feat(memory): add local embedding provider via transformers.js (#95)

Wire EMBEDDING_PROVIDER=local to an in-process embedding adapter so PR-memory works without an external embedding API. LocalEmbeddingAdapter runs a sentence-transformers model (default Xenova/all-MiniLM-L6-v2, 384-dim) through @huggingface/transformers, forcing the WASM backend and a single thread because the Alpine runner image cannot load the native onnxruntime-node binary.

The memory module now selects the embedding adapter from EMBEDDING_PROVIDER instead of hard-wiring Voyage, and EMBEDDING_MODEL becomes optional with a per-adapter default. A migration aligns pr_memory.embedding with EMBEDDING_DIMENSIONS, acting only when the dimension actually changes and truncating the (model-specific, now-incompatible) vectors since the column is NOT NULL. A persistent model-cache volume keeps the downloaded model across restarts.

Author: vineethkrishnan

.env.example

@@ -29,9 +29,17 @@ LLM_API_KEY=
 # Provider: voyage | local
 EMBEDDING_PROVIDER=voyage
 VOYAGE_API_KEY=
-# EMBEDDING_MODEL=voyage-3-lite
-# Must match the embedding model (voyage-3-lite=512, voyage-3=1024).
+# EMBEDDING_MODEL defaults per provider (voyage: voyage-3-lite, local: Xenova/all-MiniLM-L6-v2)
+# EMBEDDING_MODEL=
+# Must match the embedding model (voyage-3-lite=512, voyage-3=1024, all-MiniLM-L6-v2=384).
 # EMBEDDING_DIMENSIONS=512
+#
+# Local embeddings: runs a sentence-transformers model in-process via
+# transformers.js (WASM), no API key. Downloads the model once to the cache dir.
+# EMBEDDING_PROVIDER=local
+# EMBEDDING_MODEL=Xenova/all-MiniLM-L6-v2
+# EMBEDDING_DIMENSIONS=384
+# EMBEDDING_CACHE_DIR=/app/models
 
 # ===========================================================================
 # Database & Redis (required)

docker-compose.yml

@@ -10,6 +10,8 @@ services:
       - NODE_ENV=production
     extra_hosts:
       - 'host.docker.internal:host-gateway'
+    volumes:
+      - modelcache:/app/models
     depends_on:
       db:
         condition: service_healthy
@@ -55,3 +57,4 @@ services:
 volumes:
   pgdata:
   redisdata:
+  modelcache: