AI System Memory Profiler

Type into the mock chat below. Watch the context window actually fill up using this site's real chat configuration: an 8-message rolling history window, the same token estimator as the Token Counter, and a fixed system prompt allocation.

Context Window

System Prompt (est.) Active History (last 8) Dropped (out of window)

Stats

Messages sent0

In active window0

Dropped from history0

System prompt tokens (est.)-

Active history tokens0

Total tokens sent per request-

How This Site's Chat Actually Works

The real chat Worker keeps only the last 8 messages of history on every request (messages.slice(-8) in the client code), regardless of how long the conversation has run. Older messages aren't summarized, they're simply dropped from what the model sees. The system prompt's exact token count isn't something I can show precisely since it's set server-side and I don't have its exact text, the figure here is a reasonable estimate for a moderately detailed instruction prompt, not a measured value.