Ollama SDKs in Go: Overview and Code Examples

Here is a little guide to Ollama SDKs in Go: Packages, Usage, and Comparison.

Integrating Ollama models into Go applications is streamlined by several SDK options, each suited to different development needs. This article provides an overview of three primary Ollama SDKs for Go: the Official Ollama API package, the community go-ollama-sdk, and the OpenAI Go client configured for Ollama. For each, we describe its features and include practical code examples, concluding with a comparison to help developers choose the right fit.

1. The Official Ollama Go SDK (github.com/ollama/ollama/api)

This package is maintained by Ollama and provides a fully typed, comprehensive client for all Ollama REST API features. It includes methods for text generation, chat, model management, embeddings, and more. Its design ensures tight and reliable integration directly aligned with Ollama's API.

Features

• Typed request and response structures


• Support for streaming generation


• Model lifecycle management (pull, push, delete, list)


• Embedding generation


• Environment-based configuration

Code Example

Here’s a simple example using the official SDK to generate a response:

package main



import (

  "context"

  "fmt"

  "github.com/ollama/ollama/api"

)



func main() {

  ctx := context.Background()

  client := api.ClientFromEnvironment()



  req := &api.GenerateRequest{

    Model:  "llama3",

    Prompt: "Write a short poem about Go programming.",

    Stream: nil, // default streaming

  }



  err := client.Generate(ctx, req, func(resp api.GenerateResponse) error {

    fmt.Print(resp.Response)

    if resp.Done {

      fmt.Println("\n[Generation complete]")

    }

    return nil

  })

  if err != nil {

    panic(err)

  }

}

Strengths

• Full feature coverage


• Strong typing and JSON mapping


• Supports streaming, chat, embeddings, and model management

2. go-ollama-sdk (github.com/rozoomcool/go-ollama-sdk)

A community-supported, typed client designed specifically for Ollama's REST API. It offers straightforward methods to generate text, chat, stream responses, and manage models, making it suitable for developers who want an Ollama-centric SDK.

Features

• Typed API calls


• Generate, chat, and stream capabilities


• Model management (pull, list, delete)


• Designed for local or remote Ollama servers

Code Example

Generating a simple response:

package main



import (

  "fmt"

  "github.com/rozoomcool/go-ollama-sdk/ollama"

)



func main() {

  client := ollama.NewClient("http://localhost:11434")

  result, err := client.Generate("llama3", "Tell me a joke.")

  if err != nil {

    panic(err)

  }

  fmt.Println("Result:", result)

}

Streaming chat example:

client.GenerateStream("llama3", "Explain quantum physics.",

  func(chunk string) { fmt.Print(chunk) },

  func() { fmt.Println("\n[Stream finished]") },

)

Strengths

• Easy-to-use, Ollama-specific features


• Supports streaming and chat


• Model lifecycle management

3. OpenAI Go Client with Ollama's OpenAI-Compatible API (github.com/openai/openai-go)

Ollama offers an OpenAI-compatible API endpoint that can be accessed using the official OpenAI Go SDK. This approach leverages existing OpenAI tooling and ecosystems but requires configuring the base URL.

Features

• Compatibility with OpenAI's client libraries


• Supports chat, completion, and embeddings


• Easy to integrate if familiar with OpenAI tools

Code Example

Using the OpenAI SDK to query Ollama:

package main



import (

  "context"

  "log"

  "github.com/openai/openai-go"

  oaioption "github.com/openai/openai-go/option"

)



func main() {

  ctx := context.Background()

  client := openai.NewClient(

    oaioption.WithBaseURL("http://localhost:11434/v1"),

    oaioption.WithAPIKey("ollama"),

  )



  resp, err := client.Chat.Completions.New(ctx, openai.ChatCompletionNewParams{

    Model: "llama3",

    Messages: []openai.ChatCompletionMessageParamUnion{

      openai.UserMessage("Tell me a joke about Go."),

    },

  })

  if err != nil {

    log.Fatalf("Failed to generate: %v", err)

  }

  log.Println("Response:", resp.Choices[0].Message.Content)

}

Strengths

• Leverages the mature OpenAI ecosystem


• Easy if already using OpenAI SDKs


• Compatible with existing codebases

Comparison Summary

Final Thoughts

• For full-featured, reliable Ollama integration in Go, the official github.com/ollama/ollama/api SDK is the best choice. It offers comprehensive support for generation, chat, streaming, and model management, all in strongly typed Go code.




• For developers preferring simplicity and Ollama-specific features without needing to migrate to OpenAI, the community go-ollama-sdk provides an easy-to-use, well-typed client tailored to Ollama’s API.




• For projects already built around OpenAI tooling, or for those who want an ecosystem-wide compatible solution, the OpenAI Go SDK configured to point at Ollama's API is an effective option.

Choosing the right SDK depends on project requirements, familiarity, and the level of Ollama-specific features needed.

See More

• https://github.com/ollama/ollama/api


• https://github.com/rozoomcool/go-ollama-sdk


• https://github.com/openai/openai-go