[GH-ISSUE #16581] issue: Open WebUI/LMStudio compatibility #17965

Closed
opened 2026-04-19 23:51:56 -05:00 by GiteaMirror · 7 comments
Owner

Originally created by @b9AcE on GitHub (Aug 13, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16581

Check Existing Issues

  • I have searched the existing issues and discussions.
  • I am using the latest version of Open WebUI.

Installation Method

Docker

Open WebUI Version

0.6.22

Ollama Version (if applicable)

No response

Operating System

Nobara 42/Windows 10

Browser (if applicable)

Firefox 141.0.3/Brave 1.81.131

Confirmation

  • I have read and followed all instructions in README.md.
  • I am using the latest version of both Open WebUI and Ollama.
  • I have included the browser console logs.
  • I have included the Docker container logs.
  • I have provided every relevant configuration, setting, and environment variable used in my setup.
  • I have clearly listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc).
  • I have documented step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation. My steps:
  • Start with the initial platform/version/OS and dependencies used,
  • Specify exact install/launch/configure commands,
  • List URLs visited, user input (incl. example values/emails/passwords if needed),
  • Describe all options and toggles enabled or changed,
  • Include any files or environmental changes,
  • Identify the expected and actual result at each stage,
  • Ensure any reasonably skilled user can follow and hit the same issue.

Expected Behavior

LM Studio is expected to be usable as backend for Open WebUI, since it is listed in the documentation for Open WebUI.

Actual Behavior

Configuring LM Studio as OpenAI API compatible server in Open WebUI results in errors in LM Studio and no models listed/usable in Open WebUI.

Steps to Reproduce

Since Ollama chose to actively break their GGUF-support https://github.com/ollama/ollama/issues/11714 and haven't visibly acted to repair their own mess, I would now like to migrate to using LM Studio's integrated llama.cpp to run the LLMs served by Open WebUI instead of Ollama, but I can not get the endpoint configuration to work.

What may be complicating the matter is that in my current (completely functional apart from Ollama's recent GGUF-breakage) setup I run Open WebUI on a single-board computer, an Organge Pi 5 Plus running DietPi to then serve Open WebUI through a composerized docker, while the LLM is run by Ollama on a dual-boot Nobara 42 (Fedora-based) and Windows 10 desktop machine because it has a GPU (RTX 4070 SUPER) which of course the single-board computer doesn't (can't).
The reason I want to use Open WebUI, as I am, is because I also serve it through a cloudflared tunnel making it reachable from the Internet, so I can access my LLMs from anywhere, so just using LM Studio directly would not provide the desired functionality. The reason I don't want to run llama.cpp myself direcrly but instead rely on LM Studio's is that they take care of a lot of chores automatically like managing llama.cpp, hardware compatibility layers, etc and I don't have docker installed on the desktop.

With Ollama on that desktop machine, let's call it "desktop", I put (pseudonymized) http://desktop.network.internal:11434 in the "Settings" → "Connections" → "Manage Ollama API Connections" → "Edit Connection" → "URL" field and that has worked for many months,
but when I turn on LM Studio's server mode (with "Serve on Local Network" and "Enable CORS" turned on for maximum compatibility) and input the equivalent http://desktop.network.internal:51234 with changed port matching what LM Studio says "2025-08-13 15:58:39 [ERROR]
Unexpected endpoint or method. (GET /api/version). Returning 200 anyway" when clicking "Verify Connection" in LM Studio and when I then save anyway to go check in "New Chat" on Open WebUI there are no models listed, even though there are several models (working fine) in LM Studio directly, but I keep getting new errors in LM Studio when I click on stuff in Open WebUI, like these:

 Unexpected endpoint or method. (GET /api/tags). Returning 200 anyway
2025-08-13 15:59:55 [ERROR]
 Unexpected endpoint or method. (GET /api/ps). Returning 200 anyway
2025-08-13 16:00:48 [ERROR]
 Unexpected endpoint or method. (GET /api/version). Returning 200 anyway

