54 lines (41 loc) · 2.2 KB

Environment Variables

Source of truth: .env.example and backend/app/core/config.py.

Server

Variable	Default	Description
`SERVER_HOST`	`0.0.0.0`	API bind host
`SERVER_PORT`	`8000`	API bind port
`LOG_LEVEL`	`INFO`	Application log level

Database

Variable	Default	Description
`DB_TYPE`	`postgres`	Database backend
`DATABASE_URL`	`postgresql+asyncpg://...`	SQLAlchemy async connection URL
`POSTGRES_USER`	`postgres`	DB username
`POSTGRES_PASSWORD`	`password`	DB password
`POSTGRES_SERVER`	`postgres`	DB host
`POSTGRES_DB`	`rlm_memory`	DB name

Namespace and Cache

Variable	Default	Description
`DEFAULT_MEMORY_NAMESPACE`	`local`	Namespace used when header is missing
`MAX_NAMESPACE_LENGTH`	`64`	Max namespace size
`SEARCH_CACHE_TTL_SECONDS`	`300`	TTL for `pg_cache` entries

Model Provider

Variable	Default	Description
`OPENAI_BASE_URL`	`http://127.0.0.1:1234/v1`	OpenAI-compatible API base URL (used for LLM and embeddings unless overridden)
`OPENAI_API_KEY`	`lm-studio`	API key value used by client
`OPENAI_MODEL`	`mistralai/ministral-3-14b-reasoning`	Chat/classification model
`EMBED_OPENAI_BASE_URL`	unset	Optional separate embedding endpoint; falls back to `OPENAI_BASE_URL`
`EMBED_OPENAI_API_KEY`	unset	Optional separate embedding API key; falls back to `OPENAI_API_KEY`
`EMBEDDING_MODEL`	`text-embedding-qwen3-embedding-4b@q4_k_m`	Embedding model name
`EMBEDDING_DIMENSIONS`	`1536`	Embedding vector dimensions
`ALLOW_EMBEDDING_FALLBACK`	`true`	When `true`, stores zero-vector if embedding provider is unavailable (response shows `embedding_generated: false`). When `false`, save fails on embedding error.
`LLM_PROVIDER`	`openai-compatible`	Provider identifier (informational)

Frontend

Variable	Default	Description
`VITE_API_URL`	unset	Optional API override; default uses same-origin

Not Used in Current Architecture

The active local stack does not use Auth0, OAuth, or JWT authentication variables.