Problem
Personal files are scattered across Drive, PDFs, docs, screenshots, and notes. Standard search depends on filenames and exact keywords, making it difficult to retrieve context or supporting evidence.
Case Study
DriveMind AI
Agentic RAG Knowledge Platform for Google Drive
DriveMind AI is a full-stack assistant that connects to Google Drive, indexes personal documents, and answers questions with grounded citations using hybrid retrieval, vector search, metadata filtering, and LangGraph-based workflows.
Solo Developer2026
FastAPINext.jsLangGraphLangChainPostgreSQLQdrantGoogle Drive APIOpenAI APIDocker
Overview
Problem, solution, and outcome.
Solution
DriveMind AI creates a personal knowledge layer over Google Drive. It syncs metadata, extracts content, chunks documents, generates embeddings, combines keyword and vector retrieval, reranks evidence, and generates source-backed answers.
Outcome
An LLM application demonstrating Drive integration, incremental indexing, hybrid retrieval, and workflow orchestration.
Workflow
Curriculum to course, step by step.
Implementation
Key capabilities.
1
Google Drive Integration
Read-only Drive connection for syncing metadata and supported document content.
2
Document Ingestion
Extracts and chunks Google Docs, PDFs, DOCX, and plain text files from Drive.
3
Hybrid Retrieval
Combines full-text search, vector search, metadata filtering, fusion, and reranking.
4
LangGraph Agent Workflow
Routes queries, grades evidence, rewrites weak retrieval, and verifies citations.
5
Source-Backed Answers
Responses include citations linked to retrieved document chunks.
6
Production-Oriented Backend
Built with FastAPI, PostgreSQL, Qdrant, Docker, and modular services.
Product
Interface previews.
Onboarding
First-run setup and Google Drive connection.
Welcome
EnlargeConnect Drive
EnlargeInitial state
EnlargeGrounded Q&A
Retrieval-backed answers with citations and follow-up suggestions.
Chat with citations
EnlargeKnowledge library & settings
File browser, indexing status, and Drive connection controls.
Knowledge library
EnlargeSettings
EnlargeSources & follow-ups
Citation chips and suggested next questions after a grounded answer.
Sources & follow-ups
EnlargeTechnical Notes
Architecture decisions.
Explore more work
View the repository or return to the homepage to explore other backend, automation, and full-stack work.