The Personal Knowledge Vault
summary
A minimal viable Personal Knowledge Vault that targets caregivers and people with functional memory loss uses Neo4j graph database as the context store, OpenAI/ChatGPT for the conversational layer and handwriting OCR (Omni), and miniIO S3 storage for private files. It ingests voice and handwritten notes, creates document nodes via Celery background jobs, runs vector search and React-style agent tools, and returns answers with traced sources; demo extracts handwriting accurately and retrieves notes. Surprising context: users install ~40 apps and workflows can require 350 switches across 22 systems.
AI-written summary · about
chapters (10)
- 0:00Project vision and context
- 1:43Speaker background and expertise
- 3:27The history of computing metaphors
- 6:11Modern complexity challenges
- 8:28The role of AI and agents
- 11:02System architecture overview
- 12:28Document ingestion demo
- 16:29Retrieval and search agent
- 18:35Agentic question and answer
- 20:58Research agent workflow
related talks
31mThe Personal Knowledge Vault: Art Collection Use Case
14mEmpowering Voice AI Agents with Real-Time Knowledge from Neo4j
40mContext That Dreams
21mBeyond Recall: Creating Agentic Personas with Persistent, Evolving Memory
40mTurn 10,994 Notes Into Memory
13mStrategies for Enhanced Personal Knowledge Management
21mUI in The Age of AI
32mBuilding an Advanced Knowledge Assistant