ResearchHive
A full-stack, AI-powered research assistant that autonomously searches the web, scrapes deep content, drafts comprehensive reports, and critically evaluates its own work—all streamed in real-time.
The Problem
Internet research is often a tedious and manual process requiring extensive searching, reading, synthesizing, and evaluating of information from various sources.
The Solution
ResearchHive automates internet research using a sequential pipeline of specialized AI agents (Searcher, Reader, Writer, Critic). It streams agent thoughts and final reports in real-time to a stunning glassmorphic UI using Server-Sent Events.
Key Features
- Multi-Agent AI Pipeline (Searcher, Reader, Writer, Critic)
- Real-Time Streaming via Server-Sent Events (SSE)
- Secure Authentication with Supabase Auth
- Asynchronous Database Operations with SQLAlchemy 2.0
- Glassmorphic UI/UX with Micro-Animations
Technical Highlights
- Built a fully asynchronous FastAPI backend with non-blocking SSE streaming
- Engineered a multi-agent LangChain pipeline with specialized agent roles
- Designed a custom glassmorphic React interface with Aurora backgrounds from scratch
- Implemented robust stateless JWT authentication via query parameters for SSE
Challenges Faced
- 01
Handling real-time state streaming through SSE without blocking the main event loop
- 02
Managing database connections with Supabase transaction poolers asynchronously
Tech Stack
7 technologies used