Long conversations overflow the context window. What should you do?
Why A is the answer
Summarize older turns, and keep key facts
Summaries preserve meaning at lower token cost.
In detail
- Summaries keep meaning at a lower token cost.
- Keep exact details that matter.
Use case
An illustrative scenario, not a real incident.
A long chat nears the limit, so older turns are summarized, and key facts stay.
See it in different ways
flowchart TD
A{"How long must it last?"} -->|"This task"| B["Short-term: state and context"]
A -->|"Across sessions"| C["Long-term store"]
C --> D["Episodic: what happened"]
C --> E["Semantic: facts"]
C --> F["Procedural: how to act"]mindmap
root((Memory))
Short term
Current context
Summaries
Long term
Episodic
Semantic
Planning
Plan and execute
Revise on new factsflowchart LR S0["Summarize old turns"] --> S1["Keep key facts"] --> S2["Continue"]
Key point: Summaries save tokens.
Why the other options are wrong
- B. Delete the system prompt to free space
- The system prompt holds the rules.
- C. Raise the temperature on the model
- Temperature does not fix size.
- D. Keep sending all the history anyway
- That will fail at the limit.

