SDK 0.7.3a4 · documentation review 2026-11-14.
Runtime/source: 49578688710a8bae2373aaf9d47ff9521f566d4c.
Exact-source EC2 CI: success. Release publishes the unchanged
wheel/source bytes; later documentation commits are separate from that runtime.
Environment and checks
Linux x86-64/Python 3.12 CPU on 2 vCPU/8 GiB. The full installed-media container was limited to 2 CPU/4 GiB, non-root, read-only and network-disabled, with temporary storage. A configured limit is not measured peak consumption.
| Check | Result | Boundary |
|---|---|---|
| Full installed-media | 341 passed, 0 skipped, 410.349 s | Suite duration, not processing throughput |
| Python 3.11–3.14 core | Each:341 discovered,248 executed,93 expected optional skips | Media extras run separately in the complete suite |
| Payload parity | 70 files | Qualified source, wheel, sdist and installed bytes |
| Fresh video/media wheel installs | Online setup and offline image/video checks passed | Video-only omits Torch/NumPy/Python TrustMark |
| HTTPS deployment kit | Exports, signed reports, recovery, feedback, restart/isolation passed | Local test credentials |
| Official c2patool0.28.6 | Image and all3video modes:signature and asset hash validated | Dev issuer intentionally untrusted |
| Authenticated TSA | Real native RFC3161 with offline HTTPS fixture | Not live SSL.com enrollment |
| Explicit verification scope | Trusted claims retain all/custom legacy obligations | Missing decoders remain incomplete |
Artifact identity
| Artifact | SHA-256 | Bytes |
|---|---|---|
ariadne_mark-0.7.3a4-py3-none-any.whl |
97a83d5edf47f63c990505f9eb76d9214a431a2e535455759afc36eb86fa2793 |
162713 |
ariadne_mark-0.7.3a4.tar.gz |
3635e88f1d5653af6444a94eb415da40f3b908b62707f7b9cb39a558aeeb0e9a |
459580 |
Machine-readable summary and artifact metadata retain exact identities. The archived README describes its candidate build stage; public instructions were subsequently updated without rebuilding these artifacts.
Scope and remaining work
Economy5s remains default; Economy2s and Resilient remain selectable. Cadence, strength, payload and read-back gates were not changed. Deterministic fixtures and passing tests are not population durability or statutory safe harbors. Short window-free excerpts, crop/letterbox/remover behavior and negative-scan cost remain documented limitations. Historical benchmarks keep their own versions.
The prior 0.7.3a3 external review gave a scoped developer-alpha GO; this version addresses its follow-ups and has exact-version CI. No independent human audit or product-wide compliance certificate is claimed. Production credentials and public resolver discovery remain separate.