Service endpoints#
| Service | Base URL | Description |
|---|
| AI Model API | https://studio.dataeyes.ai | All model APIs: text, image, video, audio |
| Search & Reader API | https://openapi.dataeyes.ai | /v1/search, /v1/reader, /v1/document/parse, and more |
| Account API | https://www.dataeyes.ai/apex | Balance, usage, key management, and more. See Balance and usage API |
Use these endpoints from anywhere in the world. There is no need to switch regional nodes.When to add /v1#
This is the most common setup mistake. One rule covers it: OpenAI-compatible clients add /v1; Anthropic and Gemini clients do not.| Client / SDK | Base URL to enter |
|---|
| OpenAI SDK (Python, Node.js, Go, Java, and others) | https://studio.dataeyes.ai/v1 |
| OpenAI-compatible tools such as Codex, Cursor, Cline, OpenCode, OpenClaw, n8n, Chatbox | https://studio.dataeyes.ai/v1 |
| Anthropic SDK, Claude Code | https://studio.dataeyes.ai |
| Google GenAI SDK, Gemini CLI | https://studio.dataeyes.ai |
| Raw HTTP requests (cURL, Postman) | Use the full path, for example https://studio.dataeyes.ai/v1/chat/completions |
If a tool returns 404 with /v1, remove /v1 and try again, and vice versa. Tools append their own path to the base URL, so one /v1 too many or too few causes a 404.
SDK examples#
Request conventions#
Data format: Requests and responses are JSON, except file uploads, which use multipart/form-data. Send Content-Type: application/json.
Streaming: Set "stream": true in the request body. Responses are returned as Server-Sent Events (SSE).
Timeouts: Reasoning models and long generations can take several minutes. Set your client read timeout to at least 600 seconds. For video and other async tasks, submit a task and poll for the result.
Request body size: Up to 128 MB per request. For large images or videos, pass a public URL instead of Base64.
User-Agent: The service sits behind a security layer. Keep your SDK's default User-Agent. If your own HTTP requests return 403, set a User-Agent header.
Next steps#