Fay is an interactive CLI tool that helps users with software engineering tasks.

• Interactive command line interface - Uses an AI agent to assist with tasks
• Agent tools to help along the way

The project is built using the deno runtime. You can compile the binary, you can run the following command

deno task compile:cli

You will then get bin/fay or bin/fay.exe on windows

Fay uses the Gemini family of models from Google. To use these models, you will need to set the GOOGLE_GENERATIVE_AI_API_KEY environment variable.

export GOOGLE_GENERATIVE_AI_API_KEY=<your_api_key>

export ANTHROPIC_API_KEY=<your_api_key>

To run the interactive agent

./bin/fay

Fay can be configured on a global, and on a per-project bases. The configuration is a JSON file that can be in the following locations, in order of precedence:

• FAY_CONFIG environment variable
• .git/fay/fay.json - .fay.json
• APPDATA/fay/fay.json (windows only)
• HOME/.config/fay/fay.json
• HOME/.fay.json

The name of the model to use for the agent. A list of models can be found in agent/provider.ts.

{
  "model": "gemini-1.5-flash-latest"
}

Context files for the agent. These are files that provide additional context to the agent, they are usually project specific. The default value is ["AGENTS.md"], this will find it in the root of the agent context.

{
  "contextFiles": [
    "AGENTS.md"
  ]
}

Before starting Fay, you will need to create a new session.

fay new --title "My session title"

All of your sessions can be listed like.

fay list

You can run fay and take the initial prompt from a file. This will be as if you typed this into the first prompt. If the agent requires input after, you will get the interactive prompt.

fay run --prompt-file ./prompt.md

Slash commands can be used to perform actions that don't need the agent input. They can be typed into the interactive input.