Fix Gemini API, increase OpenAI temperature, and bug fixes

2025-06-19 23:04:41 +05:30
parent cccd5173ea
commit 566b9ff994
4 changed files with 200 additions and 24 deletions
@@ -0,0 +1,43 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project Overview
+
+GIMS (Git Made Simple) is an AI-powered Git CLI tool that automatically generates meaningful commit messages from code changes. It's a Node.js package published to npm that integrates with OpenAI GPT-4 and Google Gemini APIs.
+
+## Architecture
+
+- **Single-file CLI**: All functionality is contained in `bin/gims.js`
+- **AI Integration**: Supports both OpenAI and Google Gemini APIs with intelligent fallback strategies
+- **Git Wrapper**: Built on top of `simple-git` library for Git operations
+- **Token Management**: Implements sophisticated content chunking to handle large diffs within AI token limits
+
+## Key Components
+
+- **Command System**: Uses `commander.js` for CLI argument parsing with aliases (e.g., `g o` for `gims online`)
+- **AI Message Generation**: Multi-strategy approach that falls back from full diff → summary → status → truncated content
+- **Commit Resolution**: Supports both commit hashes and numbered indices for referencing commits
+- **Safe Operations**: Includes error handling for empty repositories and edge cases
+
+## Environment Setup
+
+Required environment variables (at least one):
+- `OPENAI_API_KEY` - For OpenAI GPT-4o-mini integration
+- `GEMINI_API_KEY` - For Google Gemini 2.0 Flash integration
+
+## Common Commands
+
+- **Install globally**: `npm install -g .`
+- **Test locally**: `node bin/gims.js --help`
+- **Test specific command**: `node bin/gims.js suggest`
+- **Run with alias**: `g o` (after global install)
+
+## Development Notes
+
+- No test framework is currently configured (package.json shows placeholder test script)
+- Node.js version requirement: >=20.0.0
+- Uses CommonJS modules (`require`/`module.exports`)
+- Dependencies are minimal and focused on core functionality
+- Token estimation uses 4 characters per token approximation
+- Maximum context limit set conservatively at 100,000 tokens