Prompts

This page adapts the original AI SDK documentation: Prompts.

Prompts are instructions that you give a large language model (LLM) to tell it what to do. It’s like when you ask someone for directions; the clearer your question, the better the directions you’ll get.

Many LLM providers offer complex interfaces for specifying prompts. They involve different roles and message types. While these interfaces are powerful, they can be hard to use and understand.

In order to simplify prompting, the AI SDK supports text, message, and system prompts.

Text Prompts

Text prompts are strings. They are ideal for simple generation use cases, e.g. repeatedly generating content for variants of the same prompt text.

You can set text prompts using the prompt property made available by AI SDK functions like streamText or generateObject. You can structure the text in any way and inject variables, e.g. using a template literal.

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  prompt: "Invent a new holiday and describe its traditions."
)
print(result.text)

You can also use template literals to provide dynamic data to your prompt.

import SwiftAISDK
import OpenAIProvider

let destination = "Kyoto"
let lengthOfStay = 5

let result = try await generateText(
  model: openai("gpt-4o"),
  prompt: "I am planning a trip to \(destination) for \(lengthOfStay) days. " +
          "Please suggest the best tourist activities for me to do."
)
print(result.text)

System Prompts

System prompts are the initial set of instructions given to models that help guide and constrain the models’ behaviors and responses. You can set system prompts using the system property. System prompts work with both the prompt and the messages properties.

import SwiftAISDK
import OpenAIProvider

let destination = "Kyoto"
let lengthOfStay = 5

let result = try await generateText(
  model: openai("gpt-4o"),
  system: "You help planning travel itineraries. Respond to the users' request with a list of the best stops to make in their destination.",
  prompt: "I am planning a trip to \(destination) for \(lengthOfStay) days. " +
          "Please suggest the best tourist activities for me to do."
)
print(result.text)

Note: When you use a message prompt, you can also use system messages instead of a system prompt.

Message Prompts

A message prompt is an array of user, assistant, and tool messages. They are great for chat interfaces and more complex, multi-modal prompts. You can use the messages property to set message prompts.

Each message has a role and a content property. The content can either be text (for user and assistant messages), or an array of relevant parts (data) for that message type.

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .text("Hi!"))),
    .assistant(AssistantModelMessage(content: .text("Hello, how can I help?"))),
    .user(UserModelMessage(content: .text("Where can I buy the best Currywurst in Berlin?")))
  ]
)
print(result.text)

Instead of sending a text in the content property, you can send an array of parts that includes a mix of text and other content parts.

Warning: Not all language models support all message and content types. For example, some models might not be capable of handling multi-modal inputs or tool messages. Learn more about the capabilities of select models.

Provider Options

You can pass through additional provider-specific metadata to enable provider-specific functionality at 3 levels.

Function Call Level

Functions like streamText or generateText accept a providerOptions property.

Adding provider options at the function call level should be used when you do not need granular control over where the provider options are applied.

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  providerOptions: [
    "openai": [
      "reasoningEffort": .string("low")
    ]
  ]
)
print(result.text)

Message Level

For granular control over applying provider options at the message level, you can pass providerOptions to the message object:

import SwiftAISDK

let messages: [ModelMessage] = [
  .system(SystemModelMessage(
    content: "Cached system message",
    providerOptions: [
      // Sets a cache control breakpoint on the system message
      "anthropic": [
        "cacheControl": .object(["type": .string("ephemeral")])
      ]
    ]
  ))
]

Message Part Level

Certain provider-specific options require configuration at the message part level:

import SwiftAISDK

let messages: [ModelMessage] = [
  .user(UserModelMessage(
    content: .parts([
      .text(TextPart(
        text: "Describe the image in detail.",
        providerOptions: ["openai": ["imageDetail": .string("low")]]
      )),
      .image(ImagePart(
        image: .url(URL(string: "https://github.com/vercel/ai/blob/main/examples/ai-core/data/comic-cat.png?raw=true")!),
        providerOptions: ["openai": ["imageDetail": .string("low")]]
      ))
    ])
  ))
]

User Messages

Text Parts

Text content is the most common type of content. It is a string that is passed to the model.

If you only need to send text content in a message, the content property can be a string, but you can also use it to send multiple content parts.

