SAEE Evolution Capability Router — DataHub Context Edition

• 

  DataHub OSS capability contract with schema, aliases, lifecycle, canonical tool, and staged-truth boundaries.

• 

  REUSE decision with canonical implementation, reason codes, lifecycle state, and explicit non-claims.

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for SAEE Evolution Capability Router — DataHub Context Edition

Inspiration

Coding agents are becoming capable of modifying complex repositories, but they need reliable knowledge of existing capabilities, lifecycle states, schemas, and canonical implementations. Without that context, they can generate a second implementation that increases maintenance and integration risk.

What it does

The project ingests a bounded capability contract into DataHub OSS. When a developer asks an AI coding agent to add a capability, the router uses the official DataHub Agent Context Kit to search the catalog, retrieve the exact entity, and inspect its schema. If DataHub returns an exact active implemented match, the router returns REUSE and generates an MCP configuration that points the coding agent to the canonical tool. Missing, design-only, or inactive matches stop at REVIEW_REQUIRED.

How we built it

• DataHub OSS v1.6.0 Quickstart
• datahub-agent-context==1.6.0.14
• Agent Context Kit tools: search, get_entities, list_schema_fields
• Python 3.12, dependency-free local web surface, pytest
• Apache-2.0 public repository

Challenges we ran into

DataHub indexing is asynchronous, so the client uses a bounded retry without falling back to local fixture results. The more important design challenge was preserving truth boundaries: a metadata match can recommend reuse, but it must not become automatic authorization or a production-readiness claim.

Accomplishments

• Real end-to-end DataHub OSS and Agent Context Kit integration
• Exact capability and alias matching
• Lifecycle-aware stop rules
• Deterministic generated configuration artifact
• Agent-readable contracts, evidence, and non-claims

What we learned

Metadata is most valuable to coding agents when it is not merely searchable, but carries lifecycle, canonical implementation, schema, and non-claim boundaries that change what the agent does next.

What's next

Evaluate the extension on additional public capability catalogs and contribute general capability-contract modeling guidance back to the DataHub community. No production or customer-validation claim is made by this roadmap.

Pre-existing work disclosure

SAEE and its canonical capability inventory predate this hackathon and are not claimed as new work. This submission is the new DataHub context extension created during the official period. It contains no copied SAEE implementation code, schemas, or MCP code, and it does not modify the SAEE mainline.

Built With

• datahub-oss