🚀 DeepSight Engine

A production-grade Agentic Deep Research Engine built with LangGraph that autonomously plans, researches, critiques, and generates citation-backed research reports using multiple collaborating AI agents.

---

🏗️ System Architecture

The architecture below illustrates the complete end-to-end workflow of DeepSight Engine.

---

---

📸 Application Preview

🏠 Home Screen

---

🔄 Research Progress

Real-time streaming of agent execution using Server-Sent Events (SSE).

---

📑 Generated Research Report

Structured, citation-backed report generated after the multi-agent workflow completes.

---

✨ Key Features

• 🧠 Dynamic query decomposition using a dedicated Planner Agent
• 🔎 Autonomous multi-step web research with Researcher Agent
• 📝 Structured report generation with Reporter Agent
• 🧐 Quality evaluation through a Critic Agent
• ✍️ Automatic refinement using an Editor Agent
• 🌐 Multi-provider LLM Gateway (Llama, OpenAI & Claude)
• ⚡ Real-time progress streaming via Server-Sent Events (SSE)
• 🧠 Redis Semantic Caching for repeated and semantically similar queries
• 📄 PDF & DOCX report export
• 📊 Usage tracking and API key management
• 🚀 Production-ready FastAPI backend

---

🤖 Multi-Agent Workflow

Unlike a traditional chatbot that relies on a single LLM response, DeepSight Engine orchestrates multiple specialized AI agents.

🧠 Planner Agent

• Understands the user's objective
• Breaks complex queries into dynamic research objectives
• Creates an execution strategy

🔎 Researcher Agent

• Executes one research objective at a time
• Performs iterative web research
• Collects and consolidates evidence

📝 Reporter Agent

• Synthesizes collected information
• Generates a structured research report

🧐 Critic Agent

• Evaluates report quality
• Detects missing information
• Decides whether refinement is required