Compaction replaces a long stretch of transcript with a shorter summary that preserves what later steps will need. It is what lets an agent run for hours in a context window that would otherwise have overflowed within the first twenty minutes.

Compaction

Preserve the goal, decisions taken and why, unresolved threads, and identifiers such as file paths and record ids. Discard the intermediate reasoning that led to a settled decision, which is the bulk of the tokens and almost never load-bearing again.

What a Good Summary Preserves

A topic of the Building AI Agents course. Planned for 1.5 hours at Hard difficulty.

University of Michigan - Ann Arbor

A category of the Building AI Agents course, covering 4 topics.

Memory, Retrieval and Context Engineering

A topic of the Building AI Agents course. Planned for 1.5 hours at Medium difficulty.

The Context Window as a Budget

A topic of the Building AI Agents course. Planned for 2 hours at Medium difficulty.

Learn Before

Related

Learn After