Configuration#
Matrix Hub is configured entirely via environment variables (see .env.example).
| Key | Description | Example |
|---|---|---|
DATABASE_URL |
SQLAlchemy URL | postgresql+psycopg://matrix:matrix@db:5432/matrixhub |
HOST / PORT |
Bind address & port | 0.0.0.0 / 7300 |
API_TOKEN |
Bearer token for admin/protected routes | supersecret |
MATRIX_REMOTES |
CSV/JSON list of index.json URLs to ingest |
https://raw.githubusercontent.com/.../index.json |
INGEST_INTERVAL_MIN |
Background ingestion interval (minutes) | 15 |
SEARCH_LEXICAL_BACKEND |
pgtrgm or none |
pgtrgm |
SEARCH_VECTOR_BACKEND |
pgvector or none |
none |
EMBED_MODEL |
Embedder model id (informational; pluggable) | all-MiniLM-L6-v2 |
MCP_GATEWAY_URL |
MCP Gateway base URL | http://mcpgateway:7200 |
MCP_GATEWAY_TOKEN |
Bearer for gateway admin API | supersecret |
Notes#
- If
API_TOKENis set, passAuthorization: Bearer <token>on admin calls (/catalog/ingest,/catalog/remotes). - For local SQLLite quick trials, you can use
sqlite+pysqlite:///./data/catalog.sqlite. Some search features (pg_trgm/pgvector) won’t be available.