Skip to content

Documentation synchronization audit — 2026-10-21

This is a documentation and website pass over the existing implementation. It does not approve scientific mappings, execute source access or change pipeline architecture.

Baseline inspected before editing

Recent history: cf18bc2 website update, 0646d2f multi agent build, 3f64c2e first prototype. Existing uncommitted inventory, contract and semantic-test changes were retained. Prior current-state and semantic-discovery audits describe earlier snapshots; their historical test counts are not current verification results.

Inspection covered README/AGENTS, MkDocs navigation, templates, CSS, assets and figure generator; the source module map and typed schemas; discovery, semantic registry, benchmark, harmonization, scoped review, streaming and representation boundaries; tests, example commands, tracks, provenance, canonical CSV/assessments and generated mirrors; retained Sentinel reviews and semantic benchmark manifests. No custom site JavaScript existed at baseline. Pages and template-sync workflows remain unchanged.

Checklist and resolution

Audit concern Finding and resolution
Implemented but hard to find Semantic discovery contracts and metadata diagnostics linked from the public narrative and separate benchmark guides
Outdated documentation Figure counts, current test status, architecture buffering wording and agent runtime paragraph corrected
Planned described too broadly Composition, graph depth, potential, diversity and flagship selection explicitly labeled research frameworks
Implementation differs Inventory scope overlap and metadata field presence distinguished from ontology equivalence and dataset suitability
Duplicated documentation Detailed formulas stay in the benchmark contract guide; new conceptual pages link there
Missing links Glossary, mappings, recipes, source evidence, benchmarking and review linked across public and technical layers
Obsolete figures Expanded system/provenance and coherent semantic figure family generated from maintained specifications
Obsolete terminology Multiple organizations and support dimensions made explicit; standardized, harmonized and embedded remain separate
Missing glossary terms Added explicit use, potential, diversity, leverage and benchmark analysis terms
Missing examples Proposed increasing-complexity observations, candidate dataset pairs and evidence-driven flagship selection added

Reader checks

Scientist: Home → biological organization → support → recipe. The source's meaning survives; shared units, labels or coordinates do not imply equivalence. Ontology expert: Landscape filters → assessment evidence → comparison → ontology benchmark → mapping workflow. Coverage ratings remain provisional. Engineer: Architecture → generated contracts/modules → streaming/review → MOSAIKS → tests. Full decomposition and ontology-service adapters remain planned.

Remaining scientific work

Ontology graph-depth/composition tests, observation-level accumulation experiments, evidence-reviewed dataset potential/diversity profiles and flagship selection have not been computed. The local diagnostics cannot substitute for them. No approved real Sentinel MOSAIKS product or completed field-to-satellite synthesis is claimed.

See verification and reader review for the completed site checks.