I have seen on both Open WebUI's website (e.g. https://docs.openwebui.com/getting-started/quick-start/starting-with-openai-compatible#step-1-choose-an-openai-compatible-server) and at LM Studio's website (e.g. https://lmstudio.ai/docs/app/api/endpoints/rest) that they should work together, but I can't get it to work together.
In desperation I have also tried adding variants of "/v1/api", "/v1", "/v0", "/api" and more on the endpoint URL specified in Open WebUI's settings, but if that is the issue I must have not stumbled on the correct one.

So, I don't know what is wrong, but either the documentation or the Open WebUI handling of the configured value seems most likely to me to need modification to reflect current reality (if it's the docs) or handle LM Studio's current implementation (if it's the software itself).

Logs & Screenshots

Docker on "verify" in Open WebUI connection
openwebui-1 | 2025-08-13 14:41:48.439 | INFO | uvicorn.protocols.http.httptools_impl:send:476 - [Public Internet IP redacted]:0 - "POST /ollama/verify HTTP/1.1" 200

LM Studio:

 Unexpected endpoint or method. (GET /api/tags). Returning 200 anyway
2025-08-13 15:59:55 [ERROR]
 Unexpected endpoint or method. (GET /api/ps). Returning 200 anyway
2025-08-13 16:00:48 [ERROR]
 Unexpected endpoint or method. (GET /api/version). Returning 200 anyway

Browser Console during a "verify" in Open WebUI (seems extremely irrelevant, but mandatory to include in this GitHib issue template, so it is):

NotFoundError: No such JSProcessActor 'BrowserToolboxDevToolsProcess'

