Agentic AI Reference Architecture

Orchestration · Software component

Dialogue Flow Manager

Software componentOrchestrationOrchestration & Toolsarc:DialogueFlowManager

An orchestration component that manages multi-turn dialogue, tracking turn context and allowing users to digress from an in-progress task flow and later resume it where it paused.

Responsibility. Coordinates each conversational turn across intent handling, memory and response generation while tolerating non-linear dialogue.

Also known as: Multi-turn dialogue management, Conversation manager

reads; writesinvokesemits telemetry towritesemits telemetry toreceives data frominvokesis evaluated byinvokesis monitored byis routed to byis scaled byreceives data fromis routed to bywritesfails over tois evaluated bysends data toConversation State Store: reads; writesConversation State StoreLLM Inference Service: invokesLLM Inference ServiceMetrics Collector: emits telemetry toMetrics CollectorWorking Memory Buffer: writesWorking Memory BufferTrace Collector: emits telemetry toTrace CollectorConversational (Chat) Interface: receives data fromConversational (Chat) In…Intent Router: invokesIntent RouterOnline Evaluator: is evaluated byOnline EvaluatorPrompt Context Builder: invokesPrompt Context BuilderSLO Monitor: is monitored bySLO MonitorA/B Test Traffic Splitter: is routed to byA/B Test Traffic SplitterAutoscaler: is scaled byAutoscalerAutomatic Speech Recognizer: receives data fromAutomatic Speech Recogni…Canary Rollout Controller: is routed to byCanary Rollout ControllerSemantic Memory Store: writesSemantic Memory StoreTemplate Response Generator: fails over toTemplate Response Genera…Load Test Runner: is evaluated byLoad Test RunnerRich Message Renderer: sends data toRich Message Renderer+1 more (see relationships)
Direct neighbourhood (hover for relationship types)

Relationships

invokes dependency

reads dependency

writes dependency

emits telemetry to dynamic

is routed to by dynamic

receives data from dynamic

sends data to dynamic

fails over to control

is scaled by control

is evaluated by assurance

is monitored by assurance

Design guidance

Classification

Patterns
Multi-turn dialogue managementInterruptible task flow
Quality attributes
Interaction capability (ISO/IEC 25010)Reliability (ISO/IEC 25010 | NIST AI RMF: valid and reliable)
Risks mitigated
Rigid linear form-like dialog flowsTreating each query in isolation

Sources

  1. Ch10.1: T. Nguyen, "Conversational UI," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 10.1. ISBN: 9798244538229.