Dev Tools

These tools and techniques reduce friction during development — from typed completions in your editor to request-level tracing in the dashboard.

Typed SDK clients#

The openai SDK ships full TypeScript types. Even if you write Python or Go, your IDE autocompletes request parameters, response fields, and error shapes — catching mistakes before they hit the wire.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://opendunes.com/v1",
  apiKey: process.env.OPENDUNES_API_KEY!,
});

// `resp` is fully typed — no casting needed
const resp = await client.chat.completions.create({
  model: "anthropic/claude-opus-4.8",
  messages: [{ role: "user", content: "Hello" }],
});

const text: string = resp.choices[0].message.content ?? "";

Request tracing with X-Request-Id#

Every response from the OpenDunes API includes an X-Request-Id header containing a UUID v7 trace identifier. Log it with every request so you can look up the exact call in your dashboard if something goes wrong.

import os
from openai import OpenAI
import httpx

# Intercept the raw HTTP response to extract headers
class TracingTransport(httpx.HTTPTransport):
    def handle_request(self, request):
        response = super().handle_request(request)
        request_id = response.headers.get("X-Request-Id", "")
        print(f"[trace] X-Request-Id: {request_id}")
        return response

http_client = httpx.Client(transport=TracingTransport())
client = OpenAI(
    base_url="https://opendunes.com/v1",
    api_key=os.environ["OPENDUNES_API_KEY"],
    http_client=http_client,
)

Note. When contacting support about a failed request, always include the X-Request-Id. It's the fastest way to locate the call server-side.

The OpenDunes dashboard#

Your dashboard gives you a live view of:

• API keys — create, rotate, or revoke keys; set a spending cap in DA on each.
• Usage — per-key token spend, request counts, and DA balance history.
• Balance — current balance in DA, with a breakdown by deposit and consumption.
• Request log — recent requests with model, tokens, latency, and trace ID.

Use the trace ID from X-Request-Id to jump directly to a specific request in the log.

Environment variables#

Keep secrets out of source code. The one variable you always need:

export OPENDUNES_API_KEY=sk-0a1b2c3d4e5f...

For local development, a .env file works well — just make sure it's in .gitignore.

# .env
OPENDUNES_API_KEY=sk-0a1b2c3d4e5f...
OPENDUNES_BASE_URL=http://localhost:8000/v1

Balance headers at a glance#

One header lets you monitor spend inline without extra API calls:

There is no cap on the number of requests, so there is no request counter to watch. A 429 is the provider's and carries Retry-After; see Limits.

Local dev base URL#

Point at http://localhost:8000/v1 when running the stack locally. Everything — auth, billing, models — behaves identically to production.

import os

base_url = os.environ.get("OPENDUNES_BASE_URL", "https://opendunes.com/v1")
client = OpenAI(base_url=base_url, api_key=os.environ["OPENDUNES_API_KEY"])

Also see#

• Usage for Agents — patterns for agent loops.
• Authentication — key scopes and management.
• Errors — error codes and how to handle them.