aimodel

A Go SDK for AI model APIs with multi-protocol support (OpenAI, Anthropic). Zero external dependencies.

Design Scope

This SDK is a thin API wrapper — it translates requests, manages connections, and normalizes responses across protocols. It intentionally does not include retry, rate limiting, request validation, caching / persistence, logging / metrics. This keeps the SDK minimal and composable. Control mechanisms belong in the layer above, where you have full context over your application's requirements.

Usage

import "github.com/vogo/aimodel"

Set env vars AI_API_KEY and AI_BASE_URL (or OPENAI_API_KEY / OPENAI_BASE_URL).

Chat Completion

client, _ := aimodel.NewClient()

resp, _ := client.ChatCompletion(context.Background(), &aimodel.ChatRequest{
    Model: aimodel.ModelOpenaiGPT4o,
    Messages: []aimodel.Message{
        {Role: aimodel.RoleUser, Content: aimodel.NewTextContent("Hello!")},
    },
})

fmt.Println(resp.Choices[0].Message.Content.Text())

Streaming

stream, _ := client.ChatCompletionStream(context.Background(), &aimodel.ChatRequest{
    Model: aimodel.ModelOpenaiGPT4o,
    Messages: []aimodel.Message{
        {Role: aimodel.RoleUser, Content: aimodel.NewTextContent("Hello!")},
    },
})
defer stream.Close()

for {
    chunk, err := stream.Recv()
    if errors.Is(err, io.EOF) {
        break
    }
    fmt.Print(chunk.Choices[0].Delta.Content.Text())
}

Anthropic Protocol

Use WithProtocol to select the Anthropic Messages API:

client, _ := aimodel.NewClient(
    aimodel.WithAPIKey("sk-ant-xxx"),
    aimodel.WithProtocol(aimodel.ProtocolAnthropic),
)

resp, _ := client.ChatCompletion(context.Background(), &aimodel.ChatRequest{
    Model: aimodel.ModelAnthropicClaude4Sonnet,
    Messages: []aimodel.Message{
        {Role: aimodel.RoleUser, Content: aimodel.NewTextContent("Hello!")},
    },
})

The same ChatCompletion / ChatCompletionStream methods work for all protocols — routing is handled internally.

Client Options

client, _ := aimodel.NewClient(
    aimodel.WithAPIKey("your-key"),
    aimodel.WithBaseURL("https://api.example.com/v1"),
    aimodel.WithProtocol(aimodel.ProtocolAnthropic),
    aimodel.WithTimeout(30 * time.Second),
)

Multi-Model Compose

The composes package dispatches requests across multiple backends with failover, random, or weighted strategies:

import "github.com/vogo/aimodel/composes"

openai, _ := aimodel.NewClient(aimodel.WithAPIKey("sk-openai"), aimodel.WithBaseURL("https://api.openai.com/v1"))
anthropic, _ := aimodel.NewClient(
    aimodel.WithAPIKey("sk-ant"),
    aimodel.WithProtocol(aimodel.ProtocolAnthropic),
)

cc, _ := composes.NewComposeClient(composes.StrategyFailover, []composes.ModelEntry{
    {Name: "gpt-4o", Client: openai},
    {Name: "claude-sonnet", Client: anthropic},
})

resp, _ := cc.ChatCompletion(ctx, req)