CoolFace
Apppublic

ImTamsi/ai-code-reviewer

sourceHugging Faceapache-2.0updated 3mo agoView on Hugging Face
0likes
App README

AI Code Reviewer (Space)

Heads-up: this Space does not ship a hosted GPU. It is a Gradio front-end that calls your own OpenAI-compatible Qwen endpoint (vLLM on AWS, local GPU, etc.). Without secrets configured, the app will show a setup banner and health-check errors.

Gradio demo of the AI Code Reviewer MCP project. Enter a public GitHub repository and the app runs LLM-powered analyses (code review, potential bugs, technical debt, security, performance, missing tests) using a Qwen3.6-27B model served behind an OpenAI-compatible endpoint.

Quick start (Space maintainer)

  1. 1.Open Settings → Variables and secrets on this Space.
  2. 2.Add the required secrets below.
  3. 3.Restart the Space (or push a commit) and wait for the build.
  4. 4.Try an example repo such as psf/requests.

Configuration

Set these as Space secrets (Settings → Variables and secrets):

VariableRequiredDescription
QWEN_BASE_URLyesOpenAI-compatible endpoint, e.g. http://<host>:8000/v1
QWEN_API_KEYyesAPI key for the endpoint
LLM_MODELnoModel id (default Qwen/Qwen3.6-27B)
GITHUB_TOKENnoRaises GitHub API rate limits

Serving recipe and JSON output contract: ImTamsi/qwen3.6-27b-code-reviewer

Sample findings dataset: ImTamsi/code-review-findings-samples

The prompts in prompts/ are copied from the MCP server package so both share the exact same review behavior.

Run locally

bash
pip install -r requirements.txt
export QWEN_BASE_URL=http://localhost:8000/v1
export QWEN_API_KEY=...
python app.py

Related