aws/context-ontology-accelerator ? reverse-engineered prompt

Build me a working version of this semantic context platform for AI agents.

I want it to let a user connect data sources, ingest documents and schemas, turn that into a searchable knowledge graph, and then serve trusted context back to an agent so it can make better decisions. It should include the main flow of scanning data, modeling the ontology, and serving queries, plus a simple web app for managing namespaces, roles, sources, and ontology content. Please make the API and backend work together cleanly, and keep the permissions model clear so different people can only see what their role allows.

If there are any missing details, use the repo docs and look up current AWS, Smithy, or Cloudscape guidance online if needed. I mainly want something that feels real and usable, with sensible defaults, good error handling, and enough structure that it could grow into a proper product.