Public docs

Martenweave Roadmap

Martenweave is intentionally backend-first. The roadmap favors a useful, trustworthy model layer before a large platform surface.

Phase 1 — Core Registry

Status: largely present in martenweave-core 0.4.0.

• initialize model repositories
• store canonical Markdown + YAML model files
• validate objects and references
• build generated SQLite and JSONL indexes
• search, query, trace, and impact-analyze objects
• export and import model review files

Phase 2 — Model Intelligence

Status: partially present and expanding.

• dataset profiling
• dataset/model gap detection
• health reports
• governance scorecards
• ownership reports
• audit and usage reports
• source registry and stale-source patterns
• lineage and impact reporting

Phase 3 — AI-Assisted Workflows

Status: proposal-first workflow exists; provider and agent integrations continue to evolve.

• evidence extraction
• PatchProposal generation
• issue drafts
• GitHub-ready change bundles
• approval gates
• ChangeRequest lifecycle
• AI usage telemetry and evaluation patterns

Phase 4 — Domain Packs And Starter Scenarios

Status: active examples exist.

• SAP Business Partner / Customer starter pack
• Supplier / Vendor starter pack
• simple product example
• generic product model example
• data governance templates
• AMS knowledge continuity scenarios
• more domain packs without making any one domain define the core

Phase 5 — Optional Workbench

Status: intentionally later.

• lightweight review UI
• review queues
• impact explorer
• model health dashboard
• graph visualization where useful

The UI should support model truth, not replace it.

Integrations And Agent Interfaces

Backend extensions may include:

• local API
• MCP server
• Google Drive / Sheets workflows
• GitHub issue and PR publishing
• database metadata imports
• JSON Schema / OpenAPI imports
• OpenLineage-style exports

These should remain bounded integrations. Integrations bring input; Martenweave stores model truth.