Skip to main content
Every model has a context window. This is the amount of conversation, file content, and tool output the model can consider for one reply. Babbily shows a context indicator near the composer so you can see how full the current chat is.

​
What the indicator means

The context indicator estimates how much of the active model’s window your chat is using. The details are approximate because models count conversation, file, and tool context differently.
  • Used shows the percentage consumed.
  • Tokens shows the estimated context used out of the model’s window.
  • Remaining shows the approximate space left for the next message and reply.
  • Compacts at shows when Babbily may summarize older parts of the chat.
The indicator updates as you send messages, attach files, receive replies, and switch models. When you use Auto, the indicator measures your chat against the smallest context window among the models Auto can choose, so it reflects the room every possible reply has.

​
How attachments use context

An attached file reaches the model as text, so its content takes up context like any other part of the chat. Once Babbily has read a file, the indicator reflects the space that content actually takes rather than a flat estimate for the file. Attachments get a share of the window rather than all of it. When files would not fit, Babbily shows the most recent ones in full and shortens the rest to an opening section, so a long thread keeps working instead of failing the message. If you need a specific part of a long file, ask about that part directly or attach only the pages or sheet that matter.

​
Automatic compaction

In saved chats, when a long thread approaches the selected model’s limit, Babbily can compact older conversation history into a shorter summary. Your visible chat history remains available, and the current conversation can keep going. A summary may omit details, so restate exact requirements when they matter. Compaction helps:
  • Preserve recent messages in detail.
  • Keep the thread within the active model’s limit.
  • Continue long-running work without forcing a new chat immediately.
A compacted summary is still part of the model’s context. Your usage percentage may drop, but it will not reset to zero.
Temporary Chats do not use automatic compaction. If they reach the limit, start a saved thread to continue.

​
If context preparation fails

If Babbily cannot safely prepare the conversation for another reply, it may stop and ask you to try again. Refresh the thread before retrying. If the issue persists, start a new saved chat with the essential context rather than assuming all earlier details carried over. Compaction is separate from saved memory: it summarizes a thread for continued work and does not mean its contents were added to your Memory Profile.

​
Switching models

Different models support different context sizes. If you switch to a smaller model, Babbily recalculates the thread against that model’s limit. If needed, Babbily may compact the thread before the next reply. With Auto, the answering model can change from one message to the next. That change alone does not make Babbily compact your chat. Saved chats still compact as usual when they approach the limit.

​
When a chat gets too full

If a thread reaches the limit, Babbily may ask you to start a new thread. Memory and plugins can still help you carry important context forward.
Long files, pasted code, large ZIP archives, and many follow-ups about a Canvas or research report can fill context quickly. Start a new thread when your topic changes or when you no longer need earlier details.
Last modified on October 7, 2026