webweavex 3.0.0

Deterministic web extraction, runtime replay, browser reconstruction, and cross-language Kaalka parity for AI agents and applications.

3.0.0

• Production release
• Engineering artifacts removed
• Version updated to 3.0.0

Changelog

2.1.0 — 2026-06-12 — Synchronized cross-language release

Changed

• Synchronized version 2.1.0 across pub.dev (Dart), PyPI (Python), and npm (JavaScript).
• Standardized README structure across all three implementations (equivalent sections: Core Capabilities, Common Workflows, Supported Platforms, Versioning) and reconciled internal test-count metrics with the certified suite (1583 tests).
• Consolidated release staging into the canonical python / javascript / dart branches and removed the release/* staging branches.
• Re-ran cross-language parity, determinism, and publication validation against this version.

Notes

• No public API behavior changes. The cross-language deterministic contract and hashes are unchanged from 2.0.x; the crypto substrate pin (kaalka 5.0.0) is intentionally unchanged.
• Shared docs (ARCHITECTURE.md, CERTIFICATION.md, AI_AGENT_GUIDE.md, API_REFERENCE.md, LICENSE) are byte-identical across all three implementations.

[Unreleased] — Cross-language canonical contract + standalone Dart

Added

• cross_language_verifier/ — automated 3-language parity harness: 1100 deterministic torture vectors (multilingual Unicode, float matrix, key ordering, nested structures), 3 runs per language, byte-level comparison. Certified: 6601/6601 fields byte-identical across Python, JavaScript, and Dart (certification_report.json, kaalka_parity_matrix.json).
• lib/src/determinism/py_float_repr.dart — Python repr(float) formatting (canonical float form shared by all languages).
• lib/src/determinism/canonical_json_encode.dart — compact canonical JSON encoder, byte-identical to Python json.dumps(..., ensure_ascii=False, separators=(",", ":"), sort_keys=True).

Changed (canonicalization contract — affects hashes of edge-case payloads)

• Dart is now fully standalone: normalizeRuntimeValue performs NFKC via pure-Dart package:unorm_dart (byte-verified against Python unicodedata and ICU); the previous silent Process.runSync('node', ...) delegation is removed. Output no longer depends on whether Node.js is installed.
• Dict keys sort by Unicode code point (Python sorted semantics) in all languages; Dart/JS previously used UTF-16 code-unit order, which inverted astral-plane keys (e.g. U+1F680) relative to U+E000–U+FFFF.
• Integral floats < 2^63 canonicalize to integers in every language (JavaScript cannot distinguish 2.0 from 2); non-finite floats serialize as null on the stable path and 0 on the fingerprint path; fractional floats use Python repr thresholds in all languages.
• dumpsDeterministic (fingerprint path) now applies real NFC normalization in Dart (was identity) and .15g rounding only to fractional values.

[Unreleased] — Final Completion Protocol · Group D (cont.) + manifest-driven reports

Added

• tools/generate_reports.py — generates API_PARITY_VALIDATION_REPORT, FINAL_TRUE_PARITY_REPORT, PARTIAL_API_AUDIT, DEFERRED_API_AUDIT from PARITY_MANIFEST.json and syncs count tokens in the legacy reports. No report carries hand-maintained counts (resolved prior count disagreements).
• lib/src/application/runtime_application.dart — executeRuntimeObjective (executable parity, Python ≡ JS ≡ Dart), buildRuntimeGoal, and the saveApplicationMemory/loadApplicationMemory
  • saveNativeRuntime/loadNativeRuntime Kaalka pairs (proven by save→load deep-equality roundtrip).

Changed

• 5 APIs promoted Deferred → Complete (execute_runtime_objective + 4 application/native save/load pairs). API parity matrix: 94 Complete · 26 Partial · 8 Deferred · 0 Missing; 857 tests; 97.21% coverage. 14 executable-proven APIs. Deferred re-audit: 8 remain = 5 live-page browser (platform ceiling) + 3 large native/application cognition entry points (convertible, pending).

[Unreleased] — Final Completion Protocol · Group D (Deferred re-audit)

Changed

• Group D re-audit of the 15 Deferred APIs: only 5 are genuinely platform-bound (live-browser page); the other 10 are snapshot/data/persistence-input deterministic functions.
• extractContainerRuntime and extractIdeRuntime ported to Python's full field set and proven Python ≡ JavaScript ≡ Dart by execution → Deferred → Complete.
• API parity matrix: 89 Complete · 26 Partial · 13 Deferred · 0 Missing; 850 tests; 97.19% coverage. 13 executable-proven APIs. See DEFERRED_API_AUDIT.md.

