What problem does it solve? Generated PySpark and SQL code in an automated SDLC pipeline needs independent review for coding standards, performance, security, and Spark optimization before it reaches testing and deployment, and this agent performs that review as stage 10 of a 15-agent Databricks pipeline. ## Core Features & Use Cases - Independent code review: Calls the Databricks Foundation Model API a second time in reviewer mode with a separate system prompt and lower temperature, so the same call that wrote the code never grades it. - Contract-validated output: Returns JSON matching a strict output contract (status, artifact path, citations, tool calls, tokens) that the LangGraph supervisor validates, with one repair retry on failure. - Lakehouse artifact persistence: Writes review_report.md and revised code files to /Volumes/{catalog}/state/artifacts/{ticket}/10_code-review-agent/ and logs runs to state.artifacts and state.agent_runs. - Use Case: After the PySpark, SQL, and Data Quality agents produce their artifacts for a ticket, this agent reviews the code for partition pruning, broadcast joins, UDF avoidance, and secrets handling, then gates downstream agents on human approval. ## Quick Start Review the PySpark and SQL artifacts for ticket PROJ-123 against Databricks coding standards and write the review report and revised code to the ticket's artifact volume.
