Auto memory

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

1. Make sure the setting auto_memory_enabled is checked (checked by default)
2. Make sure the setting max_auto_memory_files is greater than 0 (10 by default)
3. 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)