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .parts([
      .text(TextPart(text: "Where can I buy the best Currywurst in Berlin?"))
    ])))
  ]
)
print(result.text)

Image Parts

User messages can include image parts. An image can be one of the following:

• base64-encoded image:
  • string with base-64 encoded content
  • data URL string, e.g. data:image/png;base64,...
• binary image:
  • ArrayBuffer
  • Uint8Array
  • Buffer
• URL:
  • http(s) URL string, e.g. https://example.com/image.png
  • URL object, e.g. new URL('https://example.com/image.png')

Example: Binary image (Buffer)

import SwiftAISDK
import OpenAIProvider
import Foundation

let data = try Data(contentsOf: URL(fileURLWithPath: "./data/comic-cat.png"))
let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .parts([
      .text(TextPart(text: "Describe the image in detail.")),
      .image(ImagePart(image: .data(data)))
    ])))
  ]
)
print(result.text)

Example: Base-64 encoded image (string)

import SwiftAISDK
import OpenAIProvider
import Foundation

let base64 = try Data(contentsOf: URL(fileURLWithPath: "./data/comic-cat.png")).base64EncodedString()
let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .parts([
      .text(TextPart(text: "Describe the image in detail.")),
      .image(ImagePart(image: .string(base64)))
    ])))
  ]
)
print(result.text)

Example: Image URL (string)

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .parts([
      .text(TextPart(text: "Describe the image in detail.")),
      .image(ImagePart(image: .url(URL(string: "https://github.com/vercel/ai/blob/main/examples/ai-core/data/comic-cat.png?raw=true")!)))
    ])))
  ]
)
print(result.text)

File Parts

Warning: Only a few providers and models currently support file parts: Google Generative AI, OpenAI (for wav and mp3 audio with gpt-4o-audio-preview), Anthropic, OpenAI (for pdf).

User messages can include file parts. A file can be one of the following:

• base64-encoded file:
  • string with base-64 encoded content
  • data URL string, e.g. data:image/png;base64,...
• binary data:
  • ArrayBuffer
  • Uint8Array
  • Buffer
• URL:
  • http(s) URL string, e.g. https://example.com/some.pdf
  • URL object, e.g. new URL('https://example.com/some.pdf')

You need to specify the MIME type of the file you are sending.

Example: PDF file from Buffer

import SwiftAISDK
import GoogleProvider
import Foundation

let pdf = try Data(contentsOf: URL(fileURLWithPath: "./data/example.pdf"))
let result = try await generateText(
  model: google("gemini-1.5-flash"),
  messages: [
    .user(UserModelMessage(content: .parts([
      .text(TextPart(text: "What is the file about?")),
      .file(FilePart(
        data: .data(pdf),
        mediaType: "application/pdf",
        filename: "example.pdf"
      ))
    ])))
  ]
)
print(result.text)

Example: mp3 audio file from Buffer

import SwiftAISDK
import OpenAIProvider
import Foundation

let mp3 = try Data(contentsOf: URL(fileURLWithPath: "./data/galileo.mp3"))
let result = try await generateText(
  model: openai("gpt-4o-audio-preview"),
  messages: [
    .user(UserModelMessage(content: .parts([
      .text(TextPart(text: "What is the audio saying?")),
      .file(FilePart(data: .data(mp3), mediaType: "audio/mpeg"))
    ])))
  ]
)
print(result.text)

Custom Download Function (Experimental)

You can use custom download functions to implement throttling, retries, authentication, caching, and more.

The default download implementation automatically downloads files in parallel when they are not supported by the model.

Custom download function can be passed via the experimental_download property:

import SwiftAISDK
import OpenAIProvider

let customDownload: DownloadFunction = { requests in
  return try await withThrowingTaskGroup(of: (Int, DownloadResult?).self) { group in
    for (i, req) in requests.enumerated() {
      group.addTask {
        if req.isUrlSupportedByModel { return (i, nil) }
        let (data, mediaType) = try await download(req.url) // add auth/retries as needed
        return (i, DownloadResult(data: data, mediaType: mediaType))
      }
    }
    var results = Array<DownloadResult?>(repeating: nil, count: requests.count)
    for try await (i, res) in group { results[i] = res }
    return results
  }
}

let result = try await generateText(
  model: openai("gpt-4o"),
  experimentalDownload: customDownload,
  messages: [
    .user(UserModelMessage(content: .parts([
      .file(FilePart(
        data: .url(URL(string: "https://api.company.com/private/document.pdf")!),
        mediaType: "application/pdf"
      ))
    ])))
  ]
)

Note: The experimental_download option is experimental and may change in future releases.

