CodeView MCP 🪄

Powered by MCP, CodeLlama-13B (local), Llama-3.1-8b-instant (cloud)

1 Why

Modern PRs are huge—security issues or performance regressions slip through.
ReviewGenie does a 30-second AI review:

• Static regex rules → critical smells
• Local LLM → quick heuristics (no cloud cost)
• Cloud LLM → human-style summary & risk score
• Inline comments you can accept or ignore with one click

2 What it does

Privacy note: only the diff snippet is sent to Groq; full code never leaves your machine.

3 Quick Start (5 min)

git clone https://github.com/mann-uofg/codeview-mcp.git
cd codeview-mcp
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
pip install -e .

# one-liner smoke
reviewgenie/codeview ping https://github.com/psf/requests/pull/6883

Store secrets once (env-var OR keyring):

from codeview_mcp.secret import set_in_keyring

set_in_keyring("GH_TOKEN",        "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")    #GitHub PAT  
set_in_keyring("OPENAI_API_KEY",  "gsk_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx")              # Groq/OpenAI key  
set_in_keyring("OPENAI_BASE_URL", "https://api.groq.com/openai/v1")

Full tutorial: docs/QUICKSTART.md

4 Architecture

• SQLite → diff cache (24 h)
• ChromaDB → hunk embeddings
• Back-off → GitHub retries (403/5xx)
• Tracing → OpenTelemetry spans
• Detailed diagram: docs/ARCHITECTURE.md

5 Benchmark

See bench/benchmarks.md: 10 popular OSS PRs → avg ⏱ 8.1 s analyze, 💰 $0.0008 Groq cost, 96 % comment acceptance.

6 Docs

• API schema: docs/API_SCHEMA.json
• CLI reference: docs/USAGE.md
• Config & env: docs/CONFIGURATION.md
• Contributing: docs/CONTRIBUTING.md

7 Day-by-Day Log

Full changelog: docs/CHANGELOG.md

8 Roadmap

• 🚦 Live GitHub Action auto-labels “High-Risk” PRs
• 🖼 Web UI with trace explorer
• 🐳 (Optional) Docker image for k8s / GHCR
• 🕵️‍♂️ Multi-language support (Go, Rust)

Star the repo ⭐ & drop an issue if you’d like to help!