Docs/Help/Model & Context The Agent Model & Context screen chooses the AI model behind the conversation, what users may attach, and how much of the conversation the agent carries forward. These settings affect answer quality, response time and token use. Found in the agent sidebar under Conversation → Model & Context.
Model
| |
|---|
| Conversation AI Model | Required. The model used for general conversation with users. Pick a model or Organization default (inherits the organization's default model, shown in brackets). The capabilities of the selected model (for example Image in, Document in, or Text only) are shown under the field. Also the default Answering AI model for Natural conversation voice calls. |
| Enforce Thinking | Makes the AI analyze and plan its approach before responding. Improves reasoning quality but may slightly increase response time. |
| Enable User to Change AI Model | Lets users switch between AI models during a conversation. |
| Allowed Conversation AI Models | Shown only when Enable User to Change AI Model is on. Required: pick at least one model users can choose from. If image upload is allowed, include at least one image-capable model. |
Attachments
| |
|---|
| Allow Image Upload | Lets users attach images in chat and in immersive voice calls. In chat the conversation model must accept images; a warning appears when it (or none of the allowed models) does not. Turn-by-turn voice calls always accept images; Natural conversation calls only when the Voice page's Answering AI model takes them. |
| Allow Document/File Upload | Lets users attach PDF, TXT and MD files in chat and in immersive voice calls. Text is extracted independently of the model, so model capability does not matter. |
| Document Upload Restrictions | Rules every uploaded image or document is checked against, in chat and voice alike, for example content that should be blocked for security or compliance. A violating file is refused with the reason. Blank allows all supported files. |
Context Window
| |
|---|
| Summarize Previous Answer | Lets the agent summarize its previous responses, keeping long conversations clear and saving space. |
| Keep Function Tools Answer | Keeps answers produced by tools or functions in the context, for accurate follow-up responses. |
| User Message to Carry Over | Required. How many previous user messages are kept in the conversation context. |
| Max Response Token | Hard ceiling on answer length in tokens. 0 follows the model's maximum output (no limit). Must be 0 or greater. |
| Personalization note | Points to the separate Personalization page, which now holds whether the agent remembers users and the model it uses for that. |
Saving
| |
|---|
| Save | Saves all three cards at once. |
Moved Elsewhere
Related guide: Model & Context