KnowledgeDoc

KnowledgeDoc is a RAG knowledge base: upload PDFs/TXT/MD, chunk and embed with OpenAI, store vectors in PostgreSQL pgvector, and answer questions with cited passages and similarity scores.

Features

• Semantic search
• Cited passages with similarity scores
• Document chunking pipeline

Architecture

• API. Node.js + TypeScript
• Embeddings. OpenAI embeddings
• Store. PostgreSQL + pgvector
• Retrieval. Cited passage ranking

Challenges

• Chunking documents without losing context
• Returning trustworthy cited passages
• Keeping similarity search fast as corpora grow

Metrics

• Search. Semantic
• Citations. Scored
• Ingest. PDF/TXT/MD

Results

• Cited answers grounded in uploaded docs
• Reusable chunking + embedding pipeline
• pgvector-backed retrieval layer.