Monitoring LLM Conversations

If you're building conversational agents or conversational systems such as LLM chatbots, you may want to view these responses as sequential messages. Confident AI allows you to view entire conversation threads from the observatory page.

caution

This feature is only available if a conversation_id was supplied to deepeval's monitor() function during live monitoring.

import deepeval

deepeval.monitor(
    ...,
    conversation_id="your-conversation-id"
)

If you're storing converation threads in a database, most of the time the conversation_id will be the ID of your conversation data row.

Viewing the Conversation

To review a specific conversation thread, select the conversation ID on the main observatory page. This thread will display all inputs and responses of that conversation, organized chronologically.

tip

Click the inspect button to further inspect each response, which triggers a "Response Details" side modal that provides detailed information about the associated response.

Leaving Feedback for Conversations

You may also leave conversation-level feedback by clicking the top right button on the conversations page. This feedback will appear on the human feedback page.

You may also view conversational-level feedback directly in the conversations page by clicking on the feeback icon on the left nav bar.

info

To learn more about providing feedback for your LLMs in production, read this guide.