FinRAG combines Retrieval-Augmented Generation, semantic vector search, and GPT-4o to give you instant intelligence over SEC filings, earnings calls, and market reports.
No credit card required · Free during beta
Built on modern AI infrastructure
Platform Features
Retrieval-augmented generation tuned for SEC filings, earnings calls, and analyst reports. Get cited, grounded answers — not hallucinations.
Dense embeddings over your financial corpus. Find the relevant paragraph across thousands of documents in milliseconds.
Clerk-powered auth, encrypted Pinecone namespaces per user, and zero persistent logging of query content.
Token-by-token streaming with FastAPI + Server-Sent Events. Feels instant, even on long synthesis tasks.
System Architecture
01
Next.js + TailwindCSS
A streaming chat interface with real-time token output, document upload, and session history — built on the App Router with server components.
02
FastAPI Microservices
Python microservices handle document ingestion, chunking, embedding generation, and query orchestration — all behind authenticated REST endpoints.
03
LangChain + OpenAI
LangChain pipelines manage retrieval chains, context window injection, and streaming output. GPT-4o powers reasoning and synthesis over retrieved chunks.
04
Pinecone + Redis
Dense vector embeddings stored in Pinecone enable semantic similarity search. Redis caches embeddings and session state for sub-100ms repeat queries.
Technology Stack
Next.js
Frontend Framework
FastAPI
Backend Services
OpenAI
LLM Engine
LangChain
AI Orchestration
Pinecone
Vector Database
Clerk
Authentication
Redis
Caching Layer
Vercel
Deployment
Future Roadmap
Join the beta. No credit card required.