Repository navigation
LongGuard v0.3.0
Added
- Multi-OS clean-environment wheel and PyPI installation smoke-test matrix across Ubuntu, macOS, and Windows on Python 3.11, 3.12, and 3.13 (
.github/workflows/publish.yml). - Candidate-to-stable wheel diff verification tool (
scripts/check_version.py --compare-wheel) enforcing zero runtime divergence between release candidates and stable distributions. - Automated credential-free repository examples execution test suite (
tests/test_examples.py) verified in release gates. - Consolidated release trust foundation from the 0.2.x patch series:
- Tag-driven release pipeline using PyPI Trusted Publishing (OIDC) with digital build provenance attestations (
0.2.2). - Version consistency validation tool (
scripts/check_version.py) ensuring synchronization acrosspyproject.toml,src/longguard/__init__.py,uv.lock, and distribution artifacts (0.2.2). - Cross-platform 9-cell CI matrix supporting Python 3.11, 3.12, and 3.13 across Ubuntu, macOS, and Windows with line-ending and encoding fixes (
0.2.3). - Nightly framework compatibility canary against latest and lowest declared dependency bounds with automated GitHub issue reporting (
0.2.4). - Per-step overhead benchmark harness with empirical p50/p95/p99 latency reporting in
docs/performance.md(0.2.5). - Automated Markdown link checker (
scripts/check_links.py) preventing broken documentation links (0.2.5). - Community governance infrastructure including issue/PR templates, Code of Conduct, and Dependabot automation (
0.2.1).
- Tag-driven release pipeline using PyPI Trusted Publishing (OIDC) with digital build provenance attestations (
Changed
- Promoted verified
0.2.5feature-frozen candidate to stable0.3.0. - Documentation and README claims reconciled with code reality: verified Python support, explicit runtime Limits section (
0.2.2), and measured performance wording (0.2.5). - Aligned optional framework dependency lower bounds to lowest verified passing versions (
crewai>=1.15.19) (0.2.4).
Compatibility and limits
- No runtime behavior changes: core imports, detector defaults, circuit breaker state machine, budget boundaries, and report schemas are identical to 0.2.0.
- 0.2.0 runtime boundaries remain: guarded graph resume is synchronous and same-process; CrewAI support is bounded to the tested sequential ReAct executor; detector evaluation results are provisional.
- Performance benchmark percentiles are measured on shared CI runners and serve as indicative baselines, not hard SLAs.
Install the exact version
pip install longguard==0.3.0Source commit: 5b4af933eaf621e8e071d58d73d7aa8f7a18da6f. Executed checks and commands are attached in release-evidence.json; framework and platform coverage is limited to that record.
Detector benchmark labels and release acceptance require maintainer review. No production adoption, provider savings, or untested platform claims are implied.