Assistant Messages

Assistant messages are messages that have a role of assistant. They are typically previous responses from the assistant and can contain text, reasoning, and tool call parts.

Example: Assistant message with text content

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .text("Hi!"))),
    .assistant(AssistantModelMessage(content: .text("Hello, how can I help?")))
  ]
)

Example: Assistant message with text content in array

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .text("Hi!"))),
    .assistant(AssistantModelMessage(content: .parts([
      .text(TextPart(text: "Hello, how can I help?"))
    ])))
  ]
)

Example: Assistant message with tool call content

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .text("How many calories are in this block of cheese?"))),
    .assistant(AssistantModelMessage(content: .parts([
      .toolCall(ToolCallPart(
        toolCallId: "12345",
        toolName: "get-nutrition-data",
        input: .object(["cheese": .string("Roquefort")])
      ))
    ])))
  ]
)

Example: Assistant message with file content

Note: This content part is for model-generated files. Only a few models support this, and only for file types that they can generate.

import SwiftAISDK
import OpenAIProvider
import Foundation

let data = try Data(contentsOf: URL(fileURLWithPath: "./data/roquefort.jpg"))
let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .text("Generate an image of a roquefort cheese!"))),
    .assistant(AssistantModelMessage(content: .parts([
      .file(FilePart(data: .data(data), mediaType: "image/png"))
    ])))
  ]
)

Tool messages

Note: Tools (also known as function calling) are programs that you can provide an LLM to extend its built-in functionality. This can be anything from calling an external API to calling functions within your UI. Learn more about Tools in the next section.

For models that support tool calls, assistant messages can contain tool call parts, and tool messages can contain tool output parts. A single assistant message can call multiple tools, and a single tool message can contain multiple tool results.

import SwiftAISDK
import OpenAIProvider
import Foundation

let image = try Data(contentsOf: URL(fileURLWithPath: "./data/roquefort.jpg"))
let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .user(UserModelMessage(content: .parts([
      .text(TextPart(text: "How many calories are in this block of cheese?")),
      .image(ImagePart(image: .data(image)))
    ]))),
    .assistant(AssistantModelMessage(content: .parts([
      .toolCall(ToolCallPart(
        toolCallId: "12345",
        toolName: "get-nutrition-data",
        input: .object(["cheese": .string("Roquefort")])
      ))
    ]))),
    .tool(ToolModelMessage(content: [
      .toolResult(ToolResultPart(
        toolCallId: "12345",
        toolName: "get-nutrition-data",
        output: .json(value: .object([
          "name": .string("Cheese, roquefort"),
          "calories": .number(369),
          "fat": .number(31),
          "protein": .number(22)
        ]))
      ))
    ]))
  ]
)

Multi-modal Tool Results

Warning: Multi-part tool results are experimental and only supported by Anthropic.

Tool results can be multi-part and multi-modal, e.g. a text and an image. You can use the experimental_content property on tool parts to specify multi-part tool results.

import SwiftAISDK
import OpenAIProvider
import Foundation

let png = try Data(contentsOf: URL(fileURLWithPath: "./data/roquefort-nutrition-data.png")).base64EncodedString()
let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .tool(ToolModelMessage(content: [
      // JSON output for non-multipart models
      .toolResult(ToolResultPart(
        toolCallId: "12345",
        toolName: "get-nutrition-data",
        output: .json(value: .object([
          "name": .string("Cheese, roquefort"),
          "calories": .number(369),
          "fat": .number(31),
          "protein": .number(22)
        ]))
      )),
      // Multipart content for supported models (e.g., Anthropic)
      .toolResult(ToolResultPart(
        toolCallId: "12345",
        toolName: "get-nutrition-data",
        output: .content(value: [
          .text(text: "Here is an image of the nutrition data for the cheese:"),
          .media(data: png, mediaType: "image/png")
        ])
      ))
    ]))
  ]
)

System Messages

System messages are messages that are sent to the model before the user messages to guide the assistant’s behavior. You can alternatively use the system property.

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  messages: [
    .system(SystemModelMessage(content: "You help planning travel itineraries.")),
    .user(UserModelMessage(content: .text("I am planning a trip to Berlin for 3 days. Please suggest the best tourist activities for me to do.")))
  ]
)