Model-generated tool invocation
Representing assistant output that requests a function or external operation.
AI Tools · Type Alias
ToolCallPart
ToolCallPart is the AI SDK content-part shape for a model-generated tool invocation. It identifies the call, names the target tool, and stores schema-shaped arguments.
How it works
ToolCallPart models one tool invocation inside prompt content. Its discriminator identifies the part, the call ID links it to the corresponding result, the tool name identifies the requested tool, and args carries the model-generated JSON-serializable inputs.
Syntax
export interface ToolCallPart {
type: 'tool-call';
toolCallId: string;
toolName: string;
args: unknown;
}Parameters
| Name | Type | Description |
|---|---|---|
type | 'tool-call' | Discriminator for this content part; always the literal 'tool-call'. |
toolCallId | string | Identifier used to associate the call with its result. |
toolName | string | Name of the requested tool. |
args | unknown | JSON-serializable arguments matching the tool input schema. |
Returns
A ToolCallPart object representing a tool call content part.
Key points
- A ToolCallPart is one tool-invocation segment within prompt content.
- toolCallId identifies this particular invocation.
- toolName states which tool the model requested.
- args carries the inputs supplied for the invocation.
- Arguments should be JSON-compatible and conform to the selected tool's schema.
Common mistakes
- Keep toolCallId stable so the invocation can be correlated with its result.
Example: Construct a tool-call part
const part: ToolCallPart = {
type: 'tool-call',
toolCallId: 'call_123',
toolName: 'weather',
args: { city: 'Paris' },
};
console.log(part.toolName, part.args);Creates a tool-call content part with an identifier, tool name, and JSON-compatible arguments.
When to use it
Call-result correlation
Linking a tool response to the invocation that caused it.
Schema-driven argument handling
Passing model-produced inputs into a tool-validation and execution pipeline.
Interview questions
QWhat is ToolCallPart used for?
ToolCallPart represents a tool-call content part of a prompt, usually produced by a model. It contains a 'tool-call' discriminator, a call identifier, the tool name, and the call arguments.
Follow-up: How would you associate a result with the request? Use the same toolCallId on the corresponding tool-result part.
QWhat does the type field signify?
The type field must be the literal 'tool-call'. This discriminates the object from other message-content part types.
Follow-up: Why is a literal discriminator useful? It lets TypeScript and runtime logic identify the union member reliably.
QWhat does args contain?
args is an unknown value representing the tool's arguments. The documentation describes it as a JSON-serializable object matching the tool's input schema.
Follow-up: How should consumers validate args? Check it against the tool's input schema before executing the operation.
Related
ToolResultPartTextPartAssistantContentModelMessage