Skip to content

Latest commit

 

History

History
159 lines (123 loc) · 4.45 KB

Assistants.md

File metadata and controls

159 lines (123 loc) · 4.45 KB
Raw

Assistants

This guide provides step-by-step instructions on how to integrate OpenAI's Assistants API in a Kotlin application.

Prerequisites

  • OpenAI API Client for Kotlin installed.
  • An API key from OpenAI.

Steps to Follow

Step 1: Setting Up the Environment

Ensure that you have the necessary OpenAI SDK and dependencies in your Kotlin project. Set the OPENAI_API_KEY environment variable with your API key.

val token = System.getenv("OPENAI_API_KEY")
val openAI = OpenAI(token)

Step 2: Create an Assistant

val assistant = openAI.assistant(
    request = AssistantRequest(
        name = "Math Tutor",
        instructions = "You are a personal math tutor. Write and run code to answer math questions.",
        tools = listOf(AssistantTool.CodeInterpreter),
        model = ModelId("gpt-4-1106-preview")
    )
)

This code snippet creates an assistant named "Math Tutor" with instructions and tools specified.

Step 3: Create a Thread

val thread = openAI.thread()

Create a thread to start a conversation. Each user should have their own thread.

Step 4: Add a Message to the Thread

openAI.message(
    threadId = thread.id,
    request = MessageRequest(
        role = Role.User,
        content = "I need to solve the equation `3x + 11 = 14`. Can you help me?"
    )
)

Add a message to the thread. In this example, a math problem is posed.

Step 5: Run the Assistant

val run = openAI.createRun(
    thread.id,
    request = RunRequest(
        assistantId = assistant.id,
        instructions = "Please address the user as Jane Doe. The user has a premium account.",
    )
)

Create a run to process the user's message and generate a response.

Step 6: Check Run Status and Display Response

do {
    delay(1500)
    val retrievedRun = openAI.getRun(threadId = thread.id, runId = run.id)
} while (retrievedRun.status != Status.Completed)

val assistantMessages = openAI.messages(thread.id)
println("\nThe assistant's response:")
for (message in assistantMessages) {
    val textContent = message.content.first() as? MessageContent.Text ?: error("Expected MessageContent.Text")
    println(textContent.text.value)
}

Check the status of the run until it is completed, then display the assistant's response.

Conclusion

This guide covers the basics of setting up and using the OpenAI Assistants API in a Kotlin project. Remember to explore additional features and capabilities of the API for more advanced use cases.

Complete Example

Below is a complete Kotlin example following the guide:

suspend fun main() {

    // 1. Setup client
    val token = System.getenv("OPENAI_API_KEY")
    val openAI = OpenAI(token)
    
    // 2. Create an Assistant
    val assistant = openAI.assistant(
        request = AssistantRequest(
            name = "Math Tutor",
            instructions = "You are a personal math tutor. Write and run code to answer math questions.",
            tools = listOf(AssistantTool.CodeInterpreter),
            model = ModelId("gpt-4-1106-preview")
        )
    )

    // 3. Create a thread
    val thread = openAI.thread()

    // 4. Add a message to the thread
    openAI.message(
        threadId = thread.id,
        request = MessageRequest(
            role = Role.User,
            content = "I need to solve the equation `3x + 11 = 14`. Can you help me?"
        )
    )
    val messages = openAI.messages(thread.id)
    println("List of messages in the thread:")
    for (message in messages) {
        val textContent = message.content.first() as? MessageContent.Text ?: error("Expected MessageContent.Text")
        println(textContent.text.value)
    }

    // 5. Run the assistant
    val run = openAI.createRun(
        thread.id,
        request = RunRequest(
            assistantId = assistant.id,
            instructions = "Please address the user as Jane Doe. The user has a premium account.",
        )
    )

    // 6. Check the run status
    do {
        delay(1500)
        val retrievedRun = openAI.getRun(threadId = thread.id, runId = run.id)
    } while (retrievedRun.status != Status.Completed)

    // 6. Display the assistant's response
    val assistantMessages = openAI.messages(thread.id)
    println("\nThe assistant's response:")
    for (message in assistantMessages) {
        val textContent = message.content.first() as? MessageContent.Text ?: error("Expected MessageContent.Text")
        println(textContent.text.value)
    }
}