This is an extremely simple library that helps me play around with ollama.

• It only supports generating a completion.
• The only available parameter is the format one.
• The response is not streamed.

Prerequisites

1. Download and install ollama
2. Run the server ollama serve

Install

Follow this guide to setup your project using published packages.

The package for this library is: https://maven.pkg.github.com/le0nidas/ollama-kotlin-playground

Add the dependency to your build.gradle:

implementation("gr.le0nidas:ollama-kotlin-playground:0.0.2")

Usage

fun main() {
    val ollamaClient = OllamaClient()
    val request = GenerateRequest.Builder(Model("llama3.2"))
        .build(prompt = Prompt("Generate three numbers from 0 to 100"))
    val response = ollamaClient.generate(request)

    response
        .onSuccess { println(it.value) }
        .onFailure { println(it.message) }
}

/*
Here are three random numbers between 0 and 100:

1. 43
2. 91
3. 18
 */

or if you need to set a json schema for formatting the response:

fun main() {
    val ollamaClient = OllamaClient()
    val request = GenerateRequest.Builder(Model("llama3.2"))
        .withFormat(JsonSchemaFormat(
            buildJsonObject {
                put("type", "object")
                put("properties", buildJsonObject {
                    put("numbers", buildJsonObject {
                        put("type", "array")
                    })
                })
            }
        ))
        .build(prompt = Prompt("Generate three numbers from 0 to 100"))
    val response = ollamaClient.generate(request)

    response
        .onSuccess { println(it.value) }
        .onFailure { println(it.message) }
}

/*
{
    "numbers": [
        43,
        91,
        13
    ]
}
 */

FAQ

Q: Will it ever support anything else from the API?
A: Maybe. Depends on the needs of the experiments I do that involve ollama.

Q: Why did you write it then?
A: I didn't want to copy code from project to project.