
Short answer
An OpenAI-compatible API reduces rewrite work — but compatibility alone does not answer data location, DPA or storage. Teams that need a DPA verify the base URL and the contract chain before moving production traffic.
When a drop-in is enough — and when EU inference is the hard constraint
A drop-in (same SDK calls, new base URL) is enough when tooling compatibility is the main goal. EU inference becomes a hard constraint when personal or sensitive data enters prompts and privacy requires location, counterparty and storage answers.
Base URL / chat completions checklist
- Documented base URL and auth?
/v1/chat/completions(and streaming if used) tested against your client?- Model IDs from the current catalogue, not old screenshots?
- Tool-calling / structured output verified in a loop (not just HTTP 200)?
- Empty or incomplete responses handled?
- DPA path, inference location, training and storage confirmed in writing?
Migration how-to: OpenAI-compatible API migration with the OpenAI SDK.
LLMBase snapshot
| Criterion | LLMBase |
|---|---|
| API | OpenAI-compatible, base URL https://api.llmbase.ai/v1. |
| Inference location | All Inference API models run in EU data centres. Countries: Germany, Finland, France. |
| DPA | Executable in the dashboard. → Sign the DPA in the dashboard |
| Training | No training on your data. |
| Storage | LLMBase does not store prompt or response content of the Inference API. |
| Telemetry | No prompts or responses in LLMBase telemetry. |
| Contract | German company; German counterparty, EU law. |
| Billing | Pay-per-use, no monthly fees. Prepaid balance in USD. |
| Data centres | ISO 27001-certified partner data centres. |
| Models | Live catalogue on LLMBase Inference API. 9 open-weight models. |
Switching from another provider
Change base URL and API key, pick a model ID from the live catalogue, keep your existing SDK client. Estimate costs on the pricing page.
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.LLMBASE_API_KEY,
baseURL: "https://api.llmbase.ai/v1",
});
const response = await client.chat.completions.create({
model: process.env.LLMBASE_MODEL_ID,
messages: [{ role: "user", content: "Hello!" }],
});
console.log(response.choices[0]?.message?.content);
FAQ
Do I need to rewrite my code?
Usually not. OpenAI-compatible API at https://api.llmbase.ai/v1 — change the base URL and select one of the 9 open-weight models from the live catalogue.
Where do models run?
In EU data centres; inference countries: Germany, Finland, France. ISO 27001-certified partner data centres.
Is there a DPA?
Yes — in the dashboard. Sign the DPA in the dashboard
Are prompts stored or used for training?
No training on your data. LLMBase does not store prompt or response content of the Inference API.
Further reading
- LLMBase Inference API
- OpenAI-compatible API migration: endpoints, models, and SDK setup
- LLMBase pricing
- Sign the DPA in the dashboard