Official Submission: Anamnesis is Live!
Excited to announce that Anamnesis is officially submitted for Build with DataHub: The Agent Hackathon 2026!
What is Anamnesis?
A persistent memory layer for AI data agents, built natively on DataHub. Instead of reasoning in isolation every time a pipeline breaks, Anamnesis allows AI agents to search, recall, and inherit verified fix patterns across similar schema contract anomalies.
Key Features:
- 5-Stage Pipeline: Detect -> Diagnose -> Recall -> Fix -> Write
- Semantic Vector Recall: Matches exact down to 60-70% related issue signatures to adapt inherited fixes instantly.
- Auto-Indexing New Anomalies: Zero-shot resolves brand-new breaks and indexes them as new memory nodes in real time.
- Interactive 3D Constellation: Live memory graph visualization powered by Three.js.
- DataHub MCP Server Integration: Direct lineage traversal and custom IncidentMemory aspect writebacks.
Check out our live demo, source code, and walk-through video on the project page! Feedback and thoughts are highly welcomed in the comments below.
Log in or sign up for Devpost to join the conversation.