Skip to content

CHANGELOG.md

Latest commit

 

History

History
99 lines (81 loc) · 4.53 KB

CHANGELOG.md

File metadata and controls

99 lines (81 loc) · 4.53 KB
Raw

Changelog

All notable changes to TRACER are recorded here. This project follows semantic versioning.

Unreleased

Changed (breaking)

  • Require Python 3.12 or newer; drop support for Python 3.9–3.11. CI tests Python 3.12, 3.13, and 3.14, plus the minimum core dependencies on 3.12.
  • Allow NumPy 1.26.4 through 2.x so Python 3.14 can install compatible wheels. Require scikit-learn 1.4.2+ and joblib 1.3.2+ in core, and PyTorch 2.4.1+ for the optional embedding extras. Remove the obsolete NumPy downgrade advice.
  • Build releases and run the documented Python sidecar on Python 3.14.

Removed (breaking)

  • The tracer command-line executable and command implementations. Use the Python APIs for fitting, scanning, updating, reports, and serving predictions.
  • The Tracer app client, authentication, and automatic app telemetry exporters in both Python and JavaScript. Existing app credentials no longer trigger uploads. Local recording, custom sinks, and explicitly configured generic HTTP exports remain available.
  • App connection instructions and command-line documentation. All current examples use the library APIs.

0.3.3 (2026-06)

Fixed

  • build_global() and _build_accepting_stage() no longer crash with a TypeError when all surrogate candidates fail to train. They now return a structured failure state, consistent with every other non-deployable path.
  • Router.load() now raises a clear ValueError when manifest.json reports selected_method: null, instead of silently loading a stale pipeline.joblib left on disk from a previous fit.

0.3.2 (2026-06)

Fixed

  • load_traces() now reports the correct 1-indexed file line number in ValueError messages. Previously the count excluded blank lines, so the reported line was lower than the actual position of the bad record.

0.3.0 (2026-06)

Added

  • tracer.watch: a decorator that records LLM calls on the OpenTelemetry GenAI (gen_ai.*) schema, with pluggable local and generic export sinks.
  • @tracer-llm/watch: a zero-dependency JavaScript/TypeScript mirror of the watch decorator (async-context aware), so JS pipelines get the same one-line observability as Python.
  • Lazy package initialization so import tracer stays fast and only pulls in the heavy pieces when you actually use them.

Changed

  • Trace ingestion accepts additional input and label aliases.

Docs

  • New guide for the watch decorator.

0.2.0 (2026-06)

Added

  • tracer.scan(): a fast, conservative day-one read of a traces file, before any training. It groups traffic by similarity and measures, on a held-out slice it never saw, how much a near-free model can answer at your target agreement, using exact Clopper-Pearson bounds, with an optional per-1k and monthly savings estimate. Ships a self-contained HTML report with an interactive 3D map of the embedding space (hover-to-inspect cells, a Verdict/Label colour toggle, and PCA/UMAP/t-SNE layouts). Exposed as tracer.scan().
  • Distance-based OOD safety gate: at inference the router defers inputs that fall far from the training distribution (kNN distance, global and per-predicted-label thresholds) regardless of surrogate confidence, so off-distribution traffic goes to the teacher instead of getting a confident guess.
  • FitConfig.skip_candidates to drop named candidates from the zoo, including tree surrogates when a lighter sweep is desired.
  • Trace loaders accept common key aliases for both input and label (input/query/text/prompt/question and teacher/teacher_output/label/intent/output/answer).
  • Bring-your-own embeddings for tracer.scan(): pass a precomputed array with embeddings=, or select a local sentence-transformers model with model=.

Changed

  • The parity gate now certifies on an exact held-out lower bound instead of an in-sample point estimate, so a policy cannot clear the target by in-sample luck and then break the contract on real traffic. Coverage is now monotonic in the target, and a hybrid select-then-verify procedure recovers coverage at strict targets that a plain held-out split discarded.
  • The HTML report is restyled to the light Tracer theme, and the word "audit" is dropped across the report and docs.

Fixed

  • Non-monotonic coverage in the gate (a stricter target could deploy more coverage than a looser one).
  • NaN-robustness in acceptor fitting on degenerate surrogates.

0.1.3

Initial public releases: the parity-gated router (fit, update, load_router, serve), the HTML report and Sankey diagram, and the embedder factories.