Documentation

Start with the system’s scientific boundaries, run a small example, then inspect contracts and implementation. This documentation is maintained as Markdown in the same repository as the code. The public site renders that source.

Choose a path

Your goal	Start here	Continue with
Understand the project	Approach	Core concepts, current status
Understand the architecture	System architecture	Current-state audit, typed contracts
Build and run	Complete offline example	Streaming, OpenClaw / Verde
Assess the science	Harmonization	Evidence review, provenance, science tracks
Extend the implementation	Extension guide	Module map, contracts, design decisions
Verify a dataset or change	Tests and builds	MOSAIKS integrity and held-out diagnostics, Sentinel review case

Semantics: a major project resource

Explore the ontology landscape, glossary, coverage matrix and human comparison workflow. The inventory informs discovery without letting Module 1 approve mappings. The benchmark produces auditable metadata diagnostics.

The three module boundaries

Discovery & access accepts a scientific request and produces candidate source resources with native metadata and access evidence. It does not establish semantic compatibility. Discovery · Data sources

Harmonization & review turns evidence into versioned allow-listed recipes, review decisions and canonical observations. Model judgments never grant execution permission. Harmonization · Scoped review

Representation accepts harmonized outputs and produces vectors with model, preprocessing and observation lineage. Reference encoders · MOSAIKS

Execution modes

Offline generated examples are the reproducible default. Live discovery and model inference are explicit commands; unavailable services never silently become mocks. New real raster paths stream bounded windows. Historical snapshot commands remain separate. Read real-data status before using a live workflow.

Choose your route

Scientists can start with biological organization and support. Semantic-data experts can inspect the filterable landscape, ontology research framework and dataset framework. Engineers can follow typed contracts, module APIs and verification. All routes connect to the human review workflow.