Skip to main content

Context compression

The context window is the AI's short-term memory: it limits the total information the model can remember and process in a conversation. Qoder IDE currently supports a context length of 200K.

How it Works

Context compression makes the current conversation's context usage visible and helps you choose an appropriate action.

Compact Chat

  • When to use it: You are continuing work on the same topic, the conversation has become long, and the context window is nearly full.
  • What to do: Click Compact Chat to summarize the conversation. The summary retains key code logic and essential context while removing redundant content.
  • Result: Lower Credits consumption and faster AI responses.

New Chat

  • When to use it: You are switching to a new topic or task, such as moving from fixing a bug to developing a new feature.
  • What to do: Click New Chat to start a fresh conversation.
  • Result: Unrelated conversation history no longer interferes with responses or wastes Credits.
Context compression can affect generation quality. To protect quality, it is disabled early in a conversation, when Credits savings would be small, and while the AI is generating a response, when compression could interrupt or affect generation.