endpoint-model-validation.md
1.8 KB
GPT Image 2 / WanJie Endpoint & Model Validation Notes
Session: 2026-07-24. Context: user provided a new WanJie API key and asked to verify whether the existing cron job prompt would still work without changes.
Key facts
- Valid key tested:
sk-<redacted>(stored in~/.bashrcasWANJIE_API_KEY). - The environment variable name is
WANJIE_API_KEY, notWANJIEDATA_API_KEY. - Termux has TLS issues with the WanJie image host; use
curl -skorrequests.post(..., verify=False).
Validation matrix
| Endpoint | Model name | Result | Notes |
|---|---|---|---|
https://api.wanjiedata.com/v1/images/generations |
gpt-image-2 |
HTTP 410 | Aliyun Web Application Firewall HTML page, not JSON. The cron job prompt currently uses this combination. |
https://maas-openapi.wanjiedata.com/api/v1/images/generations |
gpt-image-2 |
HTTP 400 model unauthorized |
Endpoint is correct, but model name is wrong. |
https://maas-openapi.wanjiedata.com/api/v1/images/generations |
"GPT Image 2" (with space) |
HTTP 200 | Returns { "data": [{ "url": "..." }] }. Image downloads successfully (~840 KB PNG). |
Conclusion
The cron job prompt must be updated to use the correct endpoint and the exact quoted model name "GPT Image 2". Using the new key with the old prompt (wrong endpoint + wrong model name) will fail.
Correct minimal curl
bash -i -c 'curl -sk -X POST https://maas-openapi.wanjiedata.com/api/v1/images/generations \
-H "Authorization: Bearer $WANJIE_API_KEY" \
-H "Content-Type: application/json" \
-d "{\"model\":\"GPT Image 2\",\"prompt\":\"...\",\"n\":1,\"size\":\"1024x1024\"}" \
| python3 -c "import sys,json; d=json.load(sys.stdin); print(d[\"data\"][0][\"url\"])"'
Use bash -i -c in cron/terminal so .bashrc is sourced and WANJIE_API_KEY is available.