
Announcements
We're using the Bot Framework Web Chat (webchat.js) component to connect users to our custom Copilot Agent (likely built on Copilot Studio or Bot Framework SDK) via Direct Line.
We have implemented a feature to persist chat history. When a user starts a chat session:
webchat.js library. The user can scroll up and see their previous interactions.The Problem:
While the history is displayed correctly in the frontend UI, the Copilot Agent (bot backend) itself seems to have no awareness or memory of this loaded history.
If the user asks a question referring to the past conversation that is visible on the screen (e.g., "What did we discuss about project X earlier?" or "Summarize our previous conversation"), the bot cannot answer correctly. It only seems aware of the messages exchanged within the current live session.
Our Goal:
We want the Copilot Agent to be aware of the historical conversation context that we load and display in the Web Chat UI. The bot should be able to use this context to answer user questions about past interactions or potentially use it for better context in the ongoing conversation.
Question:
What are the recommended approaches or best practices within the Bot Framework ecosystem to bridge this gap between the history displayed in the Web Chat client and the conversational context/memory available to the bot backend?
How can we effectively "inject" or make the bot backend aware of the conversation history loaded from our external database?
We're looking for solutions that would work with either a Copilot Studio bot (potentially using Plugins/Actions) or a bot built with the Bot Framework SDK.
Thanks in advance for any insights or suggestions!