v1.0.0 - MVP: End-to-End RAG Chatbot

What's Included

• Full RAG pipeline: parse, chunk, embed, store, search, generate
• 7 document parsers: PDF, DOCX, XLSX, PPTX, HTML, CSV, Markdown
• Token-aware recursive chunking with boundary detection (tiktoken)
• pgvector HNSW cosine similarity search on Aurora Serverless v2
• Claude 3.5 Sonnet answer generation with source citations
• Microsoft Teams bot (personal, team, group chat)
• SharePoint auto-sync via EventBridge every 6 hours
• Modular Terraform IaC (3 modules, S3 remote state)
• Structured CloudWatch JSON logging
• Interactive CLI for local RAG testing
• Docker Compose local dev (pgvector on PostgreSQL 16)

Architecture

Teams -> Azure Bot Service -> API Gateway -> Bot Lambda
  -> Titan embedding -> pgvector search -> Claude generation -> cited answer

Quick Start

pip install -e '.[dev]'
make dev-db && make setup-db
make ingest ARGS='--source data/sample_docs'
make query