Spaces:
Sleeping
Sleeping
Agent memory
This file is the agent's persistent memory across conversations. The user can view and edit it from the UI.
The agent should read this at the start of every conversation and update it when it learns something durable about the user, the project, or the work in progress.
User profile
(filled in over time — role, background, what they're studying)
Project context
(what cases the user is currently focused on, deadlines, paper figures they're trying to reproduce)
Preferences
(default parameter values the user prefers, plotting choices, naming conventions for runs)
Findings & decisions
(non-obvious things the agent learned: parameter values that work / fail, what the user has rejected, prior optimization results worth remembering)
Open questions
(stuff the user said they'd come back to)