Using Memory

LLMs don’t retain memory between completions. In CodeCompanion, memory provides persistent, reusable context for chat buffers, via the notion of groups.

Once enabled, there are many ways that memory can be added to the chat buffer.

With the Chat Buffer

Simply modify the memory.opts.chat.default_memory value to reflect the group(s) you wish to automatically add to the chat buffer:

require("codecompanion").setup({
  memory = {
    opts = {
      chat = {
        default_memory = { "default", "claude "}
    },
  },
})

In the Chat Buffer

To add memory to an existing chat buffer, you can use the /memory slash command. This will allow multiple memory groups to be added at a time whilst also protecting against duplicate files.

From the Action Palette

There is also a Chat with memory action in the Action Palette. This lists all of the memory groups in the config that can be added to the chat buffer.

Clearing Memory

Memory can also be cleared from the chat buffer via the gM keymap. Although note, this will remove ALL context that's been designated as "memory".