Emacs AI Integration

Features

A comprehensive AI assistant for Emacs with support for multiple providers (developed for OpenAI and Claude.ai integration) and extensive features for code analysis, text editing, and project management.

• Multi-provider support: OpenAI, Claude, Gemini, Ollama, llama.cpp
• Streaming and non-streaming responses
• Code analysis: explain, fix, review (security, performance, style, bugs)
• Text editing: grammar, style, spelling fixes
• Project context awareness
• Session management with save/load
• Buffer-to-session association
• Integration with org-mode, magit, lsp-mode, projectile

Installation

1. Download ai-integration.el to your Emacs configuration directory
2. Add to your .emacs or init.el:

(load-file "~/.emacs.d/ai-integration.el")
(require 'ai-integration)

Configuration

API Keys (Recommended: Environment Variables)

export OPENAI_API_KEY="your-openai-key-here"
export ANTHROPIC_API_KEY="your-claude-key-here"
export GEMINI_API_KEY="your-gemini-key-here"

Emacs Configuration

;; Basic configuration
(setq ai-default-provider "openai")           ; or "claude", "gemini", "ollama", "llama-cpp"
(setq ai-streaming-default t)                 ; Enable streaming by default
(setq ai-auto-associate-buffers t)            ; Auto-associate buffers with sessions

;; Optional: Set API keys in Emacs (less secure than env vars)
(setq ai-openai-api-key "your-key-here")
(setq ai-claude-api-key "your-key-here")
(setq ai-gemini-api-key "your-key-here")

;; Optional: Customize models
(setq ai-openai-model "gpt-4o")
(setq ai-claude-model "claude-sonnet-4-20250514")
(setq ai-gemini-model "gemini-1.5-flash")

Local AI Providers

For Ollama:

(setq ai-ollama-endpoint "http://localhost:11434/api/chat")
(setq ai-ollama-model "llama3.2")

For llama.cpp:

(setq ai-llama-cpp-endpoint "http://localhost:8080/v1/chat/completions")
(setq ai-llama-cpp-model "llama-3.2")

Quick Start

1. Start a chat session: C-c a RET
2. Send code for analysis: Select region → C-c a c e r (explain code region)
3. Fix text: Select text → C-c a t g r (fix grammar in region)
4. Quick actions: C-c a q → select action

Key Bindings

Main Commands

• C-c a RET - Start AI chat
• C-c a N - New session
• C-c a S - Save session
• C-c a L - Load session
• C-c a P - Change provider
• C-c a O/C/G - Switch to OpenAI/Claude/Gemini
• C-c a v - Toggle streaming
• C-c a k - Cancel request

Code Commands (C-c a c)

• e r/b/p/f - Explain region/buffer/point/file
• f r/b/p/f - Fix code
• r r/b/p/f - Review code (comprehensive)
• s r/b/p/f - Security review
• p r/b/p/f - Performance review
• y r/b/p/f - Style review
• b r/b/p/f - Bug review
• TAB - Complete at point

Text Commands (C-c a t)

• s r/b/p/f - Fix style
• g r/b/p/f - Fix grammar
• l r/b/p/f - Fix spelling
• e r/b/p/f - Explain text

Send Commands (C-c a s)

• r/b/f/p - Send to existing session
• R/B/F/P - Send to new session

In AI Chat Buffers

• C-c C-c - Send input
• C-c C-v - Send input (non-streaming)
• C-c C-t - Toggle streaming
• C-c C-r - Regenerate last response
• C-c C-k - Cancel request
• C-c C-m - Select model
• C-c C-y - Copy last response
• C-c C-x s - Set system prompt
• C-c C-x t - Use template

Usage Examples

Code Analysis

;; Select a function and explain it
C-c a c e r

;; Review entire buffer for security issues
C-c a c s b

;; Fix code at point
C-c a c f p

Text Editing

;; Fix grammar in selected text
C-c a t g r

;; Improve writing style of paragraph
C-c a t s p

;; Fix spelling in entire buffer
C-c a t l b

Interactive Chat

;; Start chat session
C-c a RET

;; In the chat buffer, type your message and press:
C-c C-c    ; Send with streaming
C-c C-v    ; Send without streaming

Advanced Features

Session Management

• Sessions are automatically associated with source buffers
• Save/load conversations for later reference
• Multiple sessions per buffer supported

Provider Switching

• Change providers mid-conversation
• Each provider has optimized settings
• Support for local AI models

Templates and System Prompts

• Use predefined conversation templates
• Set custom system prompts for different roles
• Quick actions for common tasks

Troubleshooting

Enable Debug Mode

M-x ai-toggle-debug

Test Streaming

M-x ai-diagnose-streaming

Check Configuration

M-x ai-debug-session-state

Contributing

This project welcomes contributions! Please feel free to:

• Report bugs and issues
• Suggest new features
• Submit pull requests
• Improve documentation

License

MIT License - see the source file for details.

---

Note: This package requires Emacs 27.1+ and active internet connection for cloud AI providers. Local providers (vLLM, Ollama, llama.cpp) work offline once set up.