Glossary
What is a context window in AI?
A context window is the amount of text an AI language model can take into account at one time, measured in tokens, which are pieces of words. Everything the model considers, the instructions, the conversation and any documents, has to fit in it. Text outside the window is simply not seen.
Published by Sellinger AI. Numbers and settings checked on 24 October 2026 against the fact register. Report an error.
Why does the context window matter?
A sales agent needs its instructions, the facts about your product, what it knows about the prospect and the conversation so far. Long threads and large documents can exceed the window, and then something is left out. Larger windows help, but filling them with irrelevant text can also make answers worse.
How do systems work around the limit?
| Technique | What it does |
|---|---|
| Retrieval | Search the knowledge base and include only the relevant passages |
| Summaries | Replace old parts of a long thread with a short summary |
| Structured memory | Keep key facts, such as the prospect's role and objections, in fields |
| Focused instructions | Short, clear rules instead of long documents |
What does it mean for you?
- Keep knowledge base passages short and specific, so retrieval finds the right one.
- Do not paste whole manuals into a persona.
- Expect very long threads to need a person to read the history.
How Sellinger handles it
Sellinger's agent answers from the knowledge base you attach, which it searches for what is relevant to the question rather than reading the whole collection, together with your persona, rules and the conversation. See retrieval augmented generation and the knowledge base help article.
Common questions
What is a token?
Is a bigger context window always better?
Does the agent remember past conversations?
Sources and method
General AI practice. Sellinger knowledge base behaviour from the knowledge base help article, checked 24 October 2026.
Related: Retrieval augmented generation, Large language model, Prompt engineering, Knowledge base help.
