Repository navigation
Releases: LoglensAI/LogLens-AI
Releases · LoglensAI/LogLens-AI
Release list
v0.13.0
Full Changelog: v0.12.0...v0.13.0
v0.12.0 — daemon + installers
v0.11.0 - CI/CD integration + unified explainable scoring
LogLens AI now runs inside your CI/CD pipeline - analyze build/test logs, get a
machine-readable report, and fail the build on real incidents. Plus every mode
now scores on one scale and explains why each anomaly was flagged.
🚀 CI/CD integration
- GitHub Action - drop LogLens into any workflow:
- uses: LoglensAI/LogLens-AI@v0.11.0
with:
source: build.log
fail-on: error # critical | error | warning | fatal | anyPosts an incident table to the job summary, adds inline annotations, and
gates the build via exit code.
--format json- clean, machine-readable output for any CI system.--fail-on <level>- non-zero exit (code 2) when an anomaly is that severity
or worse, so a bad log fails the pipeline.
🔍 Smarter, explainable detection
- Unified scoring across fast / turbo / deep - one 0–1 scale, one reason
vocabulary. Turbo now explains itself (previously gave no reasons). ↳ why:on every anomaly by default - severity, rarity, keywords, bursts.- Mode labels - each run shows the signals it used (and supervised vs. unsupervised),
so results are comparable across modes. - Rankable incidents - the most severe levels are de-saturated (always flagged,
but scored by evidence) instead of all tying at 1.0. - Live progress on previously silent phases - no more "is it stuck?".
🐛 Fixes
- Consistent line-count reporting between modes (
read → parsed). - Documented why turbo and default modes can differ (supervised vs. unsupervised).
loglens … | headno longer prints aBrokenPipeError.
📦 Install
pip install --upgrade loglensai # 0.11.0
docker pull loglensai/loglens:0.10.0Full docs: see docs/DOCUMENTATION.md · Benchmarks: docs/BENCHMARKS.md
v0.9.0
[0.9.0] - 2026-09-22
Bug Fixes
- Platform-agnostic — UTF-8/CRLF handling, Windows resource guard, spawn fallback (b02a1f3)
Chores
- Updated the readme. (4852f22)
Documentation
- Tracking badges + reproducible benchmark links (15e0b1b)
Other
- Added new logo image (8672fc8)
- Phase 7: enforce layer boundaries in CI (import-linter); move entry points to interface (2d75603)
- Phase 6: reorganize modules into layer packages (domain/detection/application/infrastructure/interface) (0fea529)
- Fix CI mypy: skip numpy 2.x PEP695 stubs; type LiveProgress.task_id (2ad3ab0)
- Phase 5: clear mypy to zero, enforce ruff/mypy + coverage gates in CI (b51e80f)
- Phase 4: LLM redaction, webhook SSRF guard, secret + thread-safety hardening (47e5afb)
- Phase 3: fix live api inversion, extract reporting service, alerter port (08f31ba)
- Added cloud.py (fa5c50a)
- Phase 2: SRP decomposition split detector.detect, extract cloud-JSON, typed LogEntry, dedup CLI (39881d3)
- Normalize trailing newlines + import order (e846906)
- Phase 1: single severity source of truth, one masker, drop dead code, tidy scripts (e4822fa)
- Removed the data folder (9244d4b)
- Phase 0: fix silent render bug, remove dead deps, add observability + tooling (6a496f1)
- Add file-type guard + LLM provider abstraction (Closes #2) (dc50f9d)
- Merge pull request #1 from Dibya12345/main
chore: Updated the readme. (90773c0)
- Added Ci gate check (933dda9)
v0.8.1
v0.8.0
Bundled default model with auto-load, compression, feature caching, chunked embeddings