quantbt-engine 1.1.0__tar.gz → 1.2.0__tar.gz

quantbt_engine-1.2.0/CHANGELOG.md

@@ -0,0 +1,223 @@
+# Changelog
+
+All notable changes to `quantbt-engine` are documented here.
+
+## [Unreleased]
+
+## [1.2.0] - 2026-10-23
+
+### Core 1.2.0 / Native 0.4.3
+
+- Retargeted the unreleased core candidate from1.1.2to1.2.0 to reflect the
+  diagnostics, Mode4/Mode5, sampler/meta, reporting and account upgrade below.
+  Historical1.1.2receipts retain their original scope; they are not renamed
+  release evidence. Native version remains0.4.3; generated product metadata
+  follows the new pair, requiring fresh matching native artifacts.
+- Reduced full Reporting V1 Python scalar-dispatch and metadata overhead;
+  closed-ledger reason counting uses one pass instead of one scan per reason.
+  Exact before/after output, snapshot/identity and undefined-policy parity are
+  tested. Reporting formulas, financial kernels and WFO selection are unchanged;
+  timing gains are workload-specific and do not imply a universal speedup.
+
+- Added opt-in `reporting=ReportingRequest(...)` on completed endpoint/result
+  reports and quick plots. Account-period, execution, closed-trade/action and
+  drawdown episode output has explicit baseline, frequency, currency, coverage
+  and unavailable policies. This opt-in does not alter execution or WFO/meta
+  objectives; separate account repairs below intentionally correct financial
+  edge cases. Read [Reporting V1](docs/metrics/PUBLIC_REPORTING.md).
+- Reduced optional reporting copies and duplicate hashing using immutable
+  bytes-owned snapshots, a prepared boolean mask and one report/plot snapshot.
+  Cache hits still rehash mutable completed input; formulas, exact identities,
+  requested episode detail, legacy reports and financial/WFO metrics are unchanged.
+- Fixed immutable reporting snapshot reconstruction after copy/deepcopy and
+  pickle, including nested execution paths and drawdown episodes. Constructors
+  revalidate storage and recompute hashes; changed observations cannot reuse a
+  saved identity. Python reporting scans use zero-copy native scalar access
+  with preserved numerical overflow policies and exact differential coverage.
+
+- Added opt-in Mode 4 / per-fold causal meta selection and a shared sampler
+  bridge for legacy TPE, grouped multivariate TPE, CMA-ES and fixed-space Sobol.
+- Qualified causal history, ridge artifacts, reference/prepared selection
+  parity, exact reusable calendar/market witnesses, observed thread telemetry
+  and bounded sequential W3 meta on the original native accounting pass.
+- Enabled the already-qualified QMS native exports in default Cargo builds;
+  Linux core dependency is exactly `quantbt-native==0.4.3`. The ABI remains
+  versioned; later ACC-01 changes affect both Python and Rust financial source
+  and require fresh compatible artifacts. Meta remains disabled unless requested.
+- Earlier QMS wheel/sdist, installed-W3 and six-row Ubuntu receipts retain
+  their historical source scope. They do not certify the later Mode 5,
+  reporting and account repairs. Final M5-12 artifacts, remote qualification
+  and public consumers remain separate gates.
+- Added diagnostics-first opt-in Mode 5 calibration with versioned objectives,
+  selectors, diagnostic evidence/rights, frozen plans, explanation and winner
+  qualification. Shared diagnostics/reporting include Mode 4 fold consistency,
+  R/Q, decision coverage and explicit cross-mode researcher handoff. No automatic
+  sampler/selector router or universal economic improvement is claimed.
+- Extended the stable endpoint API additively: optional `execution_endpoint`
+  on `walk_forward`, keyword-only `meta_history` on `backtest`, optional
+  `reporting` on completed reports/plots, and `research_report(...)`.
+- Corrected current-capability account/configuration behavior: financial clocks,
+  funding interval coverage, contract multipliers, risk-reducing admission,
+  constraints, invalid inputs, actual attribution and reporting horizon/time
+  handling. Unsupported non-default options fail explicitly; gross-cross/netting
+  is not hedge inventory or isolated liquidation. See
