Skip to main content

Chat Run Data Flow

Status

Owner: Tech lead · Status: draft · Canonical source: training/diagrams/FleziGalaxy-Architectures/02-workflow-chat-run-sequence.html

Details the journey of user prompts through authentication, context injection, LLM generation, Redis Stream buffering, and real-time SSE streaming.

Sequence Diagram​

Existing Controls​

  • Input authorization checks (CH-02) verified before storing user input.
  • Bounded context loading prevents model context overflow and data leakage.
  • Automatic TTL expiration on transient Redis Streams.

Evidence (Files & Configuration)​

  • flezi.app.api.backend/app/features/dt_ai_runtime/routers/runtime.py
  • flezi.app.ai.agent/app/runtime/session.py

Limitations & Roadmap​

  • Real-time PII redaction scanner before prompt dispatch to external model endpoints.