[Unreleased] — Final Completion Protocol · Group C (build_browser_identity)

Changed (BREAKING)

• buildBrowserIdentity(profileId) now matches the canonical Python contract — a full port of the core.identity.* profile-generation subsystem (profile / user-agent / platform / language / timezone / webgl / canvas / font / media-device / navigator engines + entropy + fingerprint + data tables). Proven Python ≡ JavaScript ≡ Dart by execution on 4 profiles → Complete. The captured-map variant is buildBrowserIdentityFromCapture.
• API parity matrix: 87 Complete · 26 Partial · 15 Deferred · 0 Missing; 845 tests; 97.17% coverage. Group C done (11 executable-proven APIs).

[Unreleased] — Final Completion Protocol · Group B complete (executable parity)

Changed

• reconstructRuntime now exposes the canonical Python contract reconstructRuntime({semanticIr, workflowIr, synchronizationIr, executionIr, memoryIr, runtimeGraph, runtimeType, tick}); proven Python ≡ JavaScript ≡ Dart by execution → Complete. The envelope variant is reconstructRuntimeFromEnvelope (BREAKING).
• getRuntimeKernel — its observable kernel state ({runtime_type}) proven Python ≡ JavaScript ≡ Dart by execution → Complete.
• API parity matrix: 86 Complete · 27 Partial · 15 Deferred · 0 Missing; 838 tests; 97.13% coverage. Group B is fully complete (10 executable-proven APIs total).

[Unreleased] — Final Completion Protocol · Group B (executable parity)

Changed (contract alignment, BREAKING)

• computeGlobalRuntimeFingerprint, queryRuntimeGraph, and validateReplayEquivalence now match the canonical Python contracts and are proven Python ≡ JavaScript ≡ Dart by execution (validation/executable/) → promoted to Complete:
  • computeGlobalRuntimeFingerprint({extraction, graph, memory, sync, reconstruction, kaalkaSeal})
  • queryRuntimeGraph(Map graph, Map query) → {results, count, bounded}
  • validateReplayEquivalence(original, replayed) → {equivalent, checks:[{name, ok, original, replay}], bounded}
  • The Dart-native variants are retained as computeRuntimePipelineFingerprint, queryRuntimeGraphTyped, and validateReplayEquivalenceExtended.
• API parity matrix: 84 Complete · 29 Partial · 15 Deferred · 0 Missing; 831 tests; 97.13% coverage.

Added

• tools/generate_parity_manifest.py → PARITY_MANIFEST.json (single source of truth: per-API python/javascript/dart presence, contract/behavior/executable parity, classification, proof type). lib/src/parity/canonical_runtime.dart (Python-canonical Group-B ports).

[Unreleased] — Final Completion Protocol (executable parity)

Changed (contract alignment)

• buildRuntimeMemory and queryRuntimeMemory now match the canonical Python contract: buildRuntimeMemory({runtimeHistory, lineage, semanticRelations}) and queryRuntimeMemory(memory, queryType, term). Proven Python ≡ JavaScript ≡ Dart by execution (validation/executable/, EXECUTABLE_PARITY_MATRIX.md) → promoted to Complete. The previous graph-based variants are retained as buildRuntimeMemoryFabric / queryRuntimeMemoryFabric (BREAKING for callers of the old graph-based signatures).
• extract_database_runtime (postgres/mysql/sqlite/redis + degraded) and extract_kubernetes_runtime re-implemented to Python's full field set; proven Python ≡ JS ≡ Dart → Complete.
• API parity matrix: 81 Complete · 32 Partial · 15 Deferred · 0 Missing; 816 tests; 97.24% coverage.

Added

• validation/executable/ — 3-language executable parity harness (Python 2.0.1 + JavaScript via tsx + Dart) producing EXECUTABLE_PARITY_MATRIX.md; COMPLETE_API_PROOF_MATRIX.md, PARTIAL_API_AUDIT.md, DEFERRED_API_AUDIT.md auto-generated by tools/.

