Hi everyone,
I've built a Glossary Agent designed to answer questions about company-wide acronyms to make explanations fast, clear, and accessible. Right now we have a Wiki, but my goal is to have a short explanation accessible via Glean through a Slack message.
Challenge:
When I try to trigger the Glossary Agent (either by typing a specific phrase that I've instructed the Agent to listen for or mentioning the Agent’s name in Slack), Glean’s default behavior in Slack kicks in and it searches for answers as usual. The results aren’t wrong, but they tend to be much more verbose and overwhelming for simple acronym queries, compared to the focused, concise response from the Glossary Agent.
What I’ve Tried:
- Calling the Agent in Slack:
I’ve invoked the Agent by name and with targeted phrases in Slack, but Glean still reverts to its general search behavior and gives the usual, comprehensive results.
When I run the same trigger in the Glean UI, the Agent works as intended. This suggests that in Slack, Glean defaults to its typical search utility instead of passing control to the Agent. - Testing a Non-Competing Agent:
I set up another Agent for a specific action that doesn’t overlap with Glean’s natural search triggers in Slack (e.g., using: @Glean summarize for JIRA). This works properly. The Agent listens for the phrase and completes the task as configured. - Explicit Agent Name Calls:
I also tested running other Agents by calling their names directly, and in those cases, everything works as expected.
Question:
Is there a way (maybe a setting, a different trigger phrase, or some other method) to ensure that the Glossary Agent runs instead of Glean’s default search when it’s explicitly invoked?
Any advice or best practices on getting Agent responses to take priority in targeted situations would be greatly appreciated!
Thanks in advance for your suggestions.
