Context
Context is the information given to a model at inference time, the system prompt, retrieved documents, conversation history, and tool outputs that shape its answer.
Category
Industry concept
See all terms
What is context?
Context is everything a model reads before it produces a response: the system prompt, the message history so far, any documents retrieved for the request, and the output of any tools it has called. Whatever is not included in this set doesn't exist for the model at the moment it answers.
Which other concepts does context relate to?
Context is the general concept. A context window measures the capacity: the token limit on how much context a model can hold in one request. Context engineering describes the practice: the discipline of deciding what context to keep in view during a run. A context layer names the infrastructure: the syworkstem that delivers context across every session an agent runs. A context repository names a store: where that context lives before it ever reaches a model. Context agents name a role: the agents that read and write to that store on request.
How does context reach a model?
At inference, a model receives one assembled sequence of tokens: the developer's system prompt, prior turns in the conversation, whatever a retrieval step pulled in, and the results of any tool calls made in the run. Everything competes for the same space and the same attention.
A model's training data is fixed once training ends, so context is the route by which anything current, private, or company-specific reaches it. Qontext's context repository is the store that supplies that knowledge, kept current and permissioned, so what reaches the model at inference time reflects what's actually true.