Skip to main content

Recommended Config

The simplest and recommended configuration uses Google's Gemini 2.5 Pro model, which provides a strong balance of reasoning, planning, and analysis capabilities for most tasks.

  • Primary Model: Google Gemini 2.5 Pro (gemini-2.5-pro-preview-03-25) handles core agent functions and expert consultations by default when only the GEMINI_API_KEY is set.
  • Web Search: Tavily web search integration (optional) allows the agent to find relevant information online.
info

RA.Aid must be installed before using these configurations. If you haven't installed it yet, please see the Installation Guide.

Getting API Keys​

To use RA.Aid with the recommended configuration, you'll need to obtain the following API key:

  1. Gemini API Key: Obtain a free API key from Google AI Studio.

⚠️ Note: You should set up a billing account associated with your generate key to get Tier 1. By default, you'll get the Free Tier plan and probably a get a Rate Limit Error.

You might also consider keys for other services if you want to explicitly configure different models or providers:

  1. OpenAI API Key (Optional): Create an account at OpenAI's platform and generate an API key. Useful if you want to use OpenAI models, perhaps for the expert model specifically (see Expert Model Configuration).
  2. Anthropic API Key (Optional): Sign up at Anthropic's Console and generate an API key. Useful if you prefer Anthropic models.
  3. Tavily API Key (Optional): Create an account at Tavily and get your API key for web search capabilities.

Please keep your API keys secure and never share them publicly. Each service has its own pricing and usage terms.

Configuration​

Configure your API keys as environment variables. For the recommended setup, you only need:

# Recommended: For Google Gemini
export GEMINI_API_KEY='your_gemini_api_key'

# Optional: For web search capability
export TAVILY_API_KEY='your_tavily_api_key'

If you want to use other providers, you can set their keys as well:

# Optional: For OpenAI
export OPENAI_API_KEY='your_openai_api_key'

# Optional: For Anthropic
export ANTHROPIC_API_KEY='your_anthropic_api_key'

# Optional: To use a specific expert model (see Expert Model Config)
# export EXPERT_OPENAI_API_KEY='your_openai_api_key_for_expert'
# export EXPERT_GEMINI_API_KEY='your_gemini_api_key_for_expert'
# etc.

RA.Aid will automatically detect the available keys and select the default provider. If only GEMINI_API_KEY is set, it will use the gemini provider and the gemini-2.5-pro-preview-03-25 model.

Basic Usage​

Start RA.Aid in interactive chat mode:

ra-aid --chat

Or run with a single command:

ra-aid -m "Help me understand this code"

If you prefer to use aider's specialized code editing capabilities instead of RA.Aid's built-in file modification tools:

ra-aid -m "Implement this feature" --use-aider

Note: aider must be installed separately. See aider-chat for more information.

You can control logging verbosity and location using the --log-mode and --log-level options:

# Log to file (with only warnings to console)
ra-aid -m "Your task" --log-mode file --log-level debug

# Log everything to console
ra-aid -m "Your task" --log-mode console --log-level info

For more detailed logging configuration, see the Logging documentation.

For information on RA.Aid's memory management and how to reset memory when needed, see the Memory Management documentation.

For advanced reasoning capabilities, see the Expert Model Configuration to configure a specialized expert model for complex tasks.