Skip to content

Conversation Monitoring

The Chat/Test page itself also serves as a monitoring center. On the left, the conversation list displays all active or recent threads in which the Agent is involved — whether opened through UI, WhatsApp, SIP or API.

1. List overview

ColumnWhat it indicates
Title / IDIf the thread already received an automatic title (first user message) it appears; otherwise, the short id (thr_…).
Trigger iconUI, API, WhatsApp logo, SIP icon, etc. Helps identify the origin channel.
StatusActive, Running, Closed, Pending.
TimeDate/time of the last message.

Retention

For Free accounts, conversations are kept for 7 days and then automatically removed.

2. Accessing details

Click on any item in the list to open the complete history in the center panel. Behavior varies according to the trigger:

TriggerUI InteractionNotes
UIReading and sending messagesStandard test environment.
APIReading; optional sending if application allowsUseful for debugging backend integrations.
WhatsAppRead onlyAvoids duplicating messages or violating template policies.
SIP / WebRTCRead onlyReal-time audio is handled in voice channel; UI shows transcripts.

Even in read-only mode you can:

  • Close the conversation (⋮ → Close).
  • Reopen if necessary.
  • View context metrics, used tools and costs in the Sidebar.

3. Status indicators

StatusDescriptionAvailable actions
PendingThread created but no user messages.Close.
ActiveWaiting for next message.Send (if UI/API), Close.
RunningAgent processing or transmitting response.Wait; possible to force Close.
ClosedConversation finished; post-analysis executed.Reopen or delete.

4. Common use cases

4.1 Multi-channel support

Quickly filter by Trigger = WhatsApp to see all ongoing support on that channel and last interaction time.

4.2 Cost audit

Open a long thread's Sidebar to check:

  • Tokens spent / model limit.
  • Tools triggered (e.g., RAG calls, external APIs).
  • Total cost accumulated.

4.3 Debug API integrations

If the application sent messages via /stream, the list shows real-time status updates (Running). If something gets stuck, you can Close the thread manually and inspect the message log to identify the failed step.

4.4 View internal steps

Through the monitoring interface you can see all steps the Agent executed, including tool calls, partial responses and error messages. This helps understand the Agent's reasoning flow and identify possible improvements in instructions or tools.