Public docs

How Martenweave Works

Martenweave turns scattered model knowledge into checked, searchable model context.

The workflow is deliberately explicit so humans can inspect each step.

1. Bring Inputs

Inputs can include:

Raw inputs are not treated as truth. They provide evidence.

2. Normalize Into Model Objects

Martenweave represents model knowledge as canonical objects such as:

The canonical files are Markdown and YAML. They are readable in Git and can be reviewed without a hosted UI.

3. Validate Consistency

Validation checks object structure and references before trusted indexing.

Checks include:

4. Build The Index

Generated outputs are rebuildable.

They can include:

If generated files disagree with canonical files, canonical files win.

5. Analyze Gaps And Impact

Once the model is indexed, teams can ask practical questions:

6. Propose Changes

AI and import workflows create PatchProposal objects instead of silently changing canonical files.

A good proposal records:

7. Human Approval

Humans approve, reject, or revise proposals.

Approved changes become ChangeRequests and then canonical updates. This keeps the change trail visible.

8. Export Reports And Context

Martenweave can export model context for review, analysis, automation, and AI-assisted work.

Examples include:

Summary

Inputs bring evidence.

Model objects create structure.

Validators create trust.

Indexes create search and analysis.

Patch proposals keep AI and imports reviewable.