simonw/llm-gemini

What it solves

This project provides a command-line interface and Python library plugin to access Google's Gemini models directly from your terminal. It bridges the gap between the llm ecosystem and Gemini's capabilities, allowing users to interact with advanced AI models without writing custom API integration code.

How it works

It functions as a plugin for the llm tool. Once installed and configured with an API key, it allows you to invoke various Gemini models (including flash and pro versions) using simple CLI commands. It supports multimodal inputs (images, audio, video), advanced features like Google Search grounding, code execution, and URL context, as well as generating vector embeddings for data collections.

Who it’s for

This is for developers and power users who already use the llm CLI tool and want to integrate Google's Gemini models into their local workflows, automation scripts, or data processing pipelines.

Highlights

• Multimodal Support: Easily process images, audio, video, and YouTube URLs directly from the command line.
• Advanced Tooling: Enables server-side features like Google Search grounding, secure code execution, and URL context fetching.
• Embedding Generation: Provides tools to generate and store vector embeddings for similarity searches in local SQLite databases.
• Configurable Reasoning: Supports model-specific "thinking" levels or budgets to control the depth of the model's reasoning process.
• Seamless Integration: Fully compatible with the broader llm plugin ecosystem, allowing for easy switching between different model providers.