Connect an LLM Provider

Chat-in-Bio supports Anthropic, OpenAI, Google AI, and OpenRouter. You can switch providers at any time.

1. Set the API key

Add the key to your .env:

# Pick one (or set multiple for flexibility)
CHATINBIO_ANTHROPIC_API_KEY=sk-ant-...
CHATINBIO_OPENAI_API_KEY=sk-...
CHATINBIO_GOOGLE_AI_API_KEY=AIza...
CHATINBIO_OPENROUTER_API_KEY=sk-or-...

Restart the backend after changing .env.

2. Update bot config

curl -X PUT http://localhost:8000/api/admin/bot \
  -H "Authorization: Bearer $ADMIN_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_provider": "anthropic",
    "model_name": "claude-sonnet-4-20250514"
  }'

Provider and model reference

Provider value	Models	Notes
`anthropic`	`claude-sonnet-4-20250514`, `claude-haiku-4-5-20251001`	Best for A2UI tool use
`openai`	`gpt-4o`, `gpt-4o-mini`	Good balance of cost/quality
`google-gla`	`gemini-2.0-flash`, `gemini-2.5-pro-preview-06-05`	Cost-effective
`openrouter`	Any model on OpenRouter	Access 100+ models via one API key

Using OpenRouter

OpenRouter gives you access to models from many providers (Llama, Mistral, Qwen, etc.) through a single API key:

# .env
CHATINBIO_OPENROUTER_API_KEY=sk-or-v1-...

# Set bot config to use an OpenRouter model
curl -X PUT http://localhost:8000/api/admin/bot \
  -H "Authorization: Bearer $ADMIN_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_provider": "openrouter",
    "model_name": "meta-llama/llama-3.1-70b-instruct"
  }'

Tool calling support

Chat-in-Bio relies on tool calling for A2UI components. Not all models support tool calling well. Models that work reliably: Claude, GPT-4o, Gemini, Llama 3.1+, Mistral Large. Smaller models may struggle with tool-heavy interactions.

3. Set a locale (optional)

If your audience primarily speaks a specific language, set the bot locale:

curl -X PUT http://localhost:8000/api/admin/bot \
  -H "Authorization: Bearer $ADMIN_KEY" \
  -H "Content-Type: application/json" \
  -d '{"locale": "de"}'

The bot will default to responding in that language. If a visitor writes in a different language, the bot will try to match them.

4. Verify

Send a test message in the chat. If the provider is misconfigured, the agent will return a generic error message. Check the backend logs for details:

uv run litestar run --reload  # logs to stdout

1. Set the API key​

2. Update bot config​

Provider and model reference​

Using OpenRouter​

3. Set a locale (optional)​

4. Verify​