Skip to main content

One doc tagged with "summarization"

View all tags

Trimming and Summarization

An unbounded chat history eventually exceeds the model's context window, and even before that, it inflates every call's token cost and latency. Two standard strategies keep it in check.