Definition
The context window represents the maximum amount of text an LLM can consider at one time, measured in tokens. Think of it as the model's working memory: everything it needs to reference must fit within this window.
Example
If your context window is 128K tokens and you upload a 200-page document, the model can only "see" about 80 pages at once. Chunking strategies determine which parts of the document the model actually processes.
Why it matters
Context window limits determine what your agent can handle in a single interaction. Larger context windows mean more data per interaction, but also higher costs and potential quality degradation.