Show HN: Open-Source DaiDocs now beats Graphify by 218x low tokens, 83% on LME

Category: ai-ml

Tags: memory, ai, local-first, plain-text, mcp, code-graph

Score: 7.0/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 6)

DaiDocs introduces a plain-text, local-first memory format (.dai) for AI conversation history and a code graph format (.cai), enabling models to retrieve only relevant slices of memory with far fewer tokens. It claims significant token efficiency and benchmark results, and provides a router to decide which memory tier answers a query. The project is interesting for its openness, reproducibility, and potential to standardize AI memory.

Target audience: AI developers, backend devs, devops, data engineers

Repository: https://github.com/Kerneta/daidocs · Python · Apache-2.0 · 53 stars

View on Hacker News