Slack bot for answering questions from Confluence using Amazon Bedrock

This article details how to build a Slack bot that answers questions from Confluence using Amazon Bedrock. It extends a previous setup with Amazon EKS, LiteLLM, and OpenTofu. The solution uses an Amazon Bedrock Knowledge Base to crawl a Confluence space, embed pages with Amazon Titan, and store vectors in Amazon OpenSearch Serverless. LiteLLM performs server-side RAG, retrieving relevant chunks for each question and feeding them to a Bedrock model. Collmbo serves as the Slack bot front-end, communicating via Socket Mode. The architecture includes AWS Secrets Manager for credentials, Arize Phoenix for OpenTelemetry tracing, and all infrastructure managed by OpenTofu. The guide covers setup criteria, architecture diagram, and deployment steps.