Skip to content

CHANGELOG.md

Latest commit

 

History

History
71 lines (60 loc) · 3.51 KB

CHANGELOG.md

File metadata and controls

71 lines (60 loc) · 3.51 KB
Raw

Changelog

All notable changes to this project are documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

[1.0.1] - 2026-08-06

Fixed

  • setup now copies AGENTS.md into the vault — the operating agreement is part of the scaffold (non-destructive; existing files are never overwritten). Previously a fresh vault had the folder skeleton but no contract, so the agent had no operating instructions until the owner remembered to drop the file in manually.

[1.0.0] - 2026-08-06

Added

  • brain CLI — one entry point for the whole engine: brain inbox, brain search <q>, brain doctor, brain dream, brain sync, brain help. Installed into ~/.local/bin by setup.
  • brain upgrade / brain uninstall — pull the latest and re-install in place (vault files never touched), or remove everything except the vault.
  • docs/SKILLS.md — every skill in depth: pipeline, triggers, CLI mapping.
  • docs/TROUBLESHOOTING.md — skills not loading, cron, search, PATH.
  • setup --embeddings — optional local venv with sentence-transformers for real semantic search.
  • Versioning contract — VERSION file, Semantic Versioning, git tags, and this changelog. Every release is now a tagged, reviewable, revertible snapshot of the whole stack.
  • docs/RELEASING.md — the release procedure: bump, changelog, tag, verify, publish. Releasing is a checklist, not an improvisation.
  • Architecture Decision Records — docs/adr/ with 8 ADRs capturing the why behind the stack: plain Markdown + git, typed memory, nightly consolidation, secrets outside the repo, the agent credential boundary, regenerable indexes, semantic versioning, and the agent (worker) architecture.
  • docs/REPLICATION.md — the replication playbook: how to stand up this system somewhere new (new machine, another person, a team) without copying any personal knowledge — the architecture replicates, the content does not.
  • Agent architecture layer — new agents skill (worker delegation, credential boundary, context modes) and 00-Hermes/ skeleton in the vault template (contexts/, decisions/, memory/, workers/), so the multi-agent operating model used in production is part of the replicable system, not a private detail.
  • setup --version — the installer reports the stack version.

Changed

  • ARCHITECTURE.md v2 — restructured around the full production system: three-layer pipeline, typed memory schema, retrieval (graph + semantic), NOW.md, agent architecture (workers, context modes, credential boundary), versioning and replication of the stack itself. The design is the product.
  • AGENTS.md template — added the credential boundary rule ("work flows upward, credentials do not flow downward") and worker/context-mode hooks, generalizing what was proven in production.
  • README — skill count 13 → 14, updated table, versioning and docs links.

Fixed

  • None.

[0.1.0] - 2026-08-02

Added

  • Initial brainstack release: 13 Hermes skills (capture, memory, retrieval, consolidate, reflect, meetings, signal, people-crm, health, trackers, reference, index, distill), the bin/ engine (memory compiler, semantic index/search, dream cycle, inbox processor, vault health, session consolidator, decision logger), a ready-made vault-template/, and the setup installer.