Skip to main content

​
Overview

A Conversation is a sequence of agent runs sharing the same context. When conversation memory is enabled, each new message includes the full conversation history, allowing the agent to maintain context across turns.

​
Creating conversations

Conversations are created automatically when you:
  • Use the Chat interface in the dashboard
  • Send a message via the API with a conversationId

​
How memory works

When EnableConversationMemory is enabled:
  1. The system loads all previous runs in the conversation
  2. User messages and agent responses are assembled into a message history
  3. The full history is sent to the LLM as part of the prompt
  4. The agent “remembers” everything discussed previously
When disabled, each message is processed in isolation — the agent has no memory of previous interactions.
Long conversations consume more tokens per message, as the entire history is sent with each request. Monitor your credit usage for high-volume conversations.

​
Chat interface

The built-in Chat interface provides a familiar messaging experience:
  • Real-time streaming responses via SignalR
  • Conversation history sidebar
  • Support for multiple concurrent conversations
  • Powered by a dedicated Chat system agent

​
Conversation management

  • View history — browse all messages and agent responses
  • Delete conversation — removes the conversation and all associated runs
  • Export — download conversation data for analysis