cx-conversation-embedding-pipeline - Design a support conversation vectorisation pipeline
Designs a pipeline that vectorises support conversations for semantic search while managing chunking, PII redaction, and deletion.
Tags
Updated: 2026-09-18Capabilities
What this skill does
- Chunk conversations on turn boundaries
- Attach metadata filters to vectors
- Redact PII prior to embedding
- Propagate data deletions to index
- Trigger reindexing on content updates
- Evaluate retrieval recall across channels
Inputs
- Support conversation transcripts
- Conversation metadata
- PII redaction rules
- Labelled retrieval test set
Outputs
- Vectorised conversation index
- Retrieval evaluation metrics
- Pipeline design specification
Requirements
- Embedding model access
- Vector search index engine
- Data protection compliance review