Responses

​

Example

import (
    "context"
    "fmt"
    "log"

    aistats "packages/sdk-go"
    "packages/sdk-go/gen"
)

ctx := context.Background()
client := aistats.New("your-api-key", "https://api.ai-stats.dev")

req := gen.ResponsesRequest{
    Model: "openai/gpt-4.1",
    Input: []map[string]interface{}{
        {
            "role": "user",
            "content": []map[string]interface{}{
                {"type": "output_text", "text": "Summarise this text"},
            },
        },
    },
    Temperature: &[]float32{0.7}[0],
    Stream:      &[]bool{false}[0],
}

resp, _, err := client.GenerateResponse(ctx, req)
if err != nil {
    log.Fatal(err)
}
fmt.Println("Response:", resp.Output[0].Content[0].Text)

​

Streaming

// Note: Streaming is not yet implemented in the Go facade.
// Use the raw API service for streaming.

​

Key parameters

• model (required): Target model id.
• input (required): Ordered array of input items (messages, tool calls, etc.).
• temperature (0–2): Higher = more random.
• top_p (0–1) / top_k (>=1): Nucleus / k-best sampling controls.
• max_output_tokens (int): Hard cap on tokens generated per response; max_output_tokens_per_message to cap each message item.
• Tools: tools (definitions), tool_choice (auto/none/specific), max_tool_calls (int), parallel_tool_calls (bool).
• Logprobs: logprobs (bool), top_logprobs (0–20) to return per-token logprobs.
• Output: response_format (json/text), service_tier, store (bool), stream (bool).
• Metadata: metadata (object) for passthrough, reasoning (object) for effort hints.
• Gateway extras: usage (bool to request usage), meta (bool to include meta block).

​

Returns