🛰️ News Aggregator Launch Assistant

AI-powered release readiness evaluation system for news aggregation platforms.
This project analyzes launch plans using LLM agents + Retrieval-Augmented Generation (RAG) to deliver evidence-backed readiness scores, risks, and strategic recommendations.

Built as a production-style AI agent system with persistent audit history, explainable evidence traces, and a high-fidelity dashboard UI.

---

🚀 Key Capabilities

🧠 AI & Agents

• Launch Planner Validation Agent
  • Converts raw ideas into structured launch plans
  • Outputs steps, risks, owners, and success metrics
• Release Readiness Evaluation Agent
  • Scores readiness (LOW / MEDIUM / HIGH)
  • Produces confidence score + actionable recommendations
  • Uses RAG grounding for factual, explainable outputs

📚 Retrieval-Augmented Generation (RAG)

• FAISS-based semantic retrieval
• Seeded domain documents:
  • Scalability best practices
  • Publisher licensing constraints
  • Privacy & GDPR compliance
• Every recommendation is evidence-backed

🗂️ Persistence & Auditing

• All agent runs are stored in a database
• History sidebar with:
  • Title generation
  • Readiness score
  • Timestamped audit trail
• Full replay of:
  • Input plan
  • AI output
  • Retrieved evidence snippets

🖥️ Frontend Experience

• Futuristic glass-morphic dashboard
• Live agent execution
• Confidence score visualization
• Structured recommendations
• Transparent RAG evidence trace
• Retry-safe API hooks & loading state

🏗️ System Architecture

User Input
↓
React Dashboard (Vite)
↓
FastAPI Backend
├── Planner Validation Agent (Gemini)
├── Release Readiness Agent
│   ├── FAISS RAG Retrieval
│   ├── Evidence Synthesis
│   └── Gemini Reasoning
├── JSON Output Validator
└── Persistence Layer (SQLite / Postgres)
↓
Explainable AI Results + Audit History

---

🛠️ Tech Stack

Backend

• Python
• FastAPI
• Google Gemini (LLM)
• FAISS (Vector Search)
• SQLAlchemy
• Alembic
• SQLite (default) → PostgreSQL ready

Frontend

• React
• Vite
• Custom Hooks (retry logic)
• Advanced CSS (glassmorphism, animations)

---

📂 Project Structure

news-launch-assistant/
├── backend/
│   ├── Dockerfile
│   ├── requirements.txt
│   ├── main.py
│   ├── agents/
│   ├── api/
│   ├── rag/
│   ├── db/
│   └── utils/
├── frontend/
│   ├── Dockerfile
│   ├── src/
│   ├── index.html
│   └── package.json
├── docker-compose.yml
└── README.md

---

🔌 Core API Endpoints

Method	Endpoint	Description
GET	/health	Service health check
POST	/planner/validate	Validate & structure launch plan
POST	/agent/run	Execute release readiness agent
GET	/history	List past agent runs
GET	/history/{id}	Retrieve full audit trace
GET	/features	Capability discovery
POST	/ai/gemini	Gemini proxy

---

🔍 Example Agent Output

{ "readiness": "HIGH", "score": 95, "recommendations": [ "Confirm asynchronous ingestion pipelines for traffic spikes.", "Verify fallback strategies for publisher API downtime.", "Ensure continuous review of licensing agreements." ] }

Each recommendation is accompanied by retrieved RAG evidence.

---

🎓 Academic & Practical Value

• Multi-agent LLM orchestration
• RAG system design
• Explainable AI pipelines
• Full-stack AI engineering
• Production-grade persistence & auditing

Suitable for coursework, graduation projects, and AI portfolios.

---

🚀 Future Improvements

• PostgreSQL + pgvector
• Authentication & multi-user history
• Streaming agent responses
• Live document ingestion
• Model comparison (Gemini / GPT / Claude)
• CI-based readiness scoring