Right now, the Chat search works only on chat titles. Those titles are (1) auto-generated and (2) based solely on the first prompt in the conversation.
A common work pattern for me and my users is to open Glean once at the start of the day, begin a chat with whatever we need help with first, and then reuse the same chat window throughout the day for multiple, often unrelated, tasks. A typical daily chat might look like:
- Initial prompt about Topic A
- One or two follow-up prompts
- A new prompt about Topic B
- Four or five prompts related to Topic B
- Additional, unrelated prompts later in the day
While I sometimes open a new chat for a discrete piece of work, it is more common that a single chat contains several distinct “mini-sessions” of activity.
Problem
When I need to reference something from a prior chat—especially those “four or five related prompts” in the middle of a long thread—there is no reliable way to find it again. The only option today is to click through past chats one by one and manually search within each page. Because search only uses auto-generated titles based on the first prompt, it does not reflect the real content or the most important parts of the conversation.
Requested behavior
At minimum:
- Chat search should index and search the full chat body, not only the auto-generated titles.
Ideally:
- Recent chat history (and, in the future, pinned or saved chats beyond 30 days) should be indexed into my personal graph, so that:
- The Chat search box can scrub across my own prior conversations, and
- I can quickly retrieve specific prior exchanges based on the actual content of the conversation, not just the initial prompt.
This would align Chat search with real-world usage patterns, where a single chat thread often represents a day’s worth of evolving work rather than a single, tightly scoped task.
