Skip to content

Context Window

Every AI model has a context window — a limit on how much text it can hold in its working memory at once. Once a conversation fills the context window, earlier messages start to fall outside what the model can "see", which affects the quality of its responses.

The context window visualiser in OptimaChat shows you how much of this memory is currently in use.

Opening the visualiser

The context window visualiser is in the settings panel on the right side of the chat interface. If the settings panel is collapsed, click the panel handle to expand it.

What the visualiser shows

The visualiser shows the current conversation's token usage broken into four colour-coded segments:

Colour Segment What it represents
Teal System The system prompt tokens
Violet Tools Tokens used by tool definitions currently enabled
Blue User Tokens from user messages
Purple AI Tokens from assistant responses

The total token count is shown below the bar.

Switching views

Use the toggle at the top of the visualiser to switch between two layouts:

  • Bar view — a single horizontal bar showing the proportion of each segment
  • Grid view — a more detailed breakdown showing the size of each segment side by side

Context window visualiser in grid view showing each segment as a separate block with its token count

What "zoomed" means

If the conversation is long and the visualiser is showing a proportional view, it may display a note such as Showing first ~80% of max context. This means the bar is showing only a portion of the total context for readability — the full conversation is still being sent to the model.

Approaching the limit

As the context fills up, watch for the bar becoming nearly full. When the context window is exhausted:

  • The model will no longer have access to the earliest messages in the conversation
  • Response quality may degrade as important context drops out
  • You may want to start a new conversation to give the model a clean slate

The context window size is set by your administrator based on the hardware available. Larger context windows require more VRAM or RAM and may reduce the number of concurrent users the system can support.

Tools and context

Tool definitions consume context window space even when no tools have been used yet. If you have many tools enabled and the context window is becoming a concern, disabling unused tools from the tool panel will free up some space.