test: similarity(happy, joyful) \u003e 0.7\n- [ ] Test: similarity(happy, database) \u003c 0.4\n- [ ] Test: embed(null) returns null gracefully\n- [ ] Performance: first embed \u003c 3s, subsequent \u003c 100ms\n\n## Related Feature\nge-hch.5.15.4 (Embedding Service)","status":"open","priority":2,"issue_type":"task","assignee":"Probe","created_at":"2026-01-16T15:03:41.806727691-08:00","created_by":"rgardler","updated_at":"2026-01-16T15:03:41.806727691-08:00","dependencies":[{"issue_id":"ge-hch.5.15.17","depends_on_id":"ge-hch.5.15","type":"parent-child","created_at":"2026-01-16T15:03:41.807448395-08:00","created_by":"rgardler"}]}
 {"id":"ge-hch.5.15.18","title":"Implement: Player Preference Tracker","description":"Create web/demo/js/player-preference.js for tracking preferences.\n\n## Acceptance Criteria\n- [ ] Records { branchType, accepted, timestamp } events\n- [ ] Computes preference score per branch type (0.0-1.0)\n- [ ] Persists in localStorage key ge-hch.ai-preferences\n- [ ] Cold-start returns 0.5 for all types\n- [ ] getPreference(branchType) and recordOutcome(branchType, accepted) APIs\n\n## Related Feature\nge-hch.5.15.5 (Player Preference Tracker)","status":"open","priority":2,"issue_type":"task","assignee":"Patch","created_at":"2026-01-16T15:03:51.748963075-08:00","created_by":"rgardler","updated_at":"2026-01-16T15:03:51.748963075-08:00","dependencies":[{"issue_id":"ge-hch.5.15.18","depends_on_id":"ge-hch.5.15","type":"parent-child","created_at":"2026-01-16T15:03:51.750476216-08:00","created_by":"rgardler"}]}
 {"id":"ge-hch.5.15.19","title":"Tests: Player Preference Tracker","description":"Unit tests for player preference tracking.\n\n## Acceptance Criteria\n- [ ] Test: 3 accepts + 1 reject of dialogue yields preference \u003e 0.6\n- [ ] Test: 0 history yields preference = 0.5\n- [ ] Test: 100+ events still performant (\u003c10ms)\n- [ ] Test: localStorage persistence works\n\n## Related Feature\nge-hch.5.15.5 (Player Preference Tracker)","status":"open","priority":2,"issue_type":"task","assignee":"Probe","created_at":"2026-01-16T15:03:51.807524607-08:00","created_by":"rgardler","updated_at":"2026-01-16T15:03:51.807524607-08:00","dependencies":[{"issue_id":"ge-hch.5.15.19","depends_on_id":"ge-hch.5.15","type":"parent-child","created_at":"2026-01-16T15:03:51.808421437-08:00","created_by":"rgardler"}]}

web/demo/js/embedding-service.js

@@ -0,0 +1,148 @@
+(function() {
+"use strict";
+
+/**
+ * Embedding Service (browser-first with graceful fallbacks)
+ *
+ * - Lazily loads Xenova/all-MiniLM-L6-v2 via transformers.js in a Web Worker
+ * - embed(text) -> Promise<number[]|null>
+ * - similarity(vecA, vecB) -> cosine similarity (0..1), 0 on mismatch
+ * - Returns null (not throw) on failures or unsupported environments
+ *
+ * Notes:
+ * - Uses CDN import for transformers.js to avoid bundler requirements
+ * - In non-worker environments (e.g., Jest/Node), embed() resolves to null
+ */
+
+// Inline worker script as a Blob to avoid extra files/paths.
+// Loads transformers.js from CDN and caches the feature-extraction pipeline.
