Back to Projects
Chaibook LLM
AI & RAG SystemsNotebookLM-Inspired Document Intelligence & RAG Platform
Flagship AI application enabling users to upload PDFs, text documents, URLs, and YouTube videos for deep conversational context retrieval, source citations, and right-sidebar evidence verification.
System Architecture & Pipeline Mechanics
Architecture Flow:S3 Upload → BullMQ Background Ingestion Worker → Text Extraction & Chunking → OpenAI Embedding Generation → Qdrant Vector Index → Cohere Re-ranker → LLM Streaming Response.
Key Engineering Capabilities:
Multi-source ingest: PDFs, raw text, web URLs, and YouTube video transcripts
Interactive right-sidebar evidence panel displaying verbatim source citations with timestamps
Input & Output Guardrails to prevent hallucination and prompt injection
Streaming LLM synthesis with fast token response rendering
Resource card organization for multi-document research workspaces
Technology & Libraries Used
Next.jsTypeScriptQdrantOpenAI APIRAGBullMQRedisS3