Configuration

DevRecall reads ~/.devrecall/config.json. You can edit it directly, or use:

• devrecall config init — write a default config
• devrecall config show — print the current config

Tokens and API keys are not in this file — they live in ~/.devrecall/tokens/ (file-mode 0600) or the OS keychain.

Top-level shape

{
  "git": { ... },
  "slack": { ... },
  "calendar": { ... },
  "github": { ... },
  "gitlab": { ... },
  "bitbucket": { ... },
  "jira": { ... },
  "confluence": { ... },
  "linear": { ... },
  "llm": { ... },
  "embedding": { ... },
  "privacy": { ... },
  "chat": { ... },
  "server": { ... },
  "token_storage": "file"
}

Sources

git

{
  "enabled": true,
  "scan_paths": ["~/Projects", "~/work"],
  "repos": ["~/work/backend-api"],
  "emails": ["pavel@company.com", "pavel.piliak@gmail.com"]
}

Field	Notes
scan_paths	Walked recursively for .git directories
repos	Explicit repo paths (skip auto-discovery for these)
emails	Author emails that count as “you” (auto-detected if empty)

slack

{ "enabled": true, "team_id": "T0123", "team_name": "ACME" }

Connect via devrecall auth slack. To connect multiple workspaces, run the command once per workspace.

calendar

{ "enabled": true, "email": "pavel@company.com" }

Connect via devrecall auth google.

github

{ "enabled": true, "username": "pavelpilyak", "auth_mode": "oauth" }

auth_mode is one of oauth, pat, gh-cli.

gitlab / bitbucket

{
  "enabled": true,
  "base_url": "https://gitlab.example.com",
  "username": "pavel"
}

base_url is optional for cloud (defaults to gitlab.com / bitbucket.org).

jira / confluence / linear

{
  "enabled": true,
  "base_url": "https://acme.atlassian.net",
  "auth_mode": "oauth",
  "email": "pavel@company.com"
}

auth_mode is oauth or api-token (api-key for Linear). email is required for token-based Jira auth.

LLM

{
  "llm": {
    "provider": "ollama",
    "model": "gemma4",
    "base_url": "http://localhost:11434"
  }
}

Field	Values
provider	ollama, openai, anthropic
model	Provider-specific model name
base_url	Optional — for OpenAI-compatible endpoints (Groq, vLLM, etc.)

Per-task model routing and fallback chains are described in LLM strategy.

Embeddings

{
  "embedding": {
    "provider": "onnx"
  }
}

Default is onnx — bundled all-MiniLM-L6-v2 (384 dimensions), zero external dependencies. Override to ollama (all-minilm) or openai (text-embedding-3-small) if you want a different model.

Field	Values
provider	onnx (default), ollama, openai
model	Provider-specific model name; sensible default per provider
base_url	Optional — override Ollama / OpenAI-compatible endpoint
dimensions	Vector size; 0 = model default

Privacy

{
  "privacy": {
    "slack": "summary",
    "linear": "full",
    "jira": "metadata"
  }
}

Per-source storage mode. Values: full, summary, metadata. See Privacy model.

Chat

{
  "chat": {
    "sync_freshness": {
      "default_ttl": "3h",
      "per_source": { "slack": "1h", "calendar": "30m" },
      "wait": "10s"
    }
  }
}

Controls how chat refreshes data before answering. sync_freshness.disabled: true skips the pre-chat sync entirely.

Server

{ "server": { "port": 3725 } }

Local HTTP API port. Default 3725 (“DRCL” on a phone keypad). Bound to 127.0.0.1 only.

Token storage

{ "token_storage": "file" }

file (default) writes tokens to ~/.devrecall/tokens/ with 0600 permissions. keychain stores them in the OS keychain (macOS Keychain on macOS, Secret Service on Linux).

Paths

Purpose	Path
Config	~/.devrecall/config.json
Database	~/.devrecall/devrecall.db
Tokens (file mode)	~/.devrecall/tokens/<source>.json
Prompt templates	~/.devrecall/prompts/*.tmpl
Update cache	~/.devrecall/version_check.json