Configure an LLM

Get an API key

Get an API key from the Claude Console .

Configure the LLM

1. Add the API key as an environment variable:

   %env ANTHROPIC_API_KEY=your_api_key

2. Add the LLM config to your code as follows:

   import databao
   from databao import LLMConfig
    
   ...
    
   llm_config = LLMConfig(name="claude-sonnet-4-20250514", temperature=0)
   agent = databao.new_agent(llm_config=llm_config)
   ...

Get an API key

Get an API key from the API key page .

Configure the LLM

1. Add the API key as an environment variable:

   %env OPENAI_API_KEY=your_api_key

2. Add the LLM config to your code as follows:

   import databao
   from databao import LLMConfig
   ...
    
   llm_config = LLMConfig(name="gpt-4o-mini", temperature=0)
   agent = databao.new_agent(llm_config=llm_config)
   ...

Install Ollama and download the model

1. Download and install Ollama  on your local machine.

2. (Optional) Download the model:

   ollama pull qwen3:8b

   If you don’t download the model at this step, Databao will download it automatically when you use it.

   We recommend downloading the model before starting to use Databao. Some models can be large in size, often over 10 GB, and downloading them can take some time.

3. Open Ollama.

Configure the LLM

1. Add the LLM config to your code as follows:

   import databao
   from databao import LLMConfig
   ...
    
   llm = LLMConfig(name="ollama:gpt-oss:20b", temperature=0)
   agent = databao.new_agent(llm_config=llm_config)
   ...

Install an LLM server

1. Download and install an LLM server of your choice.

Configure the LLM

1. Create a LLM config file with the following parameters. If needed, replace the model name and modify other parameters.

   qwen3-8b-oai.yaml

   # Match the name used by the OAI server. This example is for LM Studio (when running with `lms server start`):
   name: qwen/qwen3-8b
   # For ollama when running with `ollama serve` use: `name: qwen3:8b`
   # N.B. If using ollama, we recommend using ollama directly as in qwen3-8b-ollama.yaml.
    
   api_base_url: http://localhost:8080/v1
   max_tokens: 32768
   temperature: 0.6
   use_responses_api: false
   timeout: 600

Connect the LLM config in Databao

1. In your code, override the default LLM config with the one you created:

   import databao
   from databao import LLMConfig
   ...
    
   llm_config = LLMConfig(name="qwen3-8b-oai")
   ...