aicm 0.1.1

AI complete me.

aicm - AI Complete Me

A local-first, privacy-focused code completion CLI tool for developers who want AI assistance without sending their code to the cloud.

🚧 Work In Progress

This project is in active development. Core features are being implemented and APIs may change. Follow the progress and contribute!

🎯 Concept & Vision

acim is designed to be the Unix-philosophy code completion tool:

• Local-only: Your code never leaves your machine
• Lightweight: Optimized for Apple Silicon and modern hardware
• Fast: Instant suggestions with efficient models
• Out-of-the-box: default is best, no configuration needed in most cases
• Composable: Works with pipes, files, and editor integrations
• Hackable: Simple Python codebase, easy to customize

Philosophy

$ echo "def quick_osrt(arr):" |acim
def quick_sort(arr):
    if len(arr) <= 1:
        return arr
    else:
        pivot = arr[len(arr) // 2]
        left = [x for x in arr if x < pivot]
        middle = [x for x in arr if x == pivot]
        right = [x for x in arr if x > pivot]
        return quick_sort(left) + middle + quick_sort(right)

Just like cat, grep, or sed — acimdoes one thing well: complete your code.

✨ Planned Features

Core

• Pipe-based input (echo "..." |acim)
• Fill-in-the-middle (FIM) completion
• Single-line suggestions
• simple indent handle
• Daemon mode (persistent model in memory)
• Configuration file support

Models

• Qwen2.5-Coder-1.5B (default)
• Qwen2.5-Coder-0.5B (lightweight option)
• Custom GGUF model support
• Automatic model downloading

Integrations

• (Neo)vim plugin
• VSCode extension
• LSP protocol support

🏗️ Architecture (Conceptual)

┌─────────────────┐
│   Input Source  │  stdin / file / editor
│  (stdin/file)   │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  acimCLI        │  Argument parsing, context detection
│  (entry point)  │
└────────┬────────┘
         │
         ▼
┌─────────────────┐     ┌──────────────────┐
│  CodeComplete   │────▶│  llama-cpp-python │
│  Engine         │     │  (local inference) │
│                 │◀────│                   │
└────────┬────────┘     └──────────────────┘
         │
         ▼
┌─────────────────┐
│   Post-process  │  Indentation fix, deduplication
│                 │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│   Output        │  stdout / editor
│                 │
└─────────────────┘

🛤️ Roadmap

Phase 1: Foundation (Current)

• Basic completion via stdin
• Model loading and inference
• Indentation handling
• Configuration system
• Error handling & logging
• Installation via pip, uv...

Phase 2: Usability

• Daemon mode for instant response
• Model management (download, switch)
• Shell completions

Phase 3: Ecosystem

• Editor plugins
• LSP server mode
• Multi-language support beyond Python
• Fine-tuning capabilities

💡 Use Cases

Quick prototyping

# Start a new function
echo "def calculate_fibonacci(n):" |acim

🖥️ System Requirements

• macOS: Apple Silicon (M1/M2/M3) or Intel
• Linux: x86_64 or ARM64
• RAM: 4GB+ (for 1.5B model), 2GB+ (for 0.5B model)
• Python: 3.10+
• Storage: ~1.5GB for models

🤝 Contributing

This is a passion project. Contributions, ideas, and feedback are welcome!

📄 License

MIT License - see LICENSE for details.

---

Made with for the local-first AI movement.