Browse / Second brain / voidbrain

voidbrain

moshehbenavraham

Local-first AI second-brain scaffold for Obsidian-style markdown vaults

Open moshehbenavraham/voidbrain → Checked against the four rules.

Works with any agent

ships AGENTS.md, the cross-tool instruction file.

In its own words

README.mdmoshehbenavraham/voidbrain

Voidbrain is a local-first AI second-brain scaffold for Obsidian-style markdown vaults. It is currently an MVP scaffold with provider privacy gates, local indexing, cited chat, source ingestion staging, staged review/apply, vault health reporting, hot cache recovery, and a local validation toolchain.

What It Is

  • Obsidian-compatible plugin scaffold built with TypeScript, Svelte, Vite, and

Vitest.

  • Spec-driven product plan under .spec_system/.
  • Local-first design: vault markdown remains the durable source of truth.
  • Safety baseline for provider privacy, staged writes, citations, indexing,

hot cache recovery, and agent-readable command surfaces.

Development

Agent documentation checks:

Local Obsidian install/update preview:

The deploy workflow builds and validates the plugin bundle, then copies only main.js, manifest.json, styles.css, and versions.json into .obsidian/plugins/voidbrain. Dry run prints the install/update plan without building, copying, cleaning, backing up, or mutating vault files. See Obsidian Install And Update Workflow.

Provider setup starts local-first. Review Provider Readiness Guide, configure a local runtime or OpenAI-compatible local endpoint first, and move to custom remote or trusted cloud providers only after provider review, trust, auth, capability, and disclosure gates pass. Untrusted cloud providers remain blocked for private vault content.

Quoting 31% of the README. Read the rest on GitHub ↗

Read from the repository's git tree on 2026-08-17.