[Bug]: NewAPI model fetching fails when local proxy is required

[sanqixy-sudo]

Module

Agent Detection & Connection

Platform

Windows

AionUi Version

2.2.2

Bug Description

AionUI fails to fetch the model list from an OpenAI-compatible/NewAPI endpoint, while the exact same endpoint and API key work normally in Codex, Cherry Studio, and curl when using my local proxy.

AionUI shows:

Backend POST /api/provider/fetch-models failed (502)
{"success":false,"error":"Upstream service unavailable.","code":"BAD_GATEWAY"}

The API endpoint itself is working correctly. The issue appears to be that AionUI's backend model-fetching request does not use or inherit my local/system proxy.

Base URL:
https://api.7cloudvps.com/v1

Local HTTP proxy:
http://127.0.0.1:7897

Steps to Reproduce

1. Start a local HTTP proxy at 127.0.0.1:7897.
2. Open AionUI on Windows.
3. Go to Models and add a Custom/OpenAI-compatible provider.
4. Set the API URL to https://api.7cloudvps.com/v1.
5. Enter a valid API key.
6. Try to fetch/search the model list.
7. AionUI returns 502 BAD_GATEWAY and cannot load the models.

For comparison, direct curl access also fails:

curl.exe -4 -v --http1.1 "https://api.7cloudvps.com/v1/models" -H "Authorization: Bearer <API_KEY>"

Result:
Recv failure: Connection was reset
schannel: failed to receive handshake, SSL/TLS connection failed

But the same request works correctly through the local proxy:

curl.exe -x http://127.0.0.1:7897 -i "https://api.7cloudvps.com/v1/models" -H "Authorization: Bearer <API_KEY>"

Result:
HTTP/1.1 200 Connection established
HTTP/1.1 200 OK

The full model list is returned successfully.

The same NewAPI endpoint also works normally in Codex and Cherry Studio.

Expected Behavior

AionUI's backend model-fetching request should use or inherit the configured/system proxy.

When a proxy such as http://127.0.0.1:7897 is available, requests to /v1/models should go through that proxy, just like Codex, Cherry Studio, and curl.

Ideally, AionUI should support HTTP_PROXY / HTTPS_PROXY or provide a dedicated per-provider proxy option for OpenAI-compatible/NewAPI providers.

Actual Behavior

AionUI appears to connect directly to the upstream API when executing /api/provider/fetch-models.

Because direct access to this endpoint is reset during the TLS handshake on my network, AionUI returns:

502 BAD_GATEWAY
Upstream service unavailable

The same endpoint works immediately when accessed through my local HTTP proxy at 127.0.0.1:7897.

So the behavior appears to be:

AionUI backend fetch-models
→ direct connection
→ TLS connection reset
→ 502 BAD_GATEWAY

While:

Codex / Cherry Studio / curl
→ local proxy 127.0.0.1:7897
→ NewAPI
→ 200 OK

Additional Context

This does not appear to be a general NewAPI compatibility issue.

I have another NewAPI endpoint that AionUI can fetch models from normally.

The problematic endpoint returns the following models successfully when accessed through the proxy:

kimi-k2
kimi-k2-thinking
kimi-k2.5
kimi-k2.6
kimi-k2.7-code
kimi-k2.7-code-highspeed
kimi-k3

Please check whether the AionUI backend/AionCore process responsible for /api/provider/fetch-models correctly supports or inherits proxy settings.

A dedicated proxy configuration for OpenAI-compatible providers would also be very useful.