> ## Documentation Index
> Fetch the complete documentation index at: https://apione.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Migrate an OpenAI-compatible application

> Change the endpoint, credentials, and model, then verify features incrementally.

Verify a minimal text request first, then test each capability your application needs. Compatible request formats do not make models' parameters and behavior identical.

## Change three settings

| Setting | APIONE configuration |
| - | - |
| Base URL | `https://api.apione.ai/v1` |
| API Key | A key created in the APIONE console |
| `model` | An ID available to the current key |

Switch the endpoint, key, and model together. Keep provider credentials paired with the correct destination. See the [SDK examples](https://apione.ai/docs/en/guides/sdks.md) and [model lookup](https://apione.ai/docs/en/models/catalog.md).

For native Anthropic or Gemini requests, use the corresponding [address and authentication](https://apione.ai/docs/en/api/overview.md). The OpenAI SDK address above does not apply to every protocol.

## Verify in stages

1. Test single-turn text and conversation history in a test environment.
2. Check output limits, finish reasons, and errors.
3. Verify required streaming, JSON, tools, and image input separately.
4. Compare quality, usage, and latency on the same tasks rather than one response.
5. Shift limited traffic, observe errors and costs, then expand.

APIONE permissions come from its API Key. OpenAI `organization` and `project` settings have no effect. Verify optional model parameters individually. See the [compatibility reference](https://apione.ai/en/api-docs/docs/openai-compat).

## Common migration issues

| Symptom | First check | Details |
| - | - | - |
| cURL works but the application fails | The SDK's actual URL, environment variables, duplicate paths, and proxy settings | [SDK setup](https://apione.ai/docs/en/guides/sdks.md) |
| Model unavailable or HTTP 403 | Model ID, key permissions, and billing group | [Model selection](https://apione.ai/docs/en/models/catalog.md) |
| Text works but streaming or tools fail | Model capabilities, event parsing, and optional parameters | [Streaming](https://apione.ai/docs/en/models/streaming.md), [tools](https://apione.ai/docs/en/models/advanced.md) |
| Charges differ from the previous provider | APIONE group prices, input/output usage, and additional charges | [Usage and billing](https://apione.ai/docs/en/account/usage.md) |

## Rollback and data

Retain the last verified configuration. Switch endpoint, key, and model together during rollback. Do not print credentials in logs or configuration comparisons.

Other gateways' backup domains, promotions, routing names, billing rules, workspace features, and data-retention commitments do not automatically apply to APIONE. Check its console and current public information.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.