LLM Providers

CommitBee supports three primary providers out of the box. All use streaming for responsive output.

⚠️ Provider Support: Native integrations are currently limited to the providers listed below.
For others not yet natively supported, a proxy is the recommended approach until official support lands.

Secure Key Storage (Recommended)

When working with cloud providers (OpenAI or Anthropic), never store API keys as plain text in config.toml.

The secure-storage feature (enabled by default) stores keys directly in your OS keychain via
platform-native backends (macOS Keychain, Windows Credential Manager, Linux Secret Service):

bash

commitbee config set-key openai      # Prompts for your OpenAI key
commitbee config set-key anthropic   # Prompts for your Anthropic key
commitbee config get-key openai      # Check if a key is stored

Key lookup order: CLI --provider flag → config file → environment variable → keychain.

Keys are held as secrecy::SecretString — memory is zeroed on drop and they appear as [REDACTED] in debug logs.

To opt out and build without keychain support, compile from source:

bash

cargo install commitbee --no-default-features --features all-languages

Ollama (default, local)

The default local setup. Your code never leaves your machine.

toml

provider = "ollama"
model = "qwen3.5:4b"
ollama_host = "http://localhost:11434"

Recommended models:

Model	Size	Notes
`qwen3.5:4b`	3.4 GB	Default — fast, clean JSON output

Thinking mode: Some models (e.g. qwen3.5:4b) support built-in reasoning, producing thinking blocks before their final response.
To enable this, set think = true in your config and raise num_predict to 8192 or higher. The default qwen3.5:4b configuration does not require this.

OpenAI & Compatible Proxies

Ensure your key is stored securely before use:

bash

commitbee config set-key openai

toml

provider = "openai"
model = "gpt-4o-mini"

Or via environment variables:

bash

export COMMITBEE_PROVIDER=openai
export COMMITBEE_MODEL=gpt-4o-mini
export OPENAI_API_KEY=sk-...

OpenAI-Compatible Providers:

Services like Groq, LM Studio, and vLLM work out of the box — just set openai_base_url to the provider’s endpoint.
For broader compatibility, a local proxy such as Bifrost can bridge providers without native OpenAI-compatible APIs.

toml

openai_base_url = "https://localhost:8080/v1"

Anthropic

Ensure your key is stored securely before use:

bash

commitbee config set-key anthropic

toml

provider = "anthropic"
model = "claude-sonnet-4-20250514"

Or via environment variables:

bash

export COMMITBEE_PROVIDER=anthropic
export ANTHROPIC_API_KEY=sk-ant-...