I worked on an AI-powered knowledge retrieval system designed to help users search and interact with information stored across business documents. The workflow combines automated document ingestion, text extraction, vector embeddings, and an AI chat interface to make document-based information easier to access.

Automated document ingestion from Microsoft OneDrive.

Document processing workflows for PDF, Excel, CSV, and text files.

Text extraction and document metadata management.

Supabase database integration for storing document records and supporting retrieval.

OpenAI embeddings and a Supabase vector store for semantic search.

An AI chat agent using an Anthropic chat model to answer questions using retrieved document context.

Retrieval tools for listing documents, retrieving file contents, and querying document records.

Conversation memory and webhook-based responses for the chat interface.

Database setup workflows for document tables, metadata, and retrieval functions.

n8n, Supabase, PostgreSQL, OpenAI Embeddings, Anthropic, Microsoft OneDrive, RAG, AI Agents, Vector Search, API Integration.

This architecture can help businesses make internal knowledge more accessible, reduce manual document searches, and create a foundation for AI assistants that answer questions using their own documents.

The system is designed around modular workflows so document ingestion, retrieval, and chat interactions can be managed independently.

AI workflow automation, document processing, retrieval-augmented generation, database integration, and connecting AI models with business data.