cvm-troubleshooter

To troubleshoot CVM pipeline failures, run the diagnostics script to collect a failure report, which classifies issues by stage such as missing raw CSVs for ingestion or dbt errors for prepare-data, and then applies a minimal rerun command.

CVM pipeline training or scoring failures with missing outputs often stem from target mart issues or empty queues and quality gates. Collecting diagnostics identifies the precise root cause and verifies the output path after applying a minimal rerun.

Yes, you can diagnose data pipeline incidents across local and CI environments using DuckDB dependencies. The process collects diagnostics from ingestion, e2e, and drift stages, returning a concise incident report with the root cause and verification path.

The best way to unblock a failed CVM pipeline is to collect diagnostics first, classify the failure by stage, and then propose the smallest rerun command with minimal impact. This returns an incident report detailing the exact fix and verification output path.

Yes, the CVM pipeline diagnostics process handles drift and reason quality or feature availability issues. It collects failure reports and applies minimal reruns to resolve these specific stage failures and verify the resulting outputs.

You should not use a minimal rerun approach when a full end-to-end workflow failure occurs that cannot be isolated to a single stage like ingestion, prepare-data, training, scoring, e2e, or drift, as the diagnostics report requires a specific stage classification.