ReproMin v0.1.0.dev9: installable pilot for failure-preserving reduction

[fly1d]

ReproMin v0.1.0.dev9 is available as a GitHub pre-release: https://github.com/fly1d/repomin/releases/tag/v0.1.0.dev9

ReproMin reduces a failing repository while repeatedly checking the configured failure contract, then exports a smaller payload with an auditable report. This release includes:

• repomin doctor for a read-only toolchain and baseline preflight
• repomin report validate and fresh-copy repomin report replay
• exact and transport-friendly payload fingerprints
• Maven, Gradle, Python, Node, Cargo, Go, Ruby, Composer, and MSBuild adapters
• a reusable GitHub Action with privacy-safe report outputs and step summaries
• repomin report compare for read-only, descriptive comparison of two or more validated reports

Try a pilot

Install the wheel in an isolated environment, then follow the real-failure pilot guide. The shortest useful path is doctor -> one reduction -> report validate; run report replay only after reviewing the unsigned report and command. When a workflow is repeated, report compare can show aggregate evidence and context drift without reading payloads, executing commands, or accessing the network.

We are looking for sanitized CI or dependency-failure workflows, small public fixtures, and reports of successful, inconclusive, or blocked trials. Please use Issue #11 for a real failure or the user workflow feedback template for value and friction. Share only the language/build system, runner, options, aggregate sizes, limitations, and sanitized evidence. Do not post credentials, private URLs, proprietary source, raw logs, commands, or environment values.

The project is pre-alpha. A passing oracle is evidence for one configured failure in one recorded environment, not a correctness, causal, production-reliability, or security guarantee. The host backend executes commands directly; Docker is defense in depth rather than a complete sandbox.

[fly1d]

Maintainer update (2026-09-01): ReproMin v0.1.0.dev7 is available as a GitHub pre-release: https://github.com/fly1d/repomin/releases/tag/v0.1.0.dev7

The release adds privacy-safe adoption evidence through repomin report validate --json, a workflow-feedback issue template for successful/inconclusive/blocked trials, and scalar GitHub Action outputs for oracle mode, retention ratios, and payload-fingerprint evidence. Replay mismatch evidence now includes expected and actual exit codes while omitting commands, expressions, logs, and environment values.

Assets passed twine check, isolated wheel/sdist installation, 560 tests, and 16/16 offline benchmark checks. Tag CI is running at https://github.com/fly1d/repomin/actions/runs/33466996058. Please use the workflow feedback template for any trial, including a blocked or inconclusive one; share only sanitized details.

[fly1d]

Follow-up (2026-09-01): tag CI for v0.1.0.dev7 has completed successfully across the Python/JDK matrix, Windows/macOS/Linux, Action smoke, offline benchmarks, and isolated wheel/source-distribution checks: https://github.com/fly1d/repomin/actions/runs/33466996058

[fly1d]

Maintainer update (2026-09-01): ReproMin v0.1.0.dev8 is published as a GitHub pre-release: https://github.com/fly1d/repomin/releases/tag/v0.1.0.dev8\n\nThe release focuses on shareable evidence: repomin report validate --format markdown produces a deterministic privacy-safe summary, and the GitHub Action can optionally append it to the runner Step Summary with a validated run link. The summary schema is now v2, with commands, expressions, logs, paths, and environment values excluded.\n\nVerification is complete on tag CI (Python/JDK/Windows/macOS/Linux matrix, Action smoke, offline benchmarks, coverage, and isolated wheel/sdist checks): https://github.com/fly1d/repomin/actions/runs/33474516876\n\nWe are looking for the next real-world signal: a sanitized CI or dependency-failure workflow, a confusing onboarding step, or a small fixture that would make the tool more useful. Please use the pilot guide and do not share credentials, proprietary source, private URLs, or raw logs. Results are oracle evidence for the current environment, not a correctness or security guarantee.

[fly1d]

Maintainer update (2026-09-01): ReproMin v0.1.0.dev9 is published as a GitHub pre-release: https://github.com/fly1d/repomin/releases/tag/v0.1.0.dev9

This release adds repomin report compare for a privacy-safe, read-only comparison of validated reduction reports. The output is intentionally descriptive: aggregate sizes, retention, attempts/mutations, cache and budget/holdout state, phase coverage, adjacent numeric deltas, and warnings for context drift. It does not inspect payloads, execute report commands, access the network, or claim speed, correctness, causality, or production reliability.

If you have a sanitized CI or dependency-failure workflow, please install the wheel or source archive from the release page and share a short pilot report in Issue #11. A small fixture, an inconclusive run, or an onboarding problem is useful too. Please omit credentials, private URLs, proprietary source, raw logs, and environment values; the host backend is not a sandbox.

[fly1d]

Follow-up (2026-09-01): the v0.1.0.dev9 tag CI has now completed successfully after a failed Windows-only attempt caused by runner startup timing. The rerun passed the Python 3.9/3.13 and JDK 11/21 matrix, Windows/macOS/Linux, GitHub Action smoke test, offline benchmarks, packaging, and isolated wheel/source-distribution checks: https://github.com/fly1d/repomin/actions/runs/33486832842

No source changes were needed; the initial failure was a timing-sensitive Windows test startup. The release remains a pre-alpha, descriptive-only pilot build.

[fly1d]

For contributors looking for a small first task, Issue #19 is now open: add the existing FastAPI/Docker fixture to the runnable examples guide. It is documentation-only, has explicit acceptance criteria, and helps Python users try the dev9 pilot without changing reducer behavior.