feat(P4-002): semantic search — pgvector embeddings + search API

Add EmbeddingService using OpenAI-compatible embeddings API (supports
text-embedding-3-small, configurable via EMBEDDING_MODEL and
EMBEDDING_API_URL env vars). Wire embedding generation into insight
creation and semantic search endpoint. POST /api/memory/search now
generates a query embedding and performs cosine distance search via
pgvector when OPENAI_API_KEY is configured.

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

apps/gateway/src/memory/embedding.service.ts

@@ -0,0 +1,69 @@
+import { Injectable, Logger } from '@nestjs/common';
+import type { EmbeddingProvider } from '@mosaic/memory';
+
+const DEFAULT_MODEL = 'text-embedding-3-small';
+const DEFAULT_DIMENSIONS = 1536;
+
+interface EmbeddingResponse {
+  data: Array<{ embedding: number[]; index: number }>;
+  model: string;
+  usage: { prompt_tokens: number; total_tokens: number };
+}
+
+/**
+ * Generates embeddings via the OpenAI-compatible embeddings API.
+ * Supports OpenAI, Azure OpenAI, and any provider with a compatible endpoint.
+ */
+@Injectable()
+export class EmbeddingService implements EmbeddingProvider {
+  private readonly logger = new Logger(EmbeddingService.name);
+  private readonly apiKey: string | undefined;
+  private readonly baseUrl: string;
+  private readonly model: string;
+
+  readonly dimensions = DEFAULT_DIMENSIONS;
+
+  constructor() {
+    this.apiKey = process.env['OPENAI_API_KEY'];
+    this.baseUrl = process.env['EMBEDDING_API_URL'] ?? 'https://api.openai.com/v1';
+    this.model = process.env['EMBEDDING_MODEL'] ?? DEFAULT_MODEL;
+  }
+
+  get available(): boolean {
+    return !!this.apiKey;
+  }
+
+  async embed(text: string): Promise<number[]> {
+    const results = await this.embedBatch([text]);
+    return results[0]!;
+  }
+
+  async embedBatch(texts: string[]): Promise<number[][]> {
+    if (!this.apiKey) {
+      this.logger.warn('No OPENAI_API_KEY configured — returning zero vectors');
+      return texts.map(() => new Array<number>(this.dimensions).fill(0));
+    }
+
+    const response = await fetch(`${this.baseUrl}/embeddings`, {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+        Authorization: `Bearer ${this.apiKey}`,
+      },
+      body: JSON.stringify({
+        model: this.model,
+        input: texts,
+        dimensions: this.dimensions,
+      }),
+    });
+
+    if (!response.ok) {
+      const body = await response.text();
+      this.logger.error(`Embedding API error: ${response.status} ${body}`);
+      throw new Error(`Embedding API returned ${response.status}`);
+    }
+
+    const json = (await response.json()) as EmbeddingResponse;
+    return json.data.sort((a, b) => a.index - b.index).map((d) => d.embedding);
+  }
+}

apps/gateway/src/memory/memory.controller.ts

@@ -15,12 +15,16 @@ import {
 import type { Memory } from '@mosaic/memory';
 import { MEMORY } from './memory.tokens.js';
 import { AuthGuard } from '../auth/auth.guard.js';
+import { EmbeddingService } from './embedding.service.js';
 import type { UpsertPreferenceDto, CreateInsightDto, SearchMemoryDto } from './memory.dto.js';
 
 @Controller('api/memory')
 @UseGuards(AuthGuard)
 export class MemoryController {
-  constructor(@Inject(MEMORY) private readonly memory: Memory) {}
+  constructor(
+    @Inject(MEMORY) private readonly memory: Memory,
+    private readonly embeddings: EmbeddingService,
+  ) {}
 
   // ─── Preferences ────────────────────────────────────────────────────
 
@@ -76,12 +80,17 @@ export class MemoryController {
 
   @Post('insights')
   async createInsight(@Query('userId') userId: string, @Body() dto: CreateInsightDto) {
+    const embedding = this.embeddings.available
+      ? await this.embeddings.embed(dto.content)
+      : undefined;
+
     return this.memory.insights.create({
       userId,
       content: dto.content,
       source: dto.source,
       category: dto.category,
       metadata: dto.metadata,
+      embedding: embedding ?? null,
     });
   }
 
@@ -96,13 +105,22 @@ export class MemoryController {
 
   @Post('search')
   async searchMemory(@Query('userId') userId: string, @Body() dto: SearchMemoryDto) {
-    // Search requires an embedding provider to be configured.
-    // For now, return empty results if no embedding is available.
-    // P4-002 will implement the full embedding + search pipeline.
-    return {
-      query: dto.query,
-      results: [],
-      message: 'Semantic search requires embedding provider (P4-002)',
-    };
+    if (!this.embeddings.available) {
+      return {
+        query: dto.query,
+        results: [],
+        message: 'Semantic search requires OPENAI_API_KEY for embeddings',
+      };
+    }
+
+    const queryEmbedding = await this.embeddings.embed(dto.query);
+    const results = await this.memory.insights.searchByEmbedding(
+      userId,
+      queryEmbedding,
+      dto.limit ?? 10,
+      dto.maxDistance ?? 0.8,
+    );
+
+    return { query: dto.query, results };
   }
 }

apps/gateway/src/memory/memory.module.ts

@@ -4,6 +4,7 @@ import type { Db } from '@mosaic/db';
 import { DB } from '../database/database.module.js';
 import { MEMORY } from './memory.tokens.js';
 import { MemoryController } from './memory.controller.js';
+import { EmbeddingService } from './embedding.service.js';
 
 @Global()
 @Module({
@@ -13,8 +14,9 @@ import { MemoryController } from './memory.controller.js';
       useFactory: (db: Db): Memory => createMemory(db),
       inject: [DB],
     },
+    EmbeddingService,
   ],
   controllers: [MemoryController],
-  exports: [MEMORY],
+  exports: [MEMORY, EmbeddingService],
 })
 export class MemoryModule {}

docs/TASKS.md

@@ -37,8 +37,8 @@
 | P3-006 | done        | Phase 3   | Settings — provider config, profile, integrations             | #88 | #31   |
 | P3-007 | done        | Phase 3   | Admin panel — user management, RBAC                           | #89 | #32   |
 | P3-008 | done        | Phase 3   | Verify Phase 3 — web dashboard functional E2E                 | —   | #33   |
-| P4-001 | not-started | Phase 4   | @mosaic/memory — preference + insight stores                  | —   | #34   |
-| P4-002 | not-started | Phase 4   | Semantic search — pgvector embeddings + search API            | —   | #35   |
+| P4-001 | in-progress | Phase 4   | @mosaic/memory — preference + insight stores                  | —   | #34   |
+| P4-002 | in-progress | Phase 4   | Semantic search — pgvector embeddings + search API            | —   | #35   |
 | P4-003 | not-started | Phase 4   | @mosaic/log — log ingest, parsing, tiered storage             | —   | #36   |
 | P4-004 | not-started | Phase 4   | Summarization pipeline — Haiku-tier LLM + cron                | —   | #37   |
 | P4-005 | not-started | Phase 4   | Memory integration — inject into agent sessions               | —   | #38   |