Support custom OpenAI-compatible embedding server with OpenAI fallback

Adds EMBEDDING_SERVER_URL and EMBEDDING_MODEL_NAME env vars, mirroring
the existing LLAMA_SERVER_URL pattern for LLM configuration.

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

.env.example

@@ -19,6 +19,12 @@ BASE_URL=192.168.1.5:8000
 LLAMA_SERVER_URL=http://192.168.1.213:8080/v1
 LLAMA_MODEL_NAME=llama-3.1-8b-instruct
 
+# Embedding Server Configuration
+# If set, uses a custom OpenAI-compatible embedding server (e.g. llama-server)
+# Falls back to OpenAI embeddings if not set
+EMBEDDING_SERVER_URL=http://192.168.1.7:8086/v1
+EMBEDDING_MODEL_NAME=all-minilm
+
 # OpenAI Configuration
 OPENAI_API_KEY=your-openai-api-key
 

blueprints/rag/logic.py

@@ -17,7 +17,17 @@ load_dotenv()
 
 logger = logging.getLogger(__name__)
 
-embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
+_embedding_server_url = os.getenv("EMBEDDING_SERVER_URL")
+_embedding_model = os.getenv("EMBEDDING_MODEL_NAME", "text-embedding-3-small")
+
+if _embedding_server_url:
+    embeddings = OpenAIEmbeddings(
+        model=_embedding_model,
+        base_url=_embedding_server_url,
+        api_key="not-needed",
+    )
+else:
+    embeddings = OpenAIEmbeddings(model=_embedding_model)
 
 # Convert Tortoise-style postgres:// URL to SQLAlchemy-style postgresql+psycopg://
 _db_url = os.getenv(

docker-compose.yml

@@ -31,6 +31,8 @@ services:
       - BASE_URL=${BASE_URL}
       - OLLAMA_URL=${OLLAMA_URL:-http://localhost:11434}
       - OPENAI_API_KEY=${OPENAI_API_KEY}
+      - EMBEDDING_SERVER_URL=${EMBEDDING_SERVER_URL}
+      - EMBEDDING_MODEL_NAME=${EMBEDDING_MODEL_NAME}
       - JWT_SECRET_KEY=${JWT_SECRET_KEY}
       - LLAMA_SERVER_URL=${LLAMA_SERVER_URL}
       - LLAMA_MODEL_NAME=${LLAMA_MODEL_NAME}