lea_code 0.0.1

A command-line coding assistant powered by Genkit and Google Gemini.

lea_code

lea_code is a Dart command-line coding assistant powered by Genkit and Google Gemini. It runs as an interactive terminal chat and gives the model a small set of local file and shell tools so it can inspect and modify code in your working directory.

Features

• Interactive CLI chat loop
• Google Gemini model selection via a command-line flag
• Optional custom system prompt
• Conversation reset with /new
• Built-in tools for:
  • running shell commands
  • reading files
  • writing files
  • searching file contents
  • finding files by name

Requirements

• Dart SDK ^3.11.4
• A Google AI API key exposed as either GOOGLE_GENAI_API_KEY or GEMINI_API_KEY

Install

dart pub global activate lea_code

Run

lea_code

You can also choose a model and provide a system prompt:

lea_code \
  --model gemini-flash-lite-latest \
  --system_prompt "You are a careful coding assistant."

CLI Options

• -m, --model: Gemini model name to use. Defaults to gemini-flash-lite-latest.
• -s, --system_prompt: Optional system prompt passed as output instructions.

Interactive Commands

• /new: clear the current conversation history
• /exit: quit the program
• /quit: quit the program

Built-in Tools

The assistant registers five tools in LeaCodeEngine:

• bash: executes a shell command with bash -c
• file_read: reads a file from disk
• file_write: writes content to a file
• string_search: searches the current directory with grep -r
• file_find: finds files and directories with find

These tools operate on the local machine, so use this project only in directories and environments you trust.

Project Structure

• bin/lea_code.dart: CLI entrypoint and REPL loop
• lib/lea_code_engine.dart: Genkit setup and tool registration
• lib/tools/: tool definitions exposed to the model
• lib/models/: typed tool input models and generated schema files

Notes

• The application exits early if neither GOOGLE_GENAI_API_KEY nor GEMINI_API_KEY is set.
• Tool output is returned directly to the model, including shell stderr when present.
• string_search and file_find run relative to the current working directory.