Migrate from OpenAI to RollTok in 5 Minutes: A Developer's Guide

If you're using OpenAI's API directly, you're probably paying more than you need to. Here's the reality:

The real win isn't just savings on OpenAI models—it's access to 49+ models through a single API. One key, one endpoint, one SDK.

That's it. You now have access to 49+ models.

Before:

from openai import OpenAI

client = OpenAI(api_key="sk-proj-...")

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)

After:

from openai import OpenAI

client = OpenAI(
    api_key="rolltok-YOUR_KEY",              # Change 1: API key
    base_url="https://api.rolltok.com/v1"     # Change 2: Base URL
)

response = client.chat.completions.create(
    model="gpt-4o",           # Or: claude-sonnet-4-5, gemini-3-flash, etc.
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)

That's it. Two lines changed.

import OpenAI from 'openai';

const client = new OpenAI({ apiKey: 'sk-proj-...' });

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: 'Hello' }]
});
python
import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: 'rolltok-YOUR_KEY',
  baseURL: 'https://api.rolltok.com/v1'
});

const response = await client.chat.completions.create({
  model: 'gpt-4o',  // Or any of 49+ models
  messages: [{ role: 'user', content: 'Hello' }]
});
curl https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer sk-proj-..." \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o","messages":[{"role":"user","content":"Hello"}]}'
curl https://api.rolltok.com/v1/chat/completions \
  -H "Authorization: Bearer rolltok-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o","messages":[{"role":"user","content":"Hello"}]}'

Here's where it gets interesting. Want to test Claude instead of GPT?

response = client.chat.completions.create(
    model="claude-sonnet-4-5",  # Just change the model name
    messages=[{"role": "user", "content": "Hello"}]
)

Want to try a cheaper model for simple tasks?

response = client.chat.completions.create(
    model="doubao-seed-2.0-mini",  # $0.03/1M tokens — 80% cheaper
    messages=[{"role": "user", "content": "Hello"}]
)

No new SDK. No new API key. No code restructuring.

Full list: rolltok.com/models

Here's a pattern we see developers use: route different tasks to different models based on complexity.

def smart_completion(prompt: str, complexity: str = "simple"):
    """Route prompts to cost-optimal models."""
    model_map = {
        "simple": "doubao-seed-2.0-mini",      # $0.03/1M — fast, cheap
        "medium": "gpt-4o-mini",                # $0.15/1M — balanced
        "complex": "claude-sonnet-4-5",         # $3.00/1M — high quality
        "reasoning": "deepseek-r1",             # $0.55/1M — chain-of-thought
    }

    response = client.chat.completions.create(
        model=model_map.get(complexity, "gpt-4o-mini"),
        messages=[{"role": "user", "content": prompt}]
    )
    return response.choices[0].message.content

summary = smart_completion("Summarize this article...", "simple")
analysis = smart_completion("Analyze this code for bugs...", "complex")
math = smart_completion("Solve this equation...", "reasoning")

Real cost impact: A developer processing 1M tokens/day might spend:

Run this quick test script:

import time

models_to_test = [
    "gpt-4o",
    "claude-sonnet-4-5",
    "gemini-3-flash",
    "doubao-seed-2.0-mini",
]

for model in models_to_test:
    start = time.time()
    response = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": "Say hello in one word"}],
        max_tokens=10
    )
    elapsed = time.time() - start
    print(f"✅ {model}: {response.choices[0].message.content} ({elapsed:.2f}s)")

Expected output:

✅ gpt-4o: Hello (0.83s)
✅ claude-sonnet-4-5: Hello (1.12s)
✅ gemini-3-flash: Hello (0.45s)
✅ doubao-seed-2.0-mini: Hello (0.67s)

If all four return responses, your migration is complete.

Yes. RollTok uses the OpenAI-compatible API format. Your prompts, system messages, function calling, streaming—all of it works identically.

Streaming works the same way:

stream = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}],
    stream=True
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="")

Supported on compatible models, same syntax:

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "What's the weather in Tokyo?"}],
    tools=[{
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get current weather",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {"type": "string"}
                },
                "required": ["location"]
            }
        }
    }]
)

Change two lines back. No lock-in. Your code is identical.

RollTok doesn't store prompts after processing. We don't use your data for training. For detailed privacy information, check rolltok.com/privacy.

Transparency matters. Here's what you should know:

If any of these are deal-breakers for your use case, we're happy to tell you to stick with your current provider until we ship those features.

Once you've migrated:

Disclosure: We're transparent about our product capabilities and limitations. The migration code above has been tested and works with the current RollTok API.