Skip to content

Akasha Seeds — v1.2 (seeds12)

Latest
Latest

Choose a tag to compare

View all tags
@akashicmaster akashicmaster released this 01 Aug 05:52
v1.2
e8dd764
This commit was signed with the committer’s verified signature.
claude Claude
SSH Key Fingerprint: 32dP45eSMmVSt/G/CGvcxl/P+MO3Nwj9xeTh/GSA2wc
Verified
Learn about vigilant mode.

Akasha Seeds — v1.2.1(seeds12)

A local-first semantic knowledge system for researchers, thinkers, and anyone who wants their notes to do more than sit in a folder.

Two seeds from this release

The seeds series now ships as two variants with identical contents:

File Backend For
akasha_seeds_seeds12.py SQLite (default) General use — the recommended, production-tested build
akasha_seeds_seeds12n.py Silica (native) Benchmark reference — Akasha's own storage engine, a drop-in twin of the SQLite backend

Both expand to the same system; only the storage engine differs. Silica matches SQLite on correctness and crash recovery in current tests, with better isolated median latency but weaker concurrent throughput and tail latency. Benchmarks ship in test/benchmark.py.

What's new in v1.2

  • MCP service — connect an LLM client directly: python akasha.py --mcp (stdio) or the portal's /mcp endpoint (HTTP). 15 tools; anonymous callers get a read-only guest session, authenticated tokens unlock scoped writes.
  • Society — group dialogue spaces. Members speak through their own cast (avatar); spaces can be chat or cowork, with guest and public gates controlled by the creator.
  • nebula — a third explorer that surveys your accumulated atoms for dense regions of meaning: candidate concept models that emerged from use, proposed for you to confirm or discard.
  • Drop-in concept models — put a concept model file in lib/akasha/concepts/ and its commands appear in the CLI immediately. Fifteen food-domain models (cheese, wine, cocktail, …) ship as working examples; enable their data with onto.pack.enable wine.
  • Silica — a self-owned native storage engine implementing the AkashaBackend instruction set, groundwork for the hardware path.
  • Incremental ontology loading — release updates now load only what changed: re-extracting a seed over an existing install skips unchanged packs entirely, and a new or edited ontology file imports in seconds without touching the rest.
  • Process lifecycle hardening — SSH drops, overwrite installs, and crashed daemons self-heal on the next start; no manual process cleanup, ever.

Getting started

Requirements: Python 3.8 or later. macOS, Linux, or Windows.

# 1. Create an empty folder and place akasha_seeds_seeds12.py inside it, then:

python akasha_seeds_seeds12.py   # extracts into the current folder

python akasha.py                 # start (subsequent launches use this command)

On first launch, Akasha loads its built-in ontology and opens an interactive shell. Updating later is the same move: run the new seed in the same folder — your data is preserved, and only changed parts reload.

Verification

  • Full test suite: 73/73 evals pass.
  • SHA-256:
    • akasha_seeds_seeds12.py — 421c620f893713e48b312d6fa5f783f39966720438b270b78dc71e03d5648244
    • akasha_seeds_seeds12n.py — 2f35ab91442f7ac28b11f13a5d668285ace363bc72804c60e0bb4f0e264b62df

Maintenance update — v1.2.1 (2026-08-03)

  • Fixed: def name "description" silently dropped the quoted description; al could create
    dangling aliases that shadowed real atoms. If you used def/al on v1.2, re-check those
    entries — dangling aliases self-heal automatically, but lost def descriptions need re-entering.
  • New opt-in ontology pack: philosophy (notions, thinkers, repères, authors).
  • Process lifecycle: every entry path now converges to one healthy daemon + portal — wedged
    processes are detected (real-request probes + boot heartbeat), reaped, and respawned
    automatically; a slow large-ontology boot is never mistaken for a hang.

Documentation

License

MIT (software and editorial content) · WordNet 3.1 (base ontology — Princeton University) · See LICENSE.md for full attribution and third-party notices.

Akasha Seeds is the independent researcher and humanities distribution. Client limit and bundled content differ across the Seeds, Thesaurus, and Server series.

Assets 4
Loading