+  [ACC-01 migration](docs/portfolio_review/ACC01_CONFIGURATION_AND_MIGRATION.md).
+- Refreshed the generated root API inventory to 527 exports (11 additions to
+  the stale generated list, no removals), with an exact metadata-only amendment
+  that preserves the separate financial source qualification and historical receipts.
+- Extracted cold option validation and portfolio report helpers, preserving
+  endpoint/private aliases, moved function bodies and retained engine code.
+  Exact source qualification is separate from historical financial-invariance
+  receipts and still does not certify installed artifacts.
+- Added maintained installed-consumer/CI wiring for the Mode 5,
+  winner/donor, reporting and account changes. Execution on final wheel/sdist
+  artifacts and remote/public qualification remain separate gates.
+- Aligned tag/release regression with the qualified CI native-wheel identity
+  pin and isolated host-PID tests. Release publication is native-first and
+  requires its own artifact and public-consumer gates; this changelog is not
+  a publication receipt or a new scientific acceptance claim. See the
+  [release checklist](docs/testpypi_release_checklist.md).
+
+## [1.1.1]
+
+### Changed
+
+- Released the completed statistical WFO follow-up: stationary-bootstrap index
+  construction now preserves the existing RNG draw sequence while avoiding
+  repeated Python work, and report-metric objectives share one report snapshot
+  per evaluation.
+- Updated the governed wheel-only companion contract to
+  `quantbt-native==0.4.2`; `quantbt-engine==1.1.1` keeps the exact
+  platform-marked dependency for supported Linux x86_64 CPython 3.11-3.13
+  consumers.
+
+### Compatibility And Verification
+
+- No public endpoint, methodology, selection, accounting, trace, or backend
+  promotion policy changed. The release remains native-first: publish the
+  complete native wheel matrix before publishing the matching core package.
+
+### Changed
+
+- NEXT-01 reduces generic every-bar and declared sparse callback overhead while
+  retaining full financial/audit output and dynamic callback compatibility.
+- NEXT-02 reuses immutable WFO market/window preparation and native score
+  requests while preserving candidate ordering, selection and fresh accounts.
+- NEXT-03 makes `src/quantbt` the sole Python package source; the reviewed
+  duplicate root modules are retired. Local developers should use an editable
+  install rather than a parent-directory import-path override.
+- Research audit exposes objective components and eight record-family links
+  while preserving existing public trial/candidate tables and legacy exports.
+
+### Compatibility And Qualification
+
+- Endpoint factory signatures are unchanged across NEXT-01 through NEXT-03.
+  Reactive WFO adds optional `preparation_policy="prepared"`; numeric reactive
+  minimal output omits diagnostic tables while retaining financial paths.
+- Local release shards and isolated wheel/sdist/editable consumer checks pass.
+  Publication requires new compatible core/native versions and remote matrix
+  certification. Performance claims remain limited to the measured workloads.
+- CI and release test checkouts retain Git history for qualification ancestry
+  verification; the validator continues to reject unverified source ancestry.
+
+## [1.1.0]
+
+### Changed
+
+- Renamed the governed public release line to `1.1.0` and synchronized package
+  metadata, runtime version, generated core/native contracts, lockfile, and
+  release documentation.
+- Prepared the governed public `quantbt-native==0.4.1` companion contract for
+  Linux x86_64 / CPython 3.11-3.13. The core declares an exact,
+  platform-marked dependency; unsupported platforms keep Python/Numba.
+- Added a native-first, wheel-only manylinux OIDC release workflow and a six-row
+  Ubuntu 22.04/24.04 Poetry consumer proof. The core release preflights the
+  matching public native artifact before it can publish.
+- Kept public release certification self-contained by excluding private,
+  out-of-tree Grid integration fixtures from its test shard profile.
+- Updated release, install, capability, troubleshooting, and handoff docs to
