Does this project support VLLM?

[NTDLS]

I've tried various settings and I can't seem to get any further than this.

VisualChatGPTStudio Output

6/27/2026 12:44:54 AM: Request URI: http://192.168.0.167:5001/v1/chat/completions
6/27/2026 12:44:54 AM: Request Method: POST
6/27/2026 12:44:54 AM: Request Headers:
6/27/2026 12:44:54 AM: Accept-Encoding: gzip, deflate
6/27/2026 12:44:54 AM: Connection: keep-alive
6/27/2026 12:44:54 AM: Authorization: Bearer 721BCE18-B403-403E-B81F-6772FB248F08

6/27/2026 12:44:54 AM: api-key: 721BCE18-B403-403E-B81F-6772FB248F08

6/27/2026 12:44:54 AM: User-Agent: Visual_chatGPT_Studio
6/27/2026 12:44:54 AM: Request Content: {"model":"Qwen/Qwen2.5-Coder-7B-Instruct-AWQ","messages":[{"role":"system","content":"Explain"},{"role":"user","content":"ConfigurationBuilder"}],"n":1,"stream":false,"max_tokens":2048}
6/27/2026 12:44:54 AM: ____________________________________________________________________________________________________
6/27/2026 12:44:54 AM: Response Headers:
6/27/2026 12:44:54 AM: connection: close
6/27/2026 12:44:54 AM: Date: Sat, 27 Jun 2026 04:44:13 GMT
6/27/2026 12:44:54 AM: Server: uvicorn
6/27/2026 12:44:54 AM: Response Status Code: BadRequest
6/27/2026 12:44:54 AM: Response Content: Invalid HTTP request received.
6/27/2026 12:44:54 AM: ____________________________________________________________________________________________________
6/27/2026 12:44:54 AM: Newtonsoft.Json.JsonReaderException: Error parsing Infinity value. Path '', line 1, position 2.
   at JeffPires.VisualChatGPTStudio.Utils.ApiHandler.<GetResponseAsync>d__4.MoveNext() in C:\Users\j.da.paz.pires\OneDrive - Avanade\Source\github\VisualChatGPTStudio\VisualChatGPTStudioShared\Utils\ApiHandler.cs:line 117
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw()
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at JeffPires.VisualChatGPTStudio.Commands.BaseGenericCommand`1.<RequestAsync>d__10.MoveNext() in C:\Users\j.da.paz.pires\OneDrive - Avanade\Source\github\VisualChatGPTStudio\VisualChatGPTStudioShared\Commands\BaseGenericCommand.cs:line 139
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw()
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at JeffPires.VisualChatGPTStudio.Commands.BaseGenericCommand`1.<ExecuteAsync>d__9.MoveNext() in C:\Users\j.da.paz.pires\OneDrive - Avanade\Source\github\VisualChatGPTStudio\VisualChatGPTStudioShared\Commands\BaseGenericCommand.cs:line 79

[jeffdapaz]

Hi!
I've never personally tested the extension with vLLM, but based on feedback from other users, it's definitely possible to use it, the only prerequisite is that the server is OpenAI-compatible, which vLLM is. In fact, looking at your log, the request the extension built is perfectly valid OpenAI-format JSON and correctly hit the standard /v1/chat/completions endpoint, so conceptually you're on the right track.
That said, the error you're seeing is actually two separate issues stacked on top of each other:

1. The real problem — your server rejected the request (HTTP 400):

Server: uvicorn
Response Status Code: BadRequest
Response Content: Invalid HTTP request received.

"Invalid HTTP request received." comes from uvicorn's low-level HTTP parser (the server vLLM runs on), not from vLLM itself. This means the request was rejected at the HTTP transport level — it never even reached the model or the JSON body. So the model field and payload were never the issue.

2. The misleading exception — Error parsing Infinity value:

Newtonsoft.Json.JsonReaderException: Error parsing Infinity value. Path '', line 1, position 2.

This is just a symptom, not the cause. The extension expects a JSON response, but the server returned plain text (Invalid HTTP request received.). The JSON parser reads In... and assumes it's about to parse the literal Infinity, then fails at position 2 (v instead of f). So this exception is simply the extension choking on a non-JSON error body, it's hiding the real 400.

How to troubleshoot:

Test the exact same request with curl against your vLLM server. If curl works but the extension doesn't, the difference lies in what the extension adds to the request, most likely a proxy setting, or the duplicate auth headers it's sending (Authorization: Bearer ... and api-key: ..., because the request is going out in Azure-style configuration). Try switching the service type to OpenAI (not Azure), clearing any proxy, and confirm vLLM is reachable over plain HTTP on that port.

Let me know how it goes!