Skip to content

README.md

Latest commit

 

History

History
54 lines (44 loc) · 1.91 KB

README.md

File metadata and controls

54 lines (44 loc) · 1.91 KB

CodeCouncil — Multi-Agent PR Review System

Built for the Global AI Hackathon with Qwen Cloud (Track 3: Agent Society)

What is CodeCouncil?

CodeCouncil is a multi-agent PR review system where 4 specialized AI agents debate pull requests and produce a consensus verdict.

Architecture

CodeCouncil Architecture

Agents

  • 🔴 SecurityAuditor — finds vulnerabilities, SQL injection, hardcoded secrets
  • 🟡 PerfEngineer — detects N+1 queries, complexity issues, memory leaks
  • 🟢 CleanCodeReviewer — enforces naming, SOLID principles, code quality
  • 🟣 TechLead — reads the full debate and makes the final verdict

How it works

  1. PR opened on GitHub → webhook fires
  2. All 3 specialist agents analyze independently (Round 1)
  3. Agents read each other's opinions and debate (Round 2)
  4. TechLead arbitrates and posts verdict to GitHub PR (Round 3)

Proof of Alibaba Cloud Usage

This project uses Qwen Cloud (Alibaba Cloud) APIs via:

  • API endpoint: https://dashscope-intl.aliyuncs.com/compatible-mode/v1
  • Models used: qwen-plus (specialist agents), qwen-plus (TechLead)
  • See: agents/base_agent.py

Benchmark Results

CodeCouncil vs Single-Agent baseline:

  • +22% issue detection rate
  • 2x more issues found per PR (5.3 vs 2.7 average)
  • Full results: benchmark_results.json

Setup

git clone https://github.com/Efrrowini/CodeCouncil.git
cd CodeCouncil
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env
# Add your API keys to .env
uvicorn main:app --reload --port 8000

Architecture

  • FastAPI — webhook receiver and API gateway
  • Qwen Cloud — powers all 4 agents via OpenAI-compatible API
  • GitHub API — fetches PR diffs and posts review comments
  • Multi-round debate — 3 rounds of agent interaction

Track

Track 3: Agent Society