[Unreleased] — Wave 3–4 (parity + OSS)

Added

• healSelector / buildSemanticAnchor — native Dart port of the canonical Python heal_selector (moved Deferred → Partial). DOM-node strategies (text_anchor, attribute_anchor, structural_fallback) are full-fidelity, proven by 11 deep-equality vectors vs the Python reference (validation/parity/selector_healing_api_vectors.json, test/parity/selector_healing_parity_test.dart). The semantic_anchor HTML sub-path matches BeautifulSoup for well-formed content; nested inline markup is the documented bounded edge.
• replayInteractions / recordInteraction — native Dart port of Python replay_interactions (moved Deferred → Partial). The returned structure is a full-fidelity pure function of the interaction log, proven by 6 deep-equality vectors vs Python 2.0.1 (validation/parity/interaction_replay_api_vectors.json). The live-page action dispatch is the documented bounded edge.
• Three-way parity validator: validation/validate_parity.dart now asserts the deterministic core against both the JavaScript and Python references (Python ≡ JavaScript ≡ Dart, 11 vectors).
• DART_REALITY_AUDIT.md, FINAL_TRUE_PARITY_REPORT.md (proof-coverage audit + honest name-parity vs signature-parity disclosure); tools/proof_coverage.py.
• OSS governance files: GOVERNANCE.md, MAINTAINERS.md, CODEOWNERS, RELEASE.md, SUPPORT.md.
• Full Phase-10 README.md (Installation, Features, API Reference, Examples, Performance, Testing, Coverage, CI/CD, Pub.dev Release, Vision).
• Validation reports regenerated from measured reality: REPOSITORY_VALIDATION_REPORT, TEST_VALIDATION_REPORT, COVERAGE_VALIDATION_REPORT, API_PARITY_VALIDATION_REPORT, README_GAP_REPORT, OSS_VALIDATION_REPORT, RELEASE_READINESS_REPORT, FINAL_STATE_OF_DART_BRANCH.

Changed

• CONTRIBUTING.md rewritten for Dart (was carrying Python pip/pytest/playwright instructions).
• API parity matrix: 79 Complete · 34 Partial · 15 Deferred · 0 Missing (Proof Coverage Audit downgraded 11 unproven Complete → Partial).
• Test suite 779 → 802; line coverage 97.23% → 97.26% (6394/6574).
• .pubignore excludes CODEOWNERS from the published package.

2.0.1 — 2026-06-09 — Dart native release

Added

• Native Dart runtime cognition infrastructure for humans and AI agents
• Kaalka v5 parity pipeline (kaalka@5.0.0 from pub.dev)
• Cross-language validation (11/11 JavaScript reference vectors)
• Browser HTTP extraction, replay equivalence, runtime memory, reconstruction
• GitHub Actions CI on dart branch

Runtime parity

• Ported 12 runtime-cognition families to native Dart with proven cross-language hash parity — Dart output hashes byte-identically to Python's compute_deterministic_hash (computeDeterministicHash(dartOut) == h(pyOut)), save/load proven by temp-file roundtrip: causality, semantic, synchronization, evolution_runtime, workflows, execution, memory-runtime, reconstruction-runtime, connectors/streaming/interaction, query/reasoning, persistence/crypto-session/identity/adaptive/distributed/ session/auth, kernel/contracts/unified-IR — +72 public APIs.
• Public-API parity vs Python __all__: 88/128 Complete, 23 Partial, 17 Deferred (OS/desktop/Electron/DevTools — documented), 0 Missing. See PUBLIC_API_MATRIX.md, DART_PARITY_AUDIT.md, DART_RELEASE_GAP_REPORT.md.
• ~145 parity-proof vectors in validation/parity/*_api_vectors.json; assertions in test/parity/*_parity_test.dart. Honestly bounded (not faked): compile_document/compile_repository (need NLP/AST), run_canonical_pipeline (deterministic core only), and document/repository sub-paths of the query family — all classified Partial in the matrix.

Quality

• Comprehensive unit-test suite: 779 tests across all subsystems and the 12 newly-ported runtime families
• 97.23% line coverage (dart test --coverage); the only sub-90% file is normalization.dart (Node-subprocess NFKC fallback, unreachable when Node is on PATH)
• dart analyze clean (zero issues) under strict-casts + strict-inference
• dart format clean across the package
• dart pub publish --dry-run clean (no packaging warnings)

Security

• No auth bypass, CAPTCHA defeat, or credential cracking