http-transforms

HTTP Transforms Example

Customize HTTP tool behavior with argument transforms, request hooks, and response processing.

What You'll Learn

• WithTransform() to map/normalize arguments before the HTTP request
• WithPreRequest() to inject headers, correlation IDs, or dynamic auth
• WithPostProcess() to transform responses after receiving
• WithRedact() to strip sensitive fields before sending to the LLM

Prerequisites

• Go 1.21+
• OpenAI API key

Running the Example

export OPENAI_API_KEY=your-key
go run .

Code Overview

Argument Transform

Map LLM arguments to the format your API expects:

cfg := sdktools.NewHTTPToolConfig("https://api.example.com/users",
    sdktools.WithTransform(func(args map[string]any) (map[string]any, error) {
        name, _ := args["name"].(string)
        parts := strings.SplitN(name, " ", 2)
        return map[string]any{
            "first_name": parts[0],
            "last_name":  parts[1],
        }, nil
    }),
)
conv.OnTool("lookup_user", cfg.Handler())

Pre-Request Hook

Modify the HTTP request before sending (auth, tracing, etc.):

cfg := sdktools.NewHTTPToolConfig("https://api.example.com/orders",
    sdktools.WithPreRequest(func(req *http.Request) error {
        req.Header.Set("X-Correlation-ID", generateID())
        req.Header.Set("Authorization", "Bearer "+getToken())
        return nil
    }),
)
conv.OnToolCtx("get_orders", cfg.HandlerCtx())

Response Post-Processing + Redaction

Strip sensitive fields and add computed values:

cfg := sdktools.NewHTTPToolConfig("https://api.example.com/customers",
    sdktools.WithRedact("ssn", "credit_card"),
    sdktools.WithPostProcess(func(resp []byte) ([]byte, error) {
        var data map[string]any
        json.Unmarshal(resp, &data)
        data["display_name"] = data["first_name"].(string) + " " + data["last_name"].(string)
        return json.Marshal(data)
    }),
)
conv.OnTool("get_customer", cfg.Handler())

Processing Pipeline

For each tool call, the pipeline runs in order:

1. Transform - Normalize arguments (WithTransform)
2. Pre-Request - Modify HTTP request (WithPreRequest)
3. Execute - Send HTTP request
4. Post-Process - Transform response (WithPostProcess)
5. Redact - Strip sensitive fields (WithRedact)

Key Concepts

1. Argument Mapping - Bridge the gap between LLM output and API input
2. Dynamic Auth - Inject tokens at request time (OAuth refresh, etc.)
3. Field Redaction - Prevent sensitive data from reaching the LLM
4. Computed Fields - Enrich responses before the LLM sees them

Next Steps

• MCP Tools Example - MCP server integration
• A2A Agent Example - Remote agent integration with auth
• SDK HTTP Auth Example - More auth patterns