Autoparts — AI‑assisted tool to turn Python files into a structured package

This repository contains a FastAPI backend (server.py + autoparts.py) and a Vite + React + Tailwind web UI (ui/). The goal is to take one or more .py files and transform them into a modular Python package, generate __main__.py, and produce a downloadable .zip. Optionally, it can use Ollama or OpenAI to suggest a package name and attempt small, safe edits.

Features

• Planning: Analyze input file(s) and propose a module plan (POST /plan, POST /plan_multi).
• Packaging: Reorganize code into a Python package and return a ZIP (POST /build, POST /build_multi).
• AI name suggestion: Suggest a package name via Ollama or OpenAI (optional).
• Compaction controls: Merge tiny modules, set target module count, line thresholds, etc.
• Web UI: Upload files, preview the plan, build, and download the ZIP — all in the browser.

---

Languages & Tech Stack

• Input language: Python (.py). The packager currently accepts Python sources only and outputs a structured Python package (with __init__.py and __main__.py).
• Backend: Python 3.10+, FastAPI, Uvicorn, python-multipart; optional Black for formatting.
• Frontend: TypeScript + React + Vite + Tailwind CSS.
• HTTP/CLI: Examples use cURL, but any HTTP client works.
• AI providers (optional): Ollama (local) and OpenAI (cloud) for name suggestions and small, safe edits.

---

Quick Start

1) Requirements

• Python ≥ 3.10
• Node.js ≥ 18 (compatible with Vite 5)
• (Optional) Ollama local server or an OpenAI API key

2) Clone & create a virtualenv

git clone <REPO_URL>.git
cd <REPO_NAME>

python -m venv .venv
# Windows
.\.venv\Scripts\activate
# macOS / Linux
source .venv/bin/activate

pip install -r requirements.txt

3) Environment variables

Create a .env file (you can copy from .env.example):

# .env
# Default AI settings for the CLI
AI_PROVIDER=openai  # or 'ollama'
AI_MODEL=gpt-4o-mini
AI_BASE_URL=https://api.openai.com/v1

# OpenAI API key
OPENAI_API_KEY=sk-...

# Ollama defaults for server endpoints
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3.1

Both Ollama and OpenAI are optional. If AI naming fails or is disabled, a local naming fallback is used.

4) Run the backend

uvicorn server:app --reload --port 8000

• Health check: GET http://localhost:8000/health
• Root: GET http://localhost:8000/ (lists available endpoints)

5) Run the frontend

cd ui
npm install
npm run dev

• The default API base is http://localhost:8000 (see API_BASE in ui/src/App.tsx).
• Production: npm run build creates ui/dist that you can serve behind any static server or reverse proxy (nginx, etc.).

---

API Usage

POST /plan

Analyze a single Python file and return a proposed module structure.

• Form field: file (single .py)

cURL

curl -X POST http://localhost:8000/plan   -F "file=@example.py"

POST /plan_multi

Create a plan for multiple files.

• Form fields: files (multiple .py)

curl -X POST http://localhost:8000/plan_multi   -F "files=@a.py" -F "files=@b.py"

POST /build

Package a single file and return a ZIP.

• Form fields:
  • file: .py file (required)
  • package_name: custom name (optional; provide it to disable AI name suggestion, or set ai_name=false)
  • compact: 0..3 compaction level (default 0)
  • pack_small_lines, max_modules, min_module_lines, target_modules: advanced compaction knobs
  • ai_name: true/false (default true)
  • ollama_base_url, ollama_model: (optional) for AI name suggestion via Ollama

cURL

curl -X POST http://localhost:8000/build   -F "file=@example.py"   -F "compact=2"   -o output.zip

POST /build_multi

Package multiple files into one package (returns ZIP).

• Form fields: files (multiple), plus the same compaction/AI options

curl -X POST http://localhost:8000/build_multi   -F "files=@a.py" -F "files=@b.py"   -F "compact=1"   -o output.zip

POST /ai_edit and POST /ai_edit_multi

Attempt limited, safe edits to the source using AI (Ollama/OpenAI) when requested.

---

Project Structure

autoparts.py       # Packaging/analysis logic (incl. AI name suggestion)
server.py          # FastAPI app and endpoints
ui/                # React + Vite + Tailwind web UI
  package.json
  src/
  ...

---

Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines. A list of people who have contributed is available in CONTRIBUTORS.md.