Setup Development Environment

The primary way to configure LLM credentials for AG2 is by setting individual environment variables for each provider. This includes variables like OPENAI_API_KEY, AZURE_OPENAI_API_KEY, GEMINI_API_KEY, and so on.

For example, to set up a Gemini API key:

export GEMINI_API_KEY="<your_api_key>"

Similarly, for OpenAI, Azure, Anthropic, etc.:

export OPENAI_API_KEY="<your_openai_api_key>"
export AZURE_OPENAI_API_KEY="<your_azure_api_key>"
export AZURE_OPENAI_API_BASE="<your_azure_base_url>"
export ANTHROPIC_API_KEY="<your_anthropic_api_key>"
# ...and so on for other providers

The AG2 test and tooling system will automatically detect these environment variables and construct the required configuration for all supported LLM providers. This method is recommended and should be your default approach for both local and CI environments.

As an alternative, AG2 also supports an environment variable called OAI_CONFIG_LIST in JSON format to store the LLM keys. OAI_CONFIG_LIST is a list of dictionaries where each dictionary contains the following keys: - model(required): The name of the OpenAI/LLM model. - api_key(optional): The API key for the OpenAI/LLM model. - api_type(optional): The type of the API key. It is used for non-OpenAI LLMs. - api_version(optional): The version of the API key. It is used for Azure API. - base_url(optional): The base URL for the OpenAI/LLM model. - tags(optional): A list of tags for the OpenAI/LLM model which can be used for filtering.

Here is an example of the OAI_CONFIG_LIST in JSON format with two OpenAI models and a Gemini model:

[
    {
        "model": "gpt-4o",
        "api_key": "<your_api_key>",
        "tags": ["gpt-4o", "tool", "vision"]
    },
    {
        "model": "gpt-5-nano",
        "api_key": "<your_api_key>",
        "tags": ["gpt-5-nano", "tool", "vision"]
    },
    {
        "api_type": "google",
        "model": "gemini-pro",
        "api_key": "<your_gemini_api_key>",
    }
]

This OAI_CONFIG_LIST can be set in two ways:

export OAI_CONFIG_LIST='[{"api_type": "openai", "model": "gpt-4o","api_key": "<your_api_key>","tags": ["gpt-4o", "tool", "vision"]},{"api_type": "openai", "model": "gpt-5-nano","api_key": "<your_api_key>","tags": ["gpt-5-nano", "tool", "vision"]},{"api_type": "google", "model": "gemini-pro","api_key": "<your_gemini_api_key>",}]'

export OAI_CONFIG_LIST="/path/to/OAI_CONFIG_LIST.json"

Alternatively, you can set up the LLM keys directly as environment variables. Following is an example of setting up the Gemini api key as an environment variable:

export GEMINI_API_KEY="<your_api_key>"

1. Setup the necessary LLM keys as mentioned above in your terminal.
2. Clone the AG2 repository and cd into the repository.
3. Open the project in Visual Studio Code by running the following command from the root of the repository:

   code .
   

4. Press Ctrl+Shift+P and select Dev Containers: Reopen in Container.
5. Select the desired python environment and wait for the container to build.
6. Once the container is built, you can start developing AG2.

1. Open the AG2 repository on GitHub and fork the repository.
2. Navigate to Settings -> Secrets and variables -> Codespaces.
3. Add the necessary LLM keys as mentioned above by clicking on the New repository secret button.
4. Navigate back to the forked repository.
5. Click on the Code button and select Open with Codespaces.
6. Once the container is built, you can start developing AG2.

1. Setup the necessary LLM keys as mentioned above in your terminal.
2. Fork the AG2 repository and clone the forked repository.
3. Create a virtual environment by running the following command from the root of the repository:

   python3 -m venv venv
   

4. Activate the virtual environment by running the following command:

   source venv/bin/activate
   

5. Install the required dependencies by running the following command:

   pip install -e ".[dev]" && pre-commit install
   

6. Once the dependencies are installed, you can start developing AG2.