You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
igardev edited this page Jul 12, 2026
·
1 revision
Skills
Overview
Llama-vscode support a simple auto memory for storing persistent data across conversations. The logic is as follows:
The persistent memory is stored by the agent (when it considers necessary or if the user explicitly asks) in .md files in a project specific auto memory folder (for example /home//.config/Code/User/workspaceStorage/<project_id>/ggml-org.llama-vscode/auto_memory). (The agent is asked to use descriptive file names)
On starting a chat with the agent, the list of the .md files names in the auto memory folder is added to the prompt.
The agent decides, based on the file name, what auto memory file to read (if any)
How to use it
Make sure the setting auto_memory_enabled is checked (checked by default)
Make sure the setting max_auto_memory_files is greater than 0 (10 by default)
Start using the agent - it will automatically store the information, which it considers important for the future conversation in auto memory (or the user could explicitly ask something to be saved in auto memory).
Settings:
auto_memory_enabled: If auto memory is enabled
max_auto_memory_files: The max number of files to keep in auto memory (this is a hint to the AI not a strict limit)