jit-setup 0.5.0

AI-powered project environment setup — scan, configure, verify, done.

     _ _ _
    (_|_) |_
    | | | __|
    | | | |_
   _/ |_|\__|
  |__/
  

jit clone — a just-in-time clone that's ready to run.

Like git clone, but your dev environment is already set up when it's done.
An AI agent that reads your project, figures out what's needed, and gets everything running.

中文文档

---

The Problem

You find a repo. Then you spend 20 minutes reading the README, figuring out which Python version to use, creating a venv, installing dependencies, hunting down .env variables, and debugging why torch won't import. Multiply that across every new project, every new teammate's machine.

The Fix

pip install jit-setup

jit clone https://github.com/user/repo

One command. It clones the repo, scans the project, talks to you about what it finds, and sets everything up — virtual environment, dependencies, env vars, the works. Like git clone, but the environment is ready when it's done. jit = just in time — you clone it, it's ready.

How It Works

jit clone <url>                  # clone + set up in one shot
jit                              # set up current directory
jit /path/to/project             # set up a specific project
jit --yes                        # auto-confirm system-level ops

Under the hood:

1. Scan — reads your project structure: languages, package managers, config files, Dockerfiles, GPU hints
2. Plan — figures out what to install, asks your preference (venv vs conda vs uv, npm vs pnpm, etc.)
3. Execute — creates environments, installs deps, writes .env files, runs setup scripts
4. Verify — runs a final check to make sure everything actually works

Supports Python, Node.js, Rust, Go, Ruby, Docker, and more.

What Makes It Different

• Zero config — no YAML, no plugins, no project-specific setup files to maintain
• Conversational — it asks before doing anything destructive, explains what it finds
• Safe by design — commands are sandboxed to the project directory; system-level ops require confirmation; dangerous commands are blocked entirely
• Free to use — ships with a community API (30 requests/day), no API key needed to get started
• Zero dependencies — pure Python stdlib, installs in seconds

Bring Your Own LLM

The free community API works out of the box. For unlimited use or a different model, configure ~/.jitx/config.json:

{
  "llm": {
    "base_url": "https://api.openai.com/v1",
    "api_key": "sk-xxx",
    "model": "gpt-4o"
  }
}

Works with any OpenAI-compatible API (OpenAI, SiliconFlow, Ollama, vLLM, etc.).

Security

jit takes a defense-in-depth approach:

• Project-scoped — file reads and command execution are restricted to the project directory
• Three-tier command policy — safe commands auto-run, system commands ask for confirmation, dangerous commands are blocked
• No data leaves your machine except LLM API calls (the conversation about your project structure)
• HMAC-signed requests to the community API to prevent abuse

Requirements

• Python 3.9+
• That's it

License

Apache-2.0