Customer integration guide · November 4, 2026. For CCSwitch, the Codex / ChatGPT desktop apps, and other chat clients that support a custom Base URL.
0. Text group vs full group
| Group | Console name | Best for | What /v1/models lists |
|---|---|---|---|
| Text group (default for new customers) | APIrove-Text | CCSwitch / Codex / chat; can also generate images | GPT chat models; can also generate images (Codex defaults to Image 2.6) |
| Full group | APIrove-Unified | API calls, image generation, Claude | Text + Claude + Image |
- When creating a key: for mostly chat, choose APIrove-Text (it can also generate images; Codex defaults to Image 2.6); choose APIrove-Unified when you need broader capabilities such as Claude.
- Existing customer keys (including gift keys) are not rotated or rebound; they keep working in their original group.
- Keys created self-service or in bulk are bound to the text group by default. If you need full capabilities, create a separate full-group key in the console, or contact support to add one.
1. Set up once: Base URL + key
- Create a key in API key management and choose the right group (see the table above).
- In CCSwitch / Codex / a similar client, set the API address to:
https://apirove.com/v1
Authentication: Authorization: Bearer YOUR_API_KEY. Keep the key only in the client's key field or a local environment variable. Do not post it in group chats, screenshots, or public repositories.
2. Copyable prompts (send to your chat model / Codex)
When you need advanced features, or want the assistant to change your client configuration for you, copy the whole block below and send it.
Full version (recommended, one-click copy)
You are my API configuration assistant. Configure my client (CCSwitch / Codex / chat client) strictly according to the rules below, and keep applying them automatically in later conversations:
[Global configuration · set once]
- Base URL: https://apirove.com/v1
- API Key: use the APIrove key I provide
- Auth header: Authorization: Bearer <API_KEY>
[Key types]
- Text group (APIrove-Text, default for new customers): /v1/models lists GPT chat models and can also generate images (Codex defaults to Image 2.6); suited to CCSwitch / Codex
- Full group (APIrove-Unified): includes text, Claude, and image generation; use this group when you need broader capabilities
- Do not rotate old keys; if full capabilities are needed, create a separate full-group key
[Default chat]
- For desktop chat / text conversations, use GPT text models (for example gpt-6.1-sol, gpt-5.6-terra, gpt-6-astra, gpt-6-sol, gpt-6.1-terra)
- Do not use an image generation model (such as gpt-image-2.6-flare) as a chat / conversation model
[Image generation]
- Text-group or full-group keys both work; Codex defaults to Image 2.6 (gpt-image-2.6-flare)
- For image generation, use: POST https://apirove.com/v1/images/generations
- Example model: gpt-image-2.6-flare (also the default when model is omitted)
- If the current client does not support images/generations: give me a copyable curl / API request; do not force an image model through the chat endpoint
[Security]
- Do not put the key in a public repository, web front-end code, or screenshots
- When configuration is done, confirm in one short reply whether the Base URL, key type, default chat model, and image generation path are all in place
Short version (system / global prompt field)
If your client only has a short system prompt field, use this one:
APIrove: Base https://apirove.com/v1 + my key. For chat, prefer the text group APIrove-Text (GPT, e.g. gpt-6.1-sol / gpt-5.6-terra); do not use gpt-image-* as a chat model. For images, use a text-group or full-group key + POST /v1/images/generations. Codex defaults to Image 2.6 (gpt-image-2.6-flare); the API also uses it when model is omitted.
<API_KEY> with your own key, or enter the key separately in the client's key field (the latter is recommended, so you never paste the key into a conversation).3. Choosing the default chat model
- Text group: for chat, pick
gpt-6.1-sol/gpt-5.6-terraor similar; do not usegpt-image-*as a chat model. Codex image generation uses Image 2.6 by default. - Full group: the list may show GPT, Claude, and Image models together; for desktop chat, choose only GPT text models.
- Do not use
gpt-image-*as a conversation model.
For more tested models, see the model directory.
4. Image generation
Use a text-group or full-group key. Image generation does not go through the chat completions endpoint; call this instead (the curl below is an Images API example using Image 2.6, gpt-image-2.6-flare, which is also the default when model is omitted; images Codex generates while you chat default to Image 2.6 too):
curl https://apirove.com/v1/images/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-image-2.6-flare",
"prompt": "a simple red circle on white background",
"size": "1024x1024",
"n": 1
}'
gpt-image-2.6-flare) by default, with no extra key setup. To turn it off, clear the checkbox for that key on the console keys page.Image model directory: /en/models/#image. Both text-group and full-group keys can generate images. If you get a model-unavailable error (such as 404), first check whether image generation was turned off on the keys page.
Externally, it is still the same APIrove address and the same text-group or full-group key, and the model names on the customer side do not change (such as gpt-image-2, gpt-image-2.5-flare, gpt-image-2.5-sunburst, gpt-image-2.6-flare). When a route has occasional jitter, the gateway automatically switches to an available path; the client does not need to change the Base URL, model, or key.
5. Checklist
- Base URL =
https://apirove.com/v1 - New chat key → group
APIrove-Text(can also generate images; Codex defaults to Image 2.6); for broader capabilities →APIrove-Unified - Text-group chat with
gpt-6.1-soland similar returns 200; image generation is also available (Codex defaults to Image 2.6) - Text-group or full-group keys can call
POST /v1/images/generations - Existing customers' keys have not been rotated
- The key is not in any public repository or screenshot