Context Windows Explained
7 min read
Think of a context window as the model’s working memory. Everything in it costs tokens: system instructions, chat history, memories, tool schema, and document chunks.
Spend it well
- Put durable rules in memory, not repeated prompts.
- Use file ingestion once, then reference summaries.
- Prefer tools over long inline answers when state updates matter.
When GreatChat summarizes long work, it preserves intent instead of growing forever. Manage your preferences in memory.
Try this in GreatChat
Everything in this article works inside your assistant — connect an app and go.