Skip to content

Latest commit

 

History

History
37 lines (27 loc) · 1.57 KB

File metadata and controls

37 lines (27 loc) · 1.57 KB

Obsidian Mind: adoption decision

Date: 2026-07-28

Verified upstream design

The upstream repository combines three layers: a plain Markdown/Obsidian vault, deterministic lifecycle hooks, and an optional QMD semantic-search layer. Its manifest is the coordination point for index naming, infrastructure boundaries, content roots, and frontmatter requirements. The architecture also keeps eager session context under a byte budget and treats QMD as a fallback rather than a hard dependency.

Source: https://github.com/breferrari/obsidian-mind and its ARCHITECTURE.md and .codex/hooks.json files.

Decision for this harness

Adopt the portable ideas, not a wholesale clone:

  • Git remains the source of truth for durable work;
  • the private chat archive may expose an Obsidian-compatible Home.md and manifest as a navigation view;
  • evidence links and privacy boundaries stay enforced by the existing archive writer and knowledge-first search;
  • QMD/MCP stays opt-in until a measured benchmark shows a real retrieval gain.

This keeps the open harness free of private conversations and prevents a second stale memory store. The live private archive contains the bridge implementation; this public note documents the reusable decision and its proof boundary.

Proof boundary

The bridge is accepted only after tests show that it writes navigation metadata, does not copy raw JSONL or secrets, and is byte-idempotent for the same input. The absence of obsidian and qmd on the current machine means GUI behavior and semantic-search quality remain unverified; neither is claimed as enabled.