Skip to content

v0.17.0

Latest
Latest

Choose a tag to compare

View all tags
@abrichr abrichr released this 17 Oct 02:05
v0.17.0
03fa181

v0.17.0 (2026-10-17)

Chores

  • Ignore every .env variant, not one at a time (#75, 7916fcc)

.gitignore covered .env and .env.backup but not .env.bak, .env.local or .env.old. A .env.bak containing populated OpenAI, Anthropic and Google keys sat in a local stash since 2026-07-13, untracked and one git add -A away from being committed.

Replace the one-at-a-time entries with a glob and negate the committed .env.example template.

Replace source-policy.public.json with the render of the canonical policy on openadapt-internal main (policy_last_updated 2026-09-03).

Co-authored-by: Claude Opus 5.5 noreply@anthropic.com

  • ml: Harden training outcome validation (1f32034)

Reject ambiguous or unmeasured training outcomes. Require valid evaluator evidence, optimizer updates, and checkpoints. Keep the release version unchanged.

  • policy: Sync generated source boundary (#77, cab99e8)

Continuous Integration

  • Pin pypa/gh-action-pypi-publish to a commit SHA (#74, 714a8f4)

This workflow referenced the publish action by the floating release/v1 tag. Every other release pipeline in the org pins a full 40-character commit SHA, so this one could change behaviour with no commit in this repo -- including a change that breaks publishing.

That risk is concrete. v1.14.0 bundles twine 6.1.0 and packaging 25.0, which reject Metadata-Version 2.5 -- the version current hatchling emits:

InvalidDistribution: Invalid distribution metadata:

'2.5' is not a valid metadata version

That pin broke the openadapt-evals 0.91.0 release. The tag, the version commit and the GitHub release all landed, then "Publish to PyPI" failed, leaving PyPI stale while every other artifact said the version shipped.

Pin v1.14.2 (twine 7.0.0 + packaging 26.2), the version already proven by live uploads of openadapt-privacy 1.0.3 and openadapt-types 0.10.1.

Co-authored-by: Claude Opus 5 noreply@anthropic.com

Documentation

  • Lead the README with research, not VLM-agent training (#79, 8811400)

The IMPORTANT banner opened with "trains and runs vision-language model agents". That is this package. It is not the OpenAdapt product.

  • Rewrite README for a human reader (#78, 199df05)
  • docs: rewrite README for a human reader

The README opened with a fifteen-line callout that spent most of its words describing openadapt-flow, so anyone who arrived for the training code read two paragraphs about a demonstration compiler first. That point is true and worth keeping, but it is one sentence, not a wall.

Three documented APIs did not exist. Checked against openadapt-ml 0.16.3 installed from PyPI into a clean venv:

  • AgentPolicy.predict(sample) is not a method. The real one is predict_action_from_sample(sample), and it returns a 4-tuple of (Action, thought, state, raw_text) rather than an object with .action and .thought attributes. - QwenVLAdapter(model_name=...) does not load a model. The constructor takes an already-loaded model and processor; the loading path is QwenVLAdapter.from_pretrained(model_name). - The smoke test cannot run on the core install. DummyAdapter raises ImportError: torch is required unless the training extra is present.

Every example is now a command or a snippet that was executed against that venv, shown with its real output. The benchmark table gains back the episode success column, which is 0% for all four configurations and was the most important number missing from the page.

The architecture tree and the ecosystem table move to docs/repo_layout.md. Nothing was deleted.

Co-Authored-By: Claude Opus 5 noreply@anthropic.com

  • docs: restore the research banner and fix four claims

AGENTS.md requires a public repo outside the seven product targets to carry a lifecycle state, and records an incident where deleting this banner cost four repos of wasted agent effort. The rewrite removed it. Put it back above the badges.

The Set-of-Marks paragraph introduced a login result and then cited registration_som_eval.json. That artifact is the registration scenario: 384 steps over 32 episodes is 12 steps each, and synthetic.py documents registration as 12 steps and 6 elements. Name the scenario and correct three elements to six.

The n=4 hardened re-runs are under 8b_hero/eval/, not 2b_dev/eval/, whose two files are both n=32.

There is no Azure trainer. cloud/ has train subcommands for lambda_labs, modal_cloud, vast_ai, and local; azure_inference.py is an async inference queue. Drop Azure from the training list and say what it is.

The Action line cut 16 of 19 fields with no ellipsis, so it read as though Action has three. Mark the cut and give the count.

Training reads --config paths relative to the checkout, and tool.uv.sources points openadapt-capture at ../openadapt-capture, so a fresh clone needs UV_NO_SOURCES=1, which is what test.yml sets. Give the clone and the flag.

  • docs: repo_layout listed an Azure trainer that does not exist

Same correction as the README. cloud/ has train subcommands for lambda_labs, modal_cloud, vast_ai, and local; azure_inference.py is an inference queue.

Co-authored-by: Claude Opus 5 noreply@anthropic.com

  • readme: Shorten the research banner to two lines (#81, cc4b7b6)

Signed-off-by: Richard Abrich richard.abrich@mldsai.com

Features

  • cloud: Kaggle GPU training path and a guard against Modal GPU on production (#82, 69c74e3)

Kaggle GPU training path; Modal GPU train/serve refuse on the production workspace without an explicit non-prod opt-in.

Signed-off-by: Richard Abrich richard.abrich@mldsai.com

Detailed Changes: v0.16.3...v0.17.0

Assets 6
Loading