Skip to content

Releases

interaction-check 0.1.0b2

Pre-release

Choose a tag to compare

@PerchAnton PerchAnton released this 04 Nov 23:22

This beta preserves Unicode in JSON and JUnit reports when a terminal uses ASCII or Windows-1252. For example, a scenario named Україна can complete and write its report in either encoding.

Install python -m pip install interaction-check==0.1.0b2. A clean PyPI install passed all six console cases and all three jsonpatch 1.39 scenarios (nine steps). The attached wheel and source distribution match the PyPI downloads byte for byte.

Package on PyPI · Source commit.

interaction-check 0.1.0b1

Pre-release

Choose a tag to compare

@PerchAnton PerchAnton released this 01 Nov 13:17
e53fd2c

Published to PyPI on 2026-11-01.

This beta documents the CLI and JSON compatibility boundary. JSON reports now identify schema: 1 and include failure_step for state mismatches. Existing scenario/adapter protocols, Mermaid examples and native-test comparisons are retained.

Published from e53fd2c77d45dc6b57dded7c089c64c78e00c24f through Trusted Publishing. The attached wheel and sdist are the exact Actions archives; both SHA-256 values match PyPI downloads. Fresh PyPI installation passed quickstart native/replay checks, the python-dispatch example, and deliberate mismatch/export/rerun checks including failure_step.

This remains a prerelease. It does not replace host-native tests or establish external adoption or shared-CI approval.

interaction-check 0.1.0a8 — readable property traces

Choose a tag to compare

@PerchAnton PerchAnton released this 28 Oct 08:19
1cda979

interaction-check 0.1.0a8 — readable property traces

This alpha packages the real python-dispatch walkthrough and optional expected/observed
Mermaid reports developed from maintainer feedback. Runner behavior is unchanged from
alpha.7; only its version identifier changes. Python 3.10+, zero runtime dependencies,
scenario schema 1 and JSONL protocol 1 remain.

  • Distinguish requested property writes, new callbacks, the cumulative callback log and
    the current property value. Equal writes, emission locks and unbinding are demonstrated.
  • Read separate expected/observed tables and sequence diagrams. A deliberately wrong
    expectation demonstrates the first mismatch; this is not a newly found upstream bug.
  • Understand the actual rat-salsa method calls, adapter costs and limits around rendering,
    focus and clipboard. Prefer ordinary native tests for a small regression; the existing
    pyee #228 is linked rather than duplicated.
  • Source archives include both walkthroughs. The release workflow verifies their contents
    after extraction; the wheel remains the standalone runner.
python -m pip install 'interaction-check==0.1.0a8'
interaction-check --version

From the attached source archive or tagged checkout, in a virtual environment:

python -m pip install -r examples/dispatch/requirements.txt
python examples/dispatch/verify.py --check

The renderer is optional, adapter-specific source code, not a new installed CLI flag.
It summarizes checked steps; the adapter may already have executed the full batch.
No intra-step timing is inferred. Native tests can remain independent of this runner.

Validation: 54 unit tests; the quickstart's three native tests and 12 replay steps;
three real python-dispatch native tests, five replay steps and intentional failure at
step 2; package metadata, fresh installation and packaged-example checks. CI covers
Linux/macOS/Windows and the existing PhotoCraft, pyee and XState checks.

This remains an experimental alpha. The GitHub wheel/source assets are the same files
uploaded to PyPI. Verify downloads against SHA256SUMS.

PyPI ·
Expected and observed diagrams ·
First mismatch ·
Changelog

Release commit: 1cda979491ba9af9b6a1de5c59cae4aaaa899aa3. CI, replay evidence, and PyPI publication passed.

Published and verified on 28 October 2026. Registry hashes match these assets. A fresh PyPI installation passed the quickstart, deliberate Cancel failure/export/replay, and the real dispatch scenario with its deliberate mismatch. See verification.json.

interaction-check 0.1.0a7 — runnable first example

Choose a tag to compare

@PerchAnton PerchAnton released this 27 Oct 07:12
a77c1e3

interaction-check 0.1.0a7 — runnable first example

This alpha packages the new first example and clearer documentation developed from
maintainer feedback. Runner behavior and CLI options are unchanged from 0.1.0a6.
Scenario schema 1, JSONL protocol 1, Python 3.10+ and zero runtime dependencies remain.

  • Run a small title editor with three ordinary Python tests and three replay cases
    (12 steps). A deliberate Cancel defect is introduced only in a temporary copy;
    both methods detect it. Export preserves the original expectations and reruns with
    the same mismatch.
  • Understand what the adapter does, when a native test is simpler, and how to share
    one actionable failure. A checked prefix does not mean execution stopped there.
  • Reproduce the README diagram from actual results and source hashes. The release
    workflow verifies the example and diagram from the packaged source archive too.

Install in a virtual environment:

python -m pip install 'interaction-check==0.1.0a7'
interaction-check --version

The wheel installs the runner. The attached source archive also contains the first
example; after extracting it, run python examples/quickstart/verify.py from its root.
The printed deliberate FAIL is expected; a successful overall demonstration exits 0.

Validation covers 49 runner tests, the native/replay demonstration, installed CLI
export/replay, package metadata and source-archive example checks. The full CI matrix
also exercises Linux/macOS/Windows and the existing PhotoCraft, pyee and XState checks.
Teaching-model results are not a new upstream bug, extra coverage or a speedup claim.

This remains an experimental alpha. No update is required merely for new runtime
behavior; there is none in this release. Pin the exact version and your adapter revision.

PyPI ·
Changelog ·
First example

interaction-check v0.1.0-alpha.5 — failure exports and installable CLI

Choose a tag to compare

@PerchAnton PerchAnton released this 26 Oct 10:46

interaction-check v0.1.0-alpha.5

This experimental release adds a portable failure handoff and an installable CLI.

  • --failure-dir saves each checked scenario prefix through its first state mismatch, preserving expectations and metadata. Existing output directories are refused; reports link to exports.
  • A native-test guide demonstrates transferring pyee/XState behavior into the consumer's normal tests. No permanent runner dependency is required by those native tests.
  • Python 3.10+ wheel and source distribution, zero runtime dependencies, interaction-check command and standalone --version.

Python metadata and JSON report version use 0.1.0a5; the tag uses v0.1.0-alpha.5. Schema 1 and JSONL protocol 1 are unchanged. Exports are prefixes, not automatically minimized or rerun reproductions. Adapter crashes/protocol errors do not produce guessed assertion scenarios.

Validation: 49 Python tests, strict package metadata checks, installation in an isolated venv outside the checkout, successful replay and replay of an exported mismatch. CI verifies the CLI/package on Linux, Windows and macOS; pyee and XState evidence checks remain separate.

Install the attached wheel in a Python virtual environment:

python -m pip install ./interaction_check-0.1.0a5-py3-none-any.whl
interaction-check --version

The wheel contains the runner and licenses; adapters/examples are in the repository. This is a GitHub prerelease, not a PyPI upload or upstream adoption. Verify downloaded assets against SHA256SUMS. Alpha CLI/report APIs may change.

Changes · Native-test guide