Repository navigation
Releases: jacksonpradolima/coleman
Release list
COLEMAN 4.19.1 — A connected guide to adaptive CI
Go from a CI test history to a working experiment, inspect its results and follow the evidence into your next decision. This release makes the platform easier to learn and its public contracts easier to use.
Highlights
- Follow a complete Random/UCB tutorial with result queries and checkpoint continuation.
- Navigate connected guides for policies, datasets, evaluation, configurable systems and integration, with research citations beside the relevant methods.
- Inspect NumPy-style parameters at selected public entry points and reject non-finite budgets before execution.
Compatibility
Requires Python 3.14 or later. Existing artifacts retain their meanings. Producer identity changes when implementation bytes change; the parameter audit covers selected public entry points rather than the entire SDK.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.19.1/coleman-4.19.1-py3-none-any.whl#sha256=ca470ba2be14ca755952f0bb3bc92615d76b90f95ebdd41dc1cdc9847713cc93"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.19.0 — Verify original CI evidence
Connect original CTest logs to typed, source-bound observations before using them in a test-prioritization experiment.
Highlights
- Keep pass, failure, timeout, exception, skipped and not-run outcomes distinct.
- Verify original archive bytes, pipeline/job identity and the correspondence between LibSSH CSV observations and CTest logs.
- Resume collection with immutable evidence and completion receipts; contradictory evidence cannot become a completed export.
Compatibility
Requires Python 3.14 or later. Native results and checkpoints keep their contracts. Source corroboration does not establish historical predecision availability or policy effectiveness.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.19.0/coleman-4.19.0-py3-none-any.whl#sha256=070e21c19fb7e3f2e392bd68706c354ed5ab7f76b86db80f5f1bdc5721b767bd"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.18.0 — Industrial test data and bounded comparisons
Explore original industrial test observations and compare bounded CI metrics with conservative uncertainty under explicit independence assumptions.
Highlights
- Convert pinned Westermo sources into separate observation and session Parquet tables, preserving abnormal outcomes and unverified duration units.
- Use bounded paired means with full-family correction, simultaneous bounds and practical-equivalence decisions.
- Recover exports without treating partial writes as completed data.
Compatibility
Requires Python 3.14 or later. Existing Student-t protocols and native artifacts keep their meanings. The new inference method requires the matching SDK; neither source conversion nor statistical machinery supplies missing independent evidence.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.18.0/coleman-4.18.0-py3-none-any.whl#sha256=ae75cd0f7eefd91be7bfc2b44b43114ea97e02588588803b17bb3d6433c79d16"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.17.0 — Comparable CI baselines with inspectable evidence
Run a frozen baseline matrix through the native TCPCI runner, inspect candidate exclusions and retain the evidence from interrupted or incomplete experiments.
Highlights
- Freeze source, development fold, candidates, seeds, budgets and cooperative limits in a versioned baseline plan.
- Compare native policies and history-based heuristics under one declared protocol, retaining rankings, selected prefixes, costs and quality observations.
- Verify completed artifacts and resume without silently replacing failed or aborted attempts.
Compatibility
Requires Python 3.14 or later. Reuses RunSpec, RNFail, Parquet and native checkpoints. Candidate compatibility remains explicit; a successful matrix run alone does not establish cross-project effectiveness.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.17.0/coleman-4.17.0-py3-none-any.whl#sha256=e5a272ddc0edf72bd909a743fa2b3bc0dfeb8a02871c487ab22683cddeea40ca"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.16.0 — Compare projects without overstating repetitions
Connect repeated native CI experiments to protected comparisons while keeping builds, seeds and independent repository families distinct.
Highlights
- Freeze complete build/seed membership, metric semantics, source clocks and repository-family identity.
- Retain completed, failed, aborted and skipped attempts with immutable raw result keys.
- Prepare equal-seed and equal-build means before paired family comparisons, preserving valid zero and negative values.
Compatibility
Requires Python 3.14 or later. Existing native metrics keep their definitions. Seeds and repeated builds are not substituted for independent observations; incomplete membership remains visible.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.16.0/coleman-4.16.0-py3-none-any.whl#sha256=73fac37fc037d4e5c061f838711371ade5ada2f129917f09ef1b49ce87350116"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.15.0 — Know what your benchmark can support
Audit source evidence before comparing CI policies: inspect units, availability clocks, execution groups and the limits of each converted dataset.
