Skip to content

fdars 0.11.0

Choose a tag to compare

View all tags
@sipemu sipemu released this 07 Sep 07:34
· 251 commits to main since this release
v0.11.0
e21c6be

fdars 0.11.0

Milestone v12.0 — Docs Depth, Card Coverage & AI Capability Skill. A documentation + AI-discovery release: no fdars-core bump and no changes to existing binding behavior. The only new importable code is the capability-discovery tooling.

Added — AI capability discovery (new code)

  • fdars_list_capabilities MCP tool on the existing server — returns the full static capability surface (modules, callables, signatures, one-line purpose). Provably LLM-free: loads a pre-generated JSON via importlib.resources, no model/network call. Covered by an extended guard-sync test (Py3.9-safe primary + Py3.10+ boundary companion) so the LLM-free compute boundary stays enforced.
  • Capability generator (scripts/generate_capability_dataset.py) → byte-stable python/fdars/_capability_map.json (31 modules / 437 callables), the single source of truth all three surfaces consume; drift-guarded by tests/test_capability_accuracy.py (regenerate-and-diff + every-method-importable).
  • fdars-capabilities Agent Skill (.claude/skills/) — spec-valid, capability-discovery trigger, explicitly disjoint from the existing fdars-advisor skill.
  • llms.txt whole-library API digest (served at /llms.txt) + an in-nav AI Capability Map page.

Added — documentation

  • 3 flagship worked-example pages (offline-runnable, FDARS_FENCE_OK): Fréchet density regression, functional-time-series forecasting, and phoneme shapelet classification.
  • 10 method pages (regression + analyze families) raised to mature-page parity — "When to use", "See also", parameter/interpretation guidance, ≥3 admonitions, ≥3 runnable fences each.
  • 21 hand-authored SVG thumbnails + gallery cards — align/represent/regression/analyze landing pages reach 100% card coverage.

Fixed — documentation accuracy

  • ~20 API-accuracy corrections across the deepened pages, verified against the shipped bindings (e.g. frechet_anova signature/defaults/return-keys, predict_fof arg order, fof_cv params, fam defaults, discover_shapelets dict return, FTS PVE from scores.var(axis=0) not weights).

Quality gates

  • Whole-site mkdocs build --strict green offline; every worked example emits FDARS_FENCE_OK.
  • All new SVGs pass the SVGO idempotence / build-determinism gate; advisor + MCP + guard-sync tests green.

Full machine-readable API digest: https://sipemu.github.io/pyfda/llms.txt

Assets 2
Loading