AI-RAG-Assistant-Chatbot: Skill for Codex

Lumina Backend Api

Overview

Use this skill for HTTP and data-layer work in server/. Follow the existing route, middleware, model, and service split instead of collapsing logic into a single file.

Load The Right Reference

• Read references/file-map.md to locate the route, service, or model that owns a behavior.
• Read references/validation.md before finishing when request contracts, persistence logic, or API docs changed.

Change Backend Behavior In The Existing Layers

• Start with the route that exposes the behavior, then trace into services, models, middleware, and utility functions.
• Keep authentication logic in auth routes or middleware rather than scattering token checks.
• Keep conversation and guest behavior separate when the repository already separates those flows.
• Keep request and response structures stable unless the task explicitly changes the API contract.

Handle Contract Changes Explicitly

• Update openapi.yaml when endpoint behavior, payload fields, or response shapes change.
• If the work is really about knowledge ingestion, Pinecone source lifecycle, or citation formatting, switch to lumina-rag-knowledge.
• If frontend callers depend on the changed contract, update the client in the same task when feasible.

Finish With Real Validation

• Run npm run build in server/ after meaningful backend changes.
• Treat npm test in server/ as a placeholder, not as evidence of correctness.
• If live services or secrets were unavailable, state that clearly and separate compile-time validation from runtime confidence.