# memory-server-rs stack — copy to .env and adjust.
# Note: all values below are also the compose defaults, so a .env is optional.

# neo4j credentials, <user>/<password> (used by the neo4j image and the server)
NEO4J_AUTH=neo4j/changeme-in-dotenv
NEO4J_PASSWORD=changeme-in-dotenv

# set to a secret to require bearer auth from MCP clients (recommended if you
# publish the port beyond localhost)
MEMORY_TOKEN=

# qdrant api key (optional; qdrant only enforces it when set)
QDRANT_API_KEY=

# host address and port for the MCP server (set HOST=0.0.0.0 for remote access)
HOST=127.0.0.1
PORT=8080

# extra HTTP Host names accepted at /mcp (comma-separated; localhost remains allowed)
#ALLOWED_HOSTS=memory.example.com,192.168.1.50:8080

# embeddings: unset = local bge-small-en-v1.5 (no memory content leaves the stack);
# set baseurl + model to use an external OpenAI-compatible /embeddings endpoint
#EMBEDDINGS_BASEURL=http://my-llm-box:8888/v1
#EMBEDDINGS_APIKEY=sk-...
#EMBEDDINGS_MODEL=emb-1

# override qdrant collection dimensionality (defaults to 384 for bge-small-en-v1.5)
#EMBEDDING_DIMENSIONS=384