Configuration

Docket is configured through config/defaults.yaml (shipped) and config/config.yaml (your overrides, gitignored).

Priority

1. defaults.yaml — built-in defaults
2. config.yaml — user overrides
3. DOCKET_* environment variables
4. CLI arguments (future)

Environment variables

Any config value can be overridden via DOCKET_{PATH}:

DOCKET_ADAPTERS_LLM_DEFAULT=openai
DOCKET_SERVER_PORT=8080
DOCKET_MEMORY_MODE=rich

You can also use interpolation in YAML:

adapters:
  llm:
    providers:
      kimi:
        config:
          apiKey: "${DOCKET_KIMI_API_KEY:-}"

Key sections

Server

server:
  port: 3000
  host: "127.0.0.1"

Logging

logging:
  level: "info"   # debug | info | warn | error
  format: "pretty" # pretty | json

Memory mode

memory:
  mode: "flat"  # flat | rich

See Memory Modes for the full rich-mode config.

RBAC

memory:
  rbac:
    enabled: false
    authStrategy: "header"   # header | jwt | apiKey
    principalHeader: "X-Principal"
    defaultPolicy: "owner-only"
    apiKeys:
      "dk_abc123": "user:alice"
    policies:
      team:
        readers: ["user:alice", "user:bob"]
        writers: ["user:alice"]

When enabled, every request to the data plane must identify a principal. Memories created without an explicit owner are assigned to the current principal. See RBAC for details.

Adapters

Each adapter category has a default and a map of providers:

adapters:
  llm:
    default: "ollama"
    providers:
      ollama:
        adapter: "@docket/llm-ollama"
        config:
          baseUrl: "http://localhost:11434"
          model: "llama3.2"

Supported categories: llm, embedder, store, blob, queue.

Generate your config visually

Prefer a point-and-click setup? The Config Builder lets you drag adapters onto a canvas, fill in optional settings, and download a complete docket.yaml plus .env.example and setup steps.

Full reference

See config/defaults.yaml in the repo for every option and default value.