Full-stack

Confluence RAG Knowledge Base

A retrieval-augmented Q&A system over Confluence that stays fresh automatically.

Personal Project––– views

Confluence RAG Knowledge Base preview

Problem: Teams knew the answer was documented in Confluence, but finding and piecing it together across spaces took about 30 minutes.
Approach: A FastAPI service that syncs pages every 24 hours through n8n, detects changes with MD5 checksums so only changed pages are re-embedded, stores vectors in PostgreSQL with pgvector, and answers with OpenAI and LangChain, citing sources. Ships as a Docker Compose stack.
Result: From roughly 30 minutes of searching to a 15-second answer with source links.

Overview

Ask a question in plain language and get an answer grounded in your Confluence pages, with links to the sources it used.

What I built

A Confluence client and document processor that chunks pages
Change detection with MD5 checksums so daily syncs stay cheap
pgvector similarity search inside PostgreSQL
A FastAPI query endpoint returning answers with citations

Read the full write-up in the blog post.

Read the write-up

Building a Confluence RAG Knowledge Base: Auto-Sync, Freshness Checks & AI-Powered Q&A
19 min read
––– views
May 12, 2026
Build a production-ready RAG system that automatically syncs Confluence pages, detects changes with MD5 checksums, stores vector embeddings in PostgreSQL (pgvector), and answers questions using OpenAI + LangChain.

More projects

Comments

Loading comments…

Edit this on GitHub ← Back to projects