Skip to content

02-configuration.md

Latest commit

 

History

History
55 lines (39 loc) · 1.68 KB

02-configuration.md

File metadata and controls

55 lines (39 loc) · 1.68 KB

Configuration

Configuration is defined by .env.example and backend/app/core/config.py.

Required Variables

DATABASE_URL=postgresql+asyncpg://postgres:password@postgres:5432/rlm_memory
OPENAI_BASE_URL=http://<your-lm-studio-host>:1234/v1
OPENAI_API_KEY=lm-studio
OPENAI_MODEL=mistralai/ministral-3-14b-reasoning
EMBEDDING_MODEL=text-embedding-qwen3-embedding-4b@q4_k_m
EMBEDDING_DIMENSIONS=1536

Optional dedicated embedding provider:

EMBED_OPENAI_BASE_URL=http://<your-embed-host>:1234/v1
EMBED_OPENAI_API_KEY=lm-studio-embed

If EMBED_OPENAI_BASE_URL or EMBED_OPENAI_API_KEY is unset/empty, embedding requests fall back to OPENAI_BASE_URL and OPENAI_API_KEY.

Namespace and Cache

DEFAULT_MEMORY_NAMESPACE=local
MAX_NAMESPACE_LENGTH=64
SEARCH_CACHE_TTL_SECONDS=300
ALLOW_EMBEDDING_FALLBACK=true

About Namespaces

A namespace scopes all memory operations so that different projects, agents, or users have independent memory stores. Memories saved in namespace project-a are invisible to namespace project-b.

  • Set the namespace per request via the X-Memory-Namespace header (REST) or the namespace MCP tool argument.
  • If omitted, the server uses DEFAULT_MEMORY_NAMESPACE (default: local).
  • Namespaces are created automatically on first use.
  • Namespaces are case-insensitive and normalized to lowercase.
  • Max length is controlled by MAX_NAMESPACE_LENGTH (default: 64 characters).

Search results are cached per namespace in the PostgreSQL pg_cache table with TTL controlled by SEARCH_CACHE_TTL_SECONDS.

Network Defaults

SERVER_HOST=0.0.0.0
SERVER_PORT=8000

Important

  • This build does not use OAuth, JWT, or Auth0.