+  state that supported users receive a pre-built wheel rather than compiling
+  Rust locally.
+
+### Compatibility
+
+- No endpoint signature, execution policy, promotion threshold, or domain
+  accounting behavior changes in this packaging-only phase.
+
+## [1.0.9] - 2026-08-21
+
+### Added
+
+- Direct PyPI discovery and installation guidance in the repository README.
+- Release-preparation guidance for the governed Rust-first routes, exact
+  installed-wheel certification, and the matching core/native compatibility
+  contract.
+
+### Changed
+
+- Promoted the bounded Rust-first public policy for certified static command
+  tapes, Native Strategy IR, batch, and causal-fold workloads while preserving
+  Python authority for callbacks, reactive strategies, generic portfolio,
+  basket, arbitrage, and options routes.
+- Updated the core product registry to the `1.0.9` release line. The staged
+  `quantbt-native` companion remains unpublished and capability-gated.
+
+### Verification
+
+- Source-mirror, generated-product-contract, documentation-link, build, and
+  installed-wheel release checks are required from the final tagged commit.
+
+## [1.0.8] - 2026-08-12
+
+### Changed
+
+- Added strict nested causal retraining for
+  `mode_1_decay + optimization_schedule="per_fold_causal"`. Mode 1 decay now
+  uses inner IS-only folds while each outer OOS segment remains untouched until
+  its parameters are frozen.
+- Added explicit chronological validation metadata so consumers can distinguish
+  retrospective global calibration, selection-adjusted OOS, and strict outer
+  OOS evaluation without breaking legacy result fields.
+- Added a bounded-RSS release test runner that executes deterministic pytest
+  shards in fresh processes; CI and publish workflows use the same selection.
+
+### Verification
+
+- Root/source mirror, prepared/reference WFO parity, nested-fold causality,
+  and the full non-real-data release suite passed before this version bump.
+
+## [1.0.7] - 2026-08-04
+
+This is the first independently installable core package release line.
+
+### Release candidate
+
+- `1.0.7rc1` was blocked before publication because its lockfile version was stale.
+- `1.0.7rc2` prepared for TestPyPI on 2026-08-03 with immutable lock validation.
+- `1.0.7rc2` passed clean TestPyPI installation and functional endpoint smoke.
+- Python 3.11-3.13 package validation is required before final publication.
+- `backend="auto"` remains Python.
+- `backend="rust"` remains explicit and experimental.
+- `quantbt-native` is not included in this core release.
+- Native crate and Python metadata remain aligned to API version `0.4.0`.
+
+### Added
+
+- Stable `from quantbt import QuantBTEndpoint` import contract.
+- NumPy/Numba native vectorized, event-driven, portfolio, arbitrage, options,
+  intrabar, and walk-forward research routes.
+- Optional extras for optimization, reports, visualization, and NautilusTrader
+  validation.
+- Prepared service contexts and report levels for repeated research workloads.
+- Explicit Python/Rust native-event selector contract with Python as the
+  release default and Rust as a capability-gated experimental backend.
+- Wheel, sdist, clean-install, source-sync, parity, and RSS release gates.
+
+### Release policy
+
+- The core `quantbt-engine` distribution is the release candidate for PyPI.
+- `quantbt-native` is not part of this core release and is not exposed through
+  a non-empty `native` extra until its wheel matrix and incremental RSS gates
+  pass.
+- `native_backend="auto"` remains Python; explicit Rust requests never
+  silently fall back to Python.

{quantbt_engine-1.1.0 → quantbt_engine-1.2.0}/MANIFEST.in

@@ -4,6 +4,8 @@ include CHANGELOG.md
 include pyproject.toml
 
 recursive-include src/quantbt *.py py.typed
+include src/quantbt/benchmarks/README.md
+include src/quantbt/benchmarks/phase7_thresholds.json
 
 global-exclude *.pem
 global-exclude *.key
@@ -14,6 +16,8 @@ global-exclude .pypirc
 prune upgrade/private
 prune upgrade/local
 prune upgrade/drafts
+prune upgrade/alphas
+exclude upgrade/MODE5_INVARIANT_AND_ALPHA_PROFILING_SPEC_V1.md
 prune benchmarks
 prune artifacts
 prune data/raw