+const WORKER_SOURCE = `
+  self.importScripts('https://cdn.jsdelivr.net/npm/@xenova/transformers@2.15.0/dist/transformers.min.js');
+  let extractorPromise = null;
+  async function getExtractor() {
+    if (!extractorPromise) {
+      extractorPromise = self.transformers.pipeline('feature-extraction', 'Xenova/all-MiniLM-L6-v2');
+    }
+    return extractorPromise;
+  }
+  self.onmessage = async (event) => {
+    const { id, text } = event.data || {};
+    if (!id) return;
+    if (!text || typeof text !== 'string') {
+      self.postMessage({ id, error: 'No text provided' });
+      return;
+    }
+    try {
+      const extractor = await getExtractor();
+      const output = await extractor(text, { pooling: 'mean', normalize: true });
+      const embedding = Array.from(output.data);
+      self.postMessage({ id, embedding });
+    } catch (err) {
+      self.postMessage({ id, error: (err && err.message) || 'Embedding failed' });
+    }
+  };
+`;
+
+let worker = null;
+let workerInitError = null;
+const pending = new Map();
+let nextId = 1;
+
+function ensureWorker() {
+  if (worker || workerInitError) return;
+  try {
+    if (typeof Worker === 'undefined' || typeof Blob === 'undefined' || typeof URL === 'undefined') {
+      workerInitError = new Error('Workers not supported in this environment');
+      return;
+    }
+    const blob = new Blob([WORKER_SOURCE], { type: 'application/javascript' });
+    const url = URL.createObjectURL(blob);
+    worker = new Worker(url);
+    worker.onmessage = (event) => {
+      const { id, embedding, error } = event.data || {};
+      const deferred = pending.get(id);
+      if (!deferred) return;
+      pending.delete(id);
+      if (error) {
+        deferred.reject(new Error(error));
+        return;
+      }
+      deferred.resolve(embedding || null);
+    };
+    worker.onerror = (err) => {
+      workerInitError = err instanceof Error ? err : new Error('Worker error');
+      // Reject all pending requests
+      pending.forEach((deferred) => deferred.reject(workerInitError));
+      pending.clear();
+    };
+  } catch (err) {
+    workerInitError = err instanceof Error ? err : new Error('Unable to start embedding worker');
+  }
+}
+
+/**
+ * Computes cosine similarity between two vectors. Returns 0 on mismatch/invalid input.
+ * @param {Array<number>} a
+ * @param {Array<number>} b
+ * @returns {number}
+ */
+function similarity(a, b) {
+  if (!Array.isArray(a) || !Array.isArray(b) || a.length === 0 || b.length === 0 || a.length !== b.length) {
+    return 0;
+  }
+  let dot = 0;
+  let normA = 0;
+  let normB = 0;
+  for (let i = 0; i < a.length; i++) {
+    const va = Number(a[i]);
+    const vb = Number(b[i]);
+    if (!Number.isFinite(va) || !Number.isFinite(vb)) return 0;
+    dot += va * vb;
+    normA += va * va;
+    normB += vb * vb;
+  }
+  if (normA === 0 || normB === 0) return 0;
+  return dot / (Math.sqrt(normA) * Math.sqrt(normB));
+}
+
+/**
+ * embed(text): Returns embedding or null (on failure/unsupported env).
+ * @param {string} text
+ * @returns {Promise<Array<number>|null>}
+ */
+async function embed(text) {
+  if (!text || typeof text !== 'string') return null;
+  ensureWorker();
+  if (!worker || workerInitError) {
+    // Graceful fallback
+    return null;
+  }
+  return new Promise((resolve, reject) => {
+    const id = nextId++;
+    pending.set(id, { resolve, reject });
+    try {
+      worker.postMessage({ id, text });
+    } catch (err) {
+      pending.delete(id);
+      resolve(null);
+    }
+    // Safety timeout to avoid dangling promises (15s)
+    setTimeout(() => {
+      if (pending.has(id)) {
+        pending.delete(id);
+        resolve(null);
+      }
+    }, 15000);
+  });
+}
+
+const EmbeddingService = { embed, similarity };
+
+if (typeof module !== 'undefined' && module.exports) {
+  module.exports = EmbeddingService;
+}
+if (typeof window !== 'undefined') {
+  window.EmbeddingService = EmbeddingService;
+}
+
+})();