AquinLabs

aq changelog

What's new in Aquin (aq). Run aq version to see what you have installed.

  1. aq v0.0.3

    aquin.app/aq + install cutover

    Account, install, and release URLs live on aquin.app. Agent chat is axi (separate package).

    New

    • Account portal at aquin.app/login.
    • Install: curl -fsSL https://github.com/Aquinf03/aq/releases/latest/download/install.sh | bash

    Changes

    • aq login opens aquin.app/login desktop handoff.
    • In-repo web/ auth app removed.
    • Agent moved to axi.
  2. aq v0.0.2

    SDK + YAML + artifacts

    A run is recipe.yaml (path map) + artifacts/ (runtime). Author with YAML and/or aquin — both first-class. Same kernel for CLI.

    New

    • Python SDK (aquin): YAML-first Aquin(".") or SDK-first Aquin.define(...); also Run, checkpoint / serve / hash / status.
    • aq init / python -m aquin init scaffolds only recipe.yaml, example.py, and artifacts/.
    • Run identity: recipe.yaml alone is enough.

    Changes

    • Removed aq tool, aq stage, aq fork, aq job, aq schedule, aq checkout. Renamed aq spawn run → aq spawn agent.
    • Init no longer scaffolds experiment.md, data/, evals/, skills/, tools/, jobs/, stages/. Default folder: aq-run.

    Fixes

    • guard.safety warns on spikes and only stops after consecutive bad steps (blowup_patience / nan_patience).
  3. aq v0.0.1

    First cut (pre-SDK)

    Everything before the SDK / path-map reboot — collapsed from the old 0.0.1–0.0.4 line into one era.

    Install

    curl -fsSL https://github.com/Aquinf03/aq/releases/latest/download/install.sh | bash

    Node 18+, Python 3, npm. Windows (Git Bash + aq.cmd) supported. aq update reinstalls latest.

    What landed

    • Train / eval / serve for tabular, LLM, vision CNNs + ViT family, and VLM (CLIP / LLaVA-style / Flamingo).
    • Machine-aware compute, train + eval TUIs, weight export, aq plot.
    • Agent chat, spawn, providers; aq version / aq doctor.
    • Pre-SDK folder layout and verbs (aq job, aq schedule, aq stage, …) — superseded by 0.0.2.