ollieb89/AI Root Cause Hints
Explain why your AI pipeline failed — not just that it failed. Correlates signals from eval, RAG, cost, PR quality, and output safety actions.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| github-token | GitHub token for posting PR comments | no | ${{ github.token }} |
| eval-failures | Number of eval failures (from ai-workflow-evals) | no | — |
| eval-suite | Eval suite name/path | no | — |
| rag-deletion-pct | Percentage of RAG corpus deleted/changed | no | — |
| rag-path | RAG corpus path | no | — |
| cost-spike | Cost spike multiplier (e.g. 2.5 = 2.5x normal spend) | no | — |
| model | LLM model name | no | — |
| pr-quality-score | PR quality score 0-100 (from ai-pr-guardian) | no | — |
| redaction-hits | Number of redaction hits (from ai-output-redacter) | no | — |
| redact-mode | Redaction mode used | no | — |
| risk-level | PR risk level: low | medium | high (from pr-context-enricher) | no | — |
| complexity-score | PR complexity score (from pr-context-enricher) | no | — |
| signals-json | Custom signals as JSON array [{source, type, severity, value?, metadata?}] | no | — |
| pr-number | PR number for comment posting | no | — |
| post-comment | Post findings as a PR comment | no | true |
| fail-on-findings | Fail the workflow step when findings are detected | no | false |
Outputs
| name | description |
|---|---|
| hints-count | Number of root cause hints generated |
| has-findings | Whether any correlated failure patterns were found |
| top-confidence | Confidence level of top finding: HIGH | MEDIUM | LOW | none |
| signal-count | Total number of signals analyzed |