Highlights
- Use source-admission reports and a 35-axis production-reality matrix to inspect what original evidence supports.
- Preserve original producer bindings and verify source-to-artifact provenance.
- Resume qualification without rewriting completed evidence, and retain an explicit insufficient-evidence outcome.
Compatibility
Requires Python 3.14 or later. Admission evaluates evidence rather than manufacturing it. A runnable conversion or replay remains distinct from protected historical evaluation and production adoption.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.15.0/coleman-4.15.0-py3-none-any.whl#sha256=ad2daedb4c251d5ac001db7d73910d912d439629aea4e0a04d84dcc66e279fde"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.14.0 — A reproducible CI stress laboratory
Examine difficult CI decisions with deterministic cases whose truth, feedback timing and accounting are explicit.
Highlights
- Exercise eleven declared factors, including rare failures, drift, churn, flakiness, costs, budgets, context and variants.
- Use 32 deterministic interaction-covering cases, including a 50,000-test scale case.
- Reuse native Parquet, policies, rewards and resumable agents instead of introducing a second experiment runner.
Compatibility
Requires Python 3.14 or later. Constructed stress cases test mechanisms and declared interactions. They are not independent industrial observations or evidence that a policy wins in production.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.14.0/coleman-4.14.0-py3-none-any.whl#sha256=de46ef9482f45372c96dd3d402f85daa18c3079733b4008ec05885e0d2d5cf66"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.13.0 — Configuration histories from source to replay
Run a bounded original LibSSH configuration history through whole-test-set and variant-test-set strategies while retaining source identity.
Highlights
- Prepare variant-aware analytical and native Parquet from pinned CSV/JSON source data.
- Inspect pipeline, job, variant and attempt identity, including missing or unusable observations.
- Resume policies across multiple budgets without one budget overwriting another budget's checkpoint.
Compatibility
Requires Python 3.14 or later. Keeps native WTS/VTS behavior and result semantics. Source chronology and scientific-admission gaps remain explicit; replay is not a production-savings measurement.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.13.0/coleman-4.13.0-py3-none-any.whl#sha256=bbe2f3cf8954fc27fc8f5a44e3713519aa30bcf2cfdd2aad140f4bcc1ae45abc"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.12.0 — Inspect constructed C faults with an oracle
Extend controlled CI research to a bounded C source workflow, with pinned compilation and an independent arithmetic oracle.
Highlights
- Capture original NIST Juliet functions in a declared GCC container profile.
- Retain raw process evidence, observer context and distinctions between expected arithmetic behavior and observed execution.
- Reopen completed observations offline and recover interrupted collection without inventing successful tests.
Compatibility
Requires Python 3.14 or later. The producer is optional and leaves native TCPCI workflows unchanged. These are constructed weaknesses, not historical industrial fault observations.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.12.0/coleman-4.12.0-py3-none-any.whl#sha256=5802f83dfa4608a979f132e6e00ea3d8a128c88f4b7bbde07afcd416c33df55e"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.
COLEMAN 4.11.0 — Reproduce Python faults in their execution context
Keep source fixes, runtime conditions and observed outcomes together when reproducing controlled Python faults.
Highlights
- Capture unchanged buggy/fixed PySnooper trees using pinned wheels and declared CPython images.
- Compare frozen ASCII and UTF-8 locale profiles with a common upstream fixture.
- Retain raw evidence and support offline validation and interruption recovery.
Compatibility
Requires Python 3.14 or later. Uses the existing controlled-observation contracts without changing native learning checkpoints. Reproduction conditions and negative outcomes remain evidence; they do not establish benchmark admission.
Get started
Install this release's attached wheel in a virtual environment:
python -m pip install "https://github.com/jacksonpradolima/coleman/releases/download/v4.11.0/coleman-4.11.0-py3-none-any.whl#sha256=d328c0d9bfd9e75658c64e35876c3151fb242c805846835c211a3622d8b75e55"The assets also include the source archive, Sigstore verification bundles and an SBOM. The dependency lock is attached as well.