Data Firefighter

• 

  Dashboard

Data incidents are rarely caused by one broken file. A deleted column can silently break pipelines, dashboards, ML features, and downstream datasets. Engineers normally have to manually trace lineage, identify owners, search repositories, determine root cause, and create a fix.

Data Firefighter turns that process into an AI-assisted incident response workflow.

When a column deletion is detected, Data Firefighter uses DataHub context to investigate the incident, trace downstream lineage, calculate blast radius, identify owners, and correlate affected assets with real repository code. An LLM then analyzes the root cause and generates a targeted SQL remediation.

The generated fix is validated before the agent creates a real GitHub branch, commit, and pull request. A human approval checkpoint prevents the agent from directly modifying production code.

The result: instead of an engineer spending hours investigating a data incident, the agent turns the incident into a validated, reviewable pull request.

Data Firefighter supports both a real DataHub GraphQL client and a deterministic mock adapter for reproducible demonstrations. The repository includes the real DataHub integration, tests, and metadata seeding workflow.

Built With

• datahub
• fastapi
• github
• graph-sql
• langgraph
• llm
• next.js
• python
• sqlite