Retrieval-Grounded Support Deflection
The rag-deflection-bot repo models support deflection as a governed operating surface. It makes knowledge freshness, escalation boundaries, and risky reply scenarios explicit instead of assuming every customer question should be answered by automation.
Primary purpose
Reduce L1 support load by routing low-risk requests through grounded knowledge answers while preserving fast escalation for billing, identity, integration, and duplication-risk scenarios.
Application shape mapping
src/app.ts serves the operator shell. src/data/sampleDeflection.ts models knowledge and case posture. src/services/ragDeflectionService.ts exposes API-ready payloads. src/services/render.ts renders the control-plane routes.
Commercial interpretation
For GTM leaders, this shows how self-service support can protect gross margin without silently shifting customer risk into unsupported automation. For technical reviewers, it shows the concrete guardrails needed before a RAG assistant touches real customer workflows.