Skip to main content

CLI tool for interacting with Gemini models

Project description

Gemini Code

A powerful AI coding assistant for your terminal, powered by Gemini 2.5 Pro with support for other LLM models.

Features

  • Interactive chat sessions in your terminal
  • Multiple model support (Gemini 2.5 Pro, Gemini 1.5 Pro, and more)
  • Intelligent context management with auto-compaction warnings and the /compact command
  • Markdown rendering in the terminal
  • Automatic tool usage by the assistant:
    • File operations (view, edit, list, grep, glob)
    • System commands (bash)
    • Web content fetching

Installation

# Clone the repository
git clone https://github.com/raizamartin/gemini-code.git
cd gemini-code

# Install the package
pip install -e .

Setup

Before using Gemini CLI, you need to set up your API keys:

# Set up Google API key for Gemini models
gemini setup YOUR_GOOGLE_API_KEY

Usage

# Start an interactive session with the default model
gemini

# Start a session with a specific model
gemini --model models/gemini-2.5-pro-exp-03-25

# Set default model
gemini set-default-model models/gemini-2.5-pro-exp-03-25

# List all available models
gemini list-models

Interactive Commands

During an interactive session, you can use these commands:

  • /exit - Exit the chat session
  • /help - Display help information
  • /compact - Summarize the conversation to reduce token usage

How It Works

Tool Usage

Unlike direct command-line tools, the Gemini CLI's tools are used automatically by the assistant to help answer your questions. For example:

  1. You ask: "What files are in the current directory?"
  2. The assistant uses the ls tool behind the scenes
  3. The assistant provides you with a formatted response

This approach makes the interaction more natural and similar to how Claude Code works.

Context Management

Gemini CLI intelligently manages the conversation context:

  1. Warning Threshold (80%): When you reach 80% of the token limit, you'll see a warning panel suggesting to use /compact
  2. Auto-Compact Prompt (95%): At 95% of the limit, the CLI will ask if you want to automatically compact the conversation
  3. Manual Compaction: You can use /compact at any time to summarize the conversation and reduce token usage

The summarization process preserves important context while significantly reducing token count, allowing for virtually unlimited conversation length.

Development

This project is under active development. More models and features will be added soon!

Recent Changes in v0.1.19

  • Improved system prompt to encourage more active tool usage
  • Added thinking/planning phase to help Gemini reason about solutions
  • Enhanced response format to prioritize creating and modifying files over printing code
  • Filtered out thinking stages from final output to keep responses clean
  • Made Gemini more proactive as a coding partner, not just an advisor

Recent Changes in v0.1.18

  • Updated default model to Gemini 2.5 Pro Experimental (models/gemini-2.5-pro-exp-03-25)
  • Updated system prompts to reference Gemini 2.5 Pro
  • Improved model usage and documentation

Recent Changes in v0.1.17

  • Added list-models command to show all available Gemini models
  • Improved error handling for models that don't exist or require permission
  • Added model initialization test to verify model availability
  • Updated help documentation with new commands

Recent Changes in v0.1.16

  • Fixed file creation issues: The CLI now properly handles creating files with content
  • Enhanced tool pattern matching: Added support for more formats that Gemini might use
  • Improved edit tool handling: Better handling of missing arguments when creating files
  • Added special case for natural language edit commands (e.g., "edit filename with content: ...")

Recent Changes in v0.1.15

  • Fixed tool execution issues: The CLI now properly processes tool calls and executes Bash commands correctly
  • Fixed argument parsing for Bash tool: Commands are now passed as a single argument to avoid parsing issues
  • Improved error handling in tools: Better handling of failures and timeouts
  • Updated model name throughout the codebase to use gemini-1.5-pro consistently

Known Issues

  • If you created a config file with earlier versions, you may need to delete it to get the correct defaults:
    rm -rf ~/.config/gemini-code
    

License

MIT

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

gemini_code-0.1.19.tar.gz (18.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

gemini_code-0.1.19-py3-none-any.whl (19.7 kB view details)

Uploaded Python 3

File details

Details for the file gemini_code-0.1.19.tar.gz.

File metadata

  • Download URL: gemini_code-0.1.19.tar.gz
  • Upload date:
  • Size: 18.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.13.1

File hashes

Hashes for gemini_code-0.1.19.tar.gz
Algorithm Hash digest
SHA256 c93ae14e1e0079c387b84727163c5e4bf931b8eb5be7368b9a0b9a7a66612bcb
MD5 c020331949f2b913fcc40501aa961ed4
BLAKE2b-256 b9dc836a76f21e51b6fc06a1f245044a54dfd70b78110691424a0c3036aa4f93

See more details on using hashes here.

File details

Details for the file gemini_code-0.1.19-py3-none-any.whl.

File metadata

  • Download URL: gemini_code-0.1.19-py3-none-any.whl
  • Upload date:
  • Size: 19.7 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.13.1

File hashes

Hashes for gemini_code-0.1.19-py3-none-any.whl
Algorithm Hash digest
SHA256 98dc465ccedbfa9ea445a2d23d544358bdc40ba43a3d0bfce5f06946f744e7d4
MD5 1da3a2304d79d58de0f19bd185bbc01e
BLAKE2b-256 8557b03ca480726c94b32a80bfa534d87e9578b9432511bd4a01646afb9c4d51

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page