| title | AI Code Reviewer Pro |
|---|---|
| emoji | ⌘ |
| colorFrom | indigo |
| colorTo | purple |
| sdk | gradio |
| sdk_version | 5.50.0 |
| app_file | app.py |
| pinned | false |
| license | mit |
A multi-language AI code review application with a clean, light-only UI. Paste code, pick a language and a review lens, and get a structured, principal-level review — bugs, security, performance, architecture, best-practices, complexity analysis, and a fully refactored version — all rendered as an interactive report you can copy or download.
Provider-agnostic: run it on Groq, Google Gemini, OpenRouter, or OpenAI. Check the provider's current pricing and rate limits before configuring a key.
- Light-only, responsive UI built with Gradio Blocks and custom CSS.
- Import a source file: load a UTF-8 code file into the editor. Its extension selects the language when supported; click Review code when ready.
- Live input count: see the 30,000-character limit before submitting a review.
- Compact results: expand findings, code quality, and report sections as needed.
- Session history: reopen any of the last 20 reviews, including its code and report, or clear the current session's review list.
- Safer review flow: empty and oversized inputs are caught before an API call, and every download gets a unique filename.
- Sample snippet to explore the app without supplying your own code.
- Works without an API key: local static checks flag selected Python syntax,
mutable defaults, recursion, bare exception handlers, and dynamic execution;
JavaScript/TypeScript
evaland SQLSELECT *are also flagged. The local scan is limited and does not assign an AI score or generate a rewrite. - Syntax-highlighted code editor that adapts to the selected language.
- 15 languages: Swift, Dart, Python, JavaScript, TypeScript, Java, Kotlin, Go, Rust, C#, C++, PHP, Ruby, SQL, HTML/CSS.
- 8 review modes: General, Bugs, Performance, Security, Best Practices, Clean Architecture, Refactoring, Testing.
- Structured output:
- Overall score out of 10
- Bugs (with severity, line, and suggested fix)
- Security issues
- Performance issues
- Architecture review (strengths, concerns, suggestions)
- Best-practice suggestions
- Complexity analysis (time/space Big-O, cyclomatic estimate + static proxy)
- Refactored code
- Markdown report with copy and one-click download.
git clone https://github.com/ananya-kaul/AI-Code-Reviewer-Pro.git
cd AI-Code-Reviewer-Pro
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txtThe app starts without a key and runs limited static checks locally. For a full AI review, it auto-detects which provider to use from whichever API key is present. You only need one:
| Provider | Cost | Get a key |
|---|---|---|
| Groq (recommended) | Free tier and usage pricing | https://console.groq.com/keys |
| Google Gemini | Depends on model and account | https://aistudio.google.com/app/apikey |
| OpenRouter | Free :free models |
https://openrouter.ai/keys |
| OpenAI | Paid | https://platform.openai.com/api-keys |
cp .env.example .env
# then edit .env and set ONE key, e.g. GROQ_API_KEY=gsk_...Or export it directly:
export GROQ_API_KEY="gsk_your_free_key_here"python app.pyOpen the printed local URL (default http://localhost:7860).
| Variable | Default | Description |
|---|---|---|
GROQ_API_KEY |
— | Groq key (free). Auto-detected first. |
GEMINI_API_KEY |
— | Google Gemini key (free). |
OPENROUTER_API_KEY |
— | OpenRouter key (free models). |
OPENAI_API_KEY |
— | OpenAI key (paid). |
LLM_PROVIDER |
(auto) | Force groq | gemini | openrouter | openai. |
LLM_MODEL |
(provider default) | Override the model for the active provider. |
GRADIO_SERVER_NAME |
0.0.0.0 |
Host interface to bind. |
GRADIO_SERVER_PORT |
7860 |
Port to serve on. |
For AI reviews, set one provider key. Auto-detection order: Groq → Gemini → OpenRouter → OpenAI. The Groq default is
openai/gpt-oss-120b; the Gemini default isgemini-3.6-flash. If either provider rejects a retired model ID, the app retries a supported model from that provider and shows the model it used.
AI-Code-Reviewer-Pro/
├── app.py # Gradio UI and callbacks
├── reviewer.py # Review engine (multi-provider, OpenAI-compatible)
├── prompts.py # Prompt engineering + JSON contract
├── utils.py # Catalogs, metrics, Markdown rendering
├── theme.css # Light responsive theme
├── requirements.txt
├── README.md
├── .env.example
├── .gitignore
└── assets/ # Logo and generated reports
prompts.pybuilds a strict system + user prompt, including a JSON schema the model must satisfy and a mode-specific review lens.reviewer.pycalls the active provider's OpenAI-compatible API when a key is available, or runs limited static checks locally. Provider responses are normalized so every expected field is present. Static metrics are computed locally.app.pyrenders the structured result into a light workspace and a downloadable Markdown report.
Review history is kept in the current browser session. Generated reports are
stored under assets/reports/ on the server and excluded from Git.
This repo ships a render.yaml Blueprint.
- Go to https://dashboard.render.com/blueprints → New Blueprint Instance.
- Connect this GitHub repo (
ananya-kaul/AI-Code-Reviewer-Pro). - Render reads
render.yamland provisions a free web service. - In the service's Environment, optionally set one provider key such as
GROQ_API_KEYto enable full AI reviews. Without a key, the local static checks remain available.
The free plan sleeps after inactivity and cold-starts on the next visit.
A portable Dockerfile is included:
docker build -t ai-code-reviewer-pro .
docker run -p 7860:7860 -e OPENAI_API_KEY=sk-your-key ai-code-reviewer-proThe app binds to $PORT when the host provides one, so it works on any
container platform out of the box.
The repo is also Spaces-ready (see the YAML header above). Note that hosting a
Gradio Space on HF's free CPU tier currently requires a PRO subscription;
create the Space, push this repo, and set OPENAI_API_KEY as a Space secret.
Released under the MIT License.