Skip to content

Repository files navigation

🏛️ PANEL

Python 3.10+Vue 3License: MITAutoGen

PRD from Agent Negotiation & Expert Logic — Multi-agent AI system that transforms project ideas into comprehensive PRDs through expert debates and judicial review.


📑 Table of Contents


✨ Features

FeatureDescription
🎯 Smart ClarificationAI asks targeted questions to understand your vision
💬 Expert Debate13+ specialized agents debate best approaches
⚖️ Judicial Review3 judges score business, technical & feasibility
📡 Real-time StreamingWatch agents discuss live via SSE
📦 Complete OutputDownload PRD + transcripts + exports as ZIP

Tech Stack: Python + FastAPI + AutoGen | Vue 3 + Vite + Tailwind | OpenAI GPT-4o


🚀 Quick Start

Prerequisites: Python 3.10+ • Node.js 18+ • Just • OpenAI API key

git clone https://github.com/CodeBlackwell/PANEL.git
cd PANEL
just setup # Install all dependencies# Edit backend/.env and add your OPENAI_API_KEY
just dev # Start backend + frontend

Open:http://localhost:5173

Manual setup (without Just)
# Backendcd backend && python -m venv venv &&source venv/bin/activate
pip install -r requirements.txt && cp .env.example .env
uvicorn app.main:app --reload --port 8000
# Frontend (new terminal)cd frontend && npm install && npm run dev

💡 Usage Example

1. Enter idea → "A marketplace for local farmers to sell directly to restaurants"
2. Configure → Select agents: Architect, Security, Backend Dev, UX
Debate rounds: 5
3. Clarify → Answer questions about target users, scale, budget
4. Watch debate → Agents discuss architecture, security, UX trade-offs
5. View scores → Business: 8/10 | Technical: 9/10 | Feasibility: 7/10
6. Download → Get PRD.md + transcripts + requirements.json

Sample output inexamples/prds/


🏗️ Project Structure

PANEL/
├── backend/
│ ├── app/
│ │ ├── main.py # FastAPI entry
│ │ ├── agents/ # AutoGen definitions
│ │ │ ├── definitions.py # Agent prompts
│ │ │ ├── teams.py # GroupChat configs
│ │ │ └── orchestrator.py # Workflow control
│ │ ├── api/routes/ # Endpoints
│ │ └── services/ # Business logic
│ └── requirements.txt
├── frontend/
│ └── src/
│ ├── views/ # Page components
│ ├── components/ # UI components
│ └── stores/ # Pinia state
├── agents/ # Agent definitions
└── examples/prds/ # Sample outputs

🤖 Agents

Debate Agents (select 2-6)

AgentFocusAgentFocus
🏗️ ArchitectSystem design, patterns🔐 SecurityAuth, compliance
⚙️ DevOpsCI/CD, infrastructure🎨 UXExperience, a11y
🧪 QATesting, quality📋 Product ManagerStories, roadmap
🗄️ Data EngineerPipelines, ETL🤖 ML EngineerModels, MLOps
🖥️ Frontend DevUI, performance⚡ Backend DevAPIs, caching
📱 Mobile DeviOS/Android📊 Business AnalystRequirements, ROI
👨‍💻 Tech LeadCode quality

Judge Agents

⚖️ Business⚖️ Technical⚖️ Feasibility
Market fit, revenueArchitecture, securityResources, timeline, risks

🔌 API Reference

MethodEndpointDescription
POST/api/v1/sessionsCreate session
GET/api/v1/sessions/{id}Get status
POST/api/v1/sessions/{id}/ideaSubmit idea
POST/api/v1/sessions/{id}/configConfigure agents
GET/api/v1/sessions/{id}/clarify/streamSSE: Clarification
POST/api/v1/sessions/{id}/answersSubmit answers
GET/api/v1/sessions/{id}/debate/streamSSE: Debate
GET/api/v1/sessions/{id}/judge/streamSSE: Evaluation
GET/api/v1/sessions/{id}/downloadDownload ZIP

⚙️ Configuration

Edit backend/.env:

OPENAI_API_KEY=your_key_hereMODEL_NAME=gpt-4oDEFAULT_DEBATE_ROUNDS=5MAX_DEBATE_ROUNDS=15STORAGE_PATH=./sessions

📦 Output

Generated ZIP structure:

prd_{session_id}.zip
├── PRD.md # Final document
├── original_prompt.txt # Your idea
├── transcripts/
│ ├── clarification.md # Q&A phase
│ ├── debate.md # Expert discussion
│ └── review.md # Judge evaluations
├── metadata/
│ └── scores.json # Numeric scores
└── exports/
├── requirements.json # Structured reqs
└── decisions.json # Key decisions

🛠️ Development

just test# Run all tests
just build # Production frontend build
just help# List all commands
RecipeDescription
just devRun backend + frontend
just backendBackend only
just frontendFrontend only
just testRun all tests
just buildProduction build
just cleanRemove generated files

📄 License

MIT License - see LICENSE for details.


Built with AutoGen + GPT-4o

About

Multi-agent AI system that stress-tests your ideas before you write a line of code and creates a hardened PRD. 13 specialized agents debate architecture, security, and UX 3 judges score the result then packages a complete PRD with transcripts and structured exports.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages