Bloom — Decision-Aware dbt Model Generator

InspirationBloom is a metadata-aware code generation agent built for the Metadata-Aware Code Generation & Development track.

Most AI tools produce disposable SQL. Bloom treats the decision as a first-class artifact. Given a business question and an Architecture Decision Record, it:

1. Queries DataHub (MCP Server / Agent Context Kit / SDK) for relevant datasets, schemas, ownership, and lineage.
2. Generates production-quality dbt model SQL + schema.yml that respect the real catalog.
3. Emits a full Decision Record that travels with the model (SQL comments + meta fields + standalone markdown).

The result is code a data team would actually merge — because both the “what” and the “why” are present and grounded in the organization’s metadata graph.

Tech stack: Python, DataHub Agent Context Kit / MCP, OpenAI (optional), dbt, Pydantic.

Repo: https://github.com/Mototown/Bloom- Sample outputs are included in examples/output/dau_by_cohort/

Built With

• datahub
• dbt
• mcp
• openai
• pydantic
• python