Skip to content

A coding assistant implemented in the terminal using the Gemini AI agent

Notifications You must be signed in to change notification settings

nathanieldorn/geminiagent

Repository files navigation

Gemini AI Agent

An AI agent built in Python with Gemini 1.12.1.

Restrictions

The agent is limited to files and folders within the working directory.

Functions

The agent may call upon the functions specified below based upon the prompts given.

Get Files Information

The agent can retrieve a file's size and directory. The same can be returned for all files in a directory.

Get File Content

The agent can return the contents of a file as a string. However, this is limited to the first 10,000 characters.

Write File

The agent can write new or edit existing files when the prompt includes the content to be written or directs the type of change to be made, e.g, Fix the bug in main.py causing the loop to break too early.

Run Python File

The agent can run files and will return the text results of the run file's output to stdout. There is a 30 second timeout. Errors will be returned to stderr.

Instructions

To run the project, use uv run main.py followed by arguments in "". See requirements below as well. A virtual environment is required and may be setup by installing UV below and entering source .venv/bin/activate within the project folder once configured.

Requirements

  • Python 3.10+
  • UV project and package manager
  • Runs a in zsh or bash, unix-like shell
  • Google GenAI (Gemini) 1.12.1
  • A Google AI Studio account
  • A Google AI API key generated (docs here)
  • Python dotenv 1.1.0

About

A coding assistant implemented in the terminal using the Gemini AI agent

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages