Model input
Provide conversation history to generateText or streamText.
AI Tools · Type Alias
ModelMessage
ModelMessage is the AI SDK Core type for conversation messages sent to models. It unifies system, user, assistant, and tool messages, including multimodal content and tool interactions.
How it works
ModelMessage is the common message type accepted by AI SDK Core operations. It is a union of system, user, assistant, and tool messages, with content shapes appropriate to each role; the SDK also exposes a Zod schema for validating the type.
Syntax
type ModelMessage = SystemModelMessage | UserModelMessage | AssistantModelMessage | ToolModelMessage;Key points
- ModelMessage is the fundamental input message structure for AI SDK Core functions.
- The type covers the message variants accepted in Core functions' messages fields.
- Use the modelMessageSchema export when runtime Zod validation is needed.
- SystemModelMessage represents system-level information.
- UserModelMessage supports text-only content and multimodal text, image, or file content.
- AssistantModelMessage can represent generated text, tool calls, or both.
- ToolModelMessage carries results produced by tool calls.
Common mistakes
- For functions that provide a top-level instructions option, prefer that option rather than putting instructions in a system message.
Example: Define model messages
import { ModelMessage } from 'ai';
const messages: ModelMessage[] = [
{ role: 'system', content: 'You are a helpful assistant.' },
{ role: 'user', content: 'Explain TypeScript generics.' },
{ role: 'assistant', content: 'Generics let types work with a variety of values.' },
];This creates a typed conversation containing system, user, and assistant messages.
When to use it
Multimodal prompts
Represent multimodal user prompts containing images or files.
Tool workflows
Describe assistant tool calls and subsequent tool results.
Related
SystemModelMessageUserModelMessageAssistantModelMessageToolModelMessageTextPartImagePartFilePartToolCallPart