add-derived-signal

To define custom code-ranking signals for RAG, you implement the DerivedSignalDescriptor interface to create normalized 0-1 metrics based on git history or static code structure, then register them in the trajectory provider's signal array.

Yes, you can use git metrics like churn and ownership to rerank code in RAG pipelines. The skill facilitates integrating these domain-specific git history metrics into the retrieval pipeline as normalized signals to surface stable code.

The best way to prioritize stable code over frequently modified hotspots is defining a custom signal that calculates recency from git history, dampening scores for high-churn files during the reranking phase of the retrieval pipeline.

Yes, TeaRAGs supports static code analysis metrics for reranking. You can implement flexible trajectory signals for static metrics like chunk size and documentation density alongside git-based metrics within the retrieval pipeline.

To register a new derived signal in TypeScript, create a file in the appropriate trajectory directory implementing the DerivedSignalDescriptor interface, then add the signal to the trajectory provider's signal array for pipeline integration.

You need custom normalized signals for code search when standard retrieval fails to surface specific code patterns, requiring domain-specific metrics like documentation quality or ownership to rank high-quality code accurately during reranking.