Public docs

Import and Export

Martenweave treats import and export as model review workflows, not hidden synchronization.

The core rule is simple: imports should create evidence, profiles, or PatchProposals. Exports should produce reviewable files, reports, schemas, or context. Canonical model changes still go through validation and approval.

Available Now

The current core includes commands and services for:

• profiling CSV and XLSX datasets
• detecting dataset-to-model gaps
• importing Google Drive files for profiling
• importing Google Sheets as PatchProposals
• importing spreadsheet edits as PatchProposals
• exporting canonical model objects to CSV or XLSX
• exporting JSON Schema for canonical object types
• exporting canonical model objects to Google Sheets
• generating search and lineage JSONL exports
• generating GitHub-ready change bundles
• publishing issue drafts and pull request bundles through review workflows

The exact command names and options are documented in the core repository and CLI help.

Supported By Examples Or Manual Workflow

These workflows are practical today, but may still require manual setup, local files, or review steps:

• mapping spreadsheet review
• dataset profiling before mock load
• source-to-target field coverage checks
• export of model objects for steward review
• import of reviewed sheet edits as PatchProposals
• GitHub issue drafts for model gaps
• local automation around generated reports and JSONL files

Planned Or Integration Direction

These are directionally aligned with the product, but should not be described as fully implemented hosted integrations:

• hosted workbench review queues
• enterprise connector marketplace
• automated bidirectional sync with MDM platforms
• direct SAP customizing extraction packs
• large-scale catalog or lineage platform replacement
• turnkey Jira, ServiceNow, or Confluence sync

Martenweave can grow toward these paths, but the current open-source core is a local-first package and CLI.

Import Safety Model

Raw files are not model truth.

Good import behavior:

• profile the source
• record evidence
• infer draft model objects only when appropriate
• create PatchProposal objects for review
• validate proposed references and types
• let humans approve changes

Bad import behavior:

• silently overwrite canonical files
• treat source spreadsheets as always correct
• bypass validation
• hide assumptions
• create unreviewed AI changes

Export Safety Model

Exports should be reproducible from canonical model state.

Useful exports include:

• steward review workbooks
• search documents for retrieval
• lineage edges for analysis
• gap summaries
• impact reports
• schemas for integration
• AI context files

Generated files are disposable. Canonical model files remain the source of record.