Skip to content

CONFIG.md

Latest commit

 

History

History
executable file
·
77 lines (43 loc) · 1.55 KB

CONFIG.md

File metadata and controls

executable file
·
77 lines (43 loc) · 1.55 KB
Raw

Configuration

BrainX V5 is configured through environment variables.

Recommended workflow:

  • keep a local .env for development
  • for production/system services, inject env vars via your process manager (systemd, Railway, Docker, etc.)

Required

DATABASE_URL

Postgres connection string.

Example:

DATABASE_URL=postgresql://brainx:brainx_change_me@127.0.0.1:5432/brainx

Note:

  • existing deployments may still use a legacy physical database name such as brainx
  • that naming drift does not block BrainX V5 itself, but docs and code should not assume a specific DB name unless a migration was actually executed

OPENAI_API_KEY

Used by lib/openai-rag.js to call the OpenAI embeddings endpoint.

Embeddings

OPENAI_EMBEDDING_MODEL

Default: text-embedding-3-small

OPENAI_EMBEDDING_DIMENSIONS

Default: 1536

Must match the schema type:

  • brainx_memories.embedding vector(1536)

If you change dimensions:

  1. change schema
  2. rebuild embeddings for existing rows

Shared env file

BRAINX_ENV

Path to a shared env file.

lib/db.js and lib/openai-rag.js both support loading env from BRAINX_ENV if the main variables are missing.

This is useful when multiple agents share one secrets file.

Inject formatting

BRAINX_INJECT_DEFAULT_TIER

Default: warm_or_hot.

If unset and you don’t pass --tier, the inject command:

  1. searches hot
  2. searches warm
  3. merges results unique by id

BRAINX_INJECT_MAX_CHARS_PER_ITEM

Default: 2000.

BRAINX_INJECT_MAX_LINES_PER_ITEM

Default: 80.