Additional Information

No response

Originally created by @b9AcE on GitHub (Aug 13, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/16581 ### Check Existing Issues - [x] I have searched the existing issues and discussions. - [x] I am using the latest version of Open WebUI. ### Installation Method Docker ### Open WebUI Version 0.6.22 ### Ollama Version (if applicable) _No response_ ### Operating System Nobara 42/Windows 10 ### Browser (if applicable) Firefox 141.0.3/Brave 1.81.131 ### Confirmation - [x] I have read and followed all instructions in `README.md`. - [x] I am using the latest version of **both** Open WebUI and Ollama. - [x] I have included the browser console logs. - [x] I have included the Docker container logs. - [x] I have **provided every relevant configuration, setting, and environment variable used in my setup.** - [x] I have clearly **listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup** (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc). - [x] I have documented **step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation**. My steps: - Start with the initial platform/version/OS and dependencies used, - Specify exact install/launch/configure commands, - List URLs visited, user input (incl. example values/emails/passwords if needed), - Describe all options and toggles enabled or changed, - Include any files or environmental changes, - Identify the expected and actual result at each stage, - Ensure any reasonably skilled user can follow and hit the same issue. ### Expected Behavior LM Studio is expected to be usable as backend for Open WebUI, since it is listed in the documentation for Open WebUI. ### Actual Behavior Configuring LM Studio as OpenAI API compatible server in Open WebUI results in errors in LM Studio and no models listed/usable in Open WebUI. ### Steps to Reproduce Since Ollama chose to actively break their GGUF-support https://github.com/ollama/ollama/issues/11714 and haven't visibly acted to repair their own mess, I would now like to migrate to using LM Studio's integrated llama.cpp to run the LLMs served by Open WebUI instead of Ollama, but I can not get the endpoint configuration to work. What may be complicating the matter is that in my current (completely functional apart from Ollama's recent GGUF-breakage) setup I run Open WebUI on a single-board computer, an Organge Pi 5 Plus running DietPi to then serve Open WebUI through a composerized docker, while the LLM is run by Ollama on a dual-boot Nobara 42 (Fedora-based) and Windows 10 desktop machine because it has a GPU (RTX 4070 SUPER) which of course the single-board computer doesn't (can't). The reason I want to use Open WebUI, as I am, is because I also serve it through a cloudflared tunnel making it reachable from the Internet, so I can access my LLMs from anywhere, so just using LM Studio directly would not provide the desired functionality. The reason I don't want to run llama.cpp myself direcrly but instead rely on LM Studio's is that they take care of a lot of chores automatically like managing llama.cpp, hardware compatibility layers, etc and I don't have docker installed on the desktop. With Ollama on that desktop machine, let's call it "desktop", I put (pseudonymized) http://desktop.network.internal:11434 in the "Settings" → "Connections" → "Manage Ollama API Connections" → "Edit Connection" → "URL" field and that has worked for many months, but when I turn on LM Studio's server mode (with "Serve on Local Network" and "Enable CORS" turned on for maximum compatibility) and input the equivalent http://desktop.network.internal:51234 with changed port matching what LM Studio says "2025-08-13 15:58:39 [ERROR] Unexpected endpoint or method. (GET /api/version). Returning 200 anyway" when clicking "Verify Connection" in LM Studio and when I then save anyway to go check in "New Chat" on Open WebUI there are no models listed, even though there are several models (working fine) in LM Studio directly, but I keep getting new errors in LM Studio when I click on stuff in Open WebUI, like these: ``` 2025-08-13 15:59:55 [ERROR] Unexpected endpoint or method. (GET /api/tags). Returning 200 anyway 2025-08-13 15:59:55 [ERROR] Unexpected endpoint or method. (GET /api/ps). Returning 200 anyway 2025-08-13 16:00:48 [ERROR] Unexpected endpoint or method. (GET /api/version). Returning 200 anyway ``` I have seen on both Open WebUI's website (e.g. https://docs.openwebui.com/getting-started/quick-start/starting-with-openai-compatible#step-1-choose-an-openai-compatible-server) and at LM Studio's website (e.g. https://lmstudio.ai/docs/app/api/endpoints/rest) that they should work together, but I can't get it to work together. In desperation I have also tried adding variants of "/v1/api", "/v1", "/v0", "/api" and more on the endpoint URL specified in Open WebUI's settings, but if that is the issue I must have not stumbled on the correct one. So, I don't know what is wrong, but either the documentation or the Open WebUI handling of the configured value seems most likely to me to need modification to reflect current reality (if it's the docs) or handle LM Studio's current implementation (if it's the software itself). ### Logs & Screenshots Docker on "verify" in Open WebUI connection ``` openwebui-1 | 2025-08-13 14:41:48.439 | INFO | uvicorn.protocols.http.httptools_impl:send:476 - [Public Internet IP redacted]:0 - "POST /ollama/verify HTTP/1.1" 200``` LM Studio: ```2025-08-13 15:59:55 [ERROR] Unexpected endpoint or method. (GET /api/tags). Returning 200 anyway 2025-08-13 15:59:55 [ERROR] Unexpected endpoint or method. (GET /api/ps). Returning 200 anyway 2025-08-13 16:00:48 [ERROR] Unexpected endpoint or method. (GET /api/version). Returning 200 anyway ``` Browser Console during a "verify" in Open WebUI (seems extremely irrelevant, but mandatory to include in this GitHib issue template, so it is): ``` NotFoundError: No such JSProcessActor 'BrowserToolboxDevToolsProcess' ``` ### Additional Information _No response_
GiteaMirror added the bug label 2026-04-19 23:51:56 -05:00
Author
Owner

@tjbck commented on GitHub (Aug 13, 2025):

LMStudio is NOT Ollama

<!-- gh-comment-id:3184309265 --> @tjbck commented on GitHub (Aug 13, 2025): LMStudio is NOT Ollama
Author
Owner

@b9AcE commented on GitHub (Aug 13, 2025):

LMStudio is NOT Ollama

Yes..? That is also NOT claimed in any way. Completely irrelevant response.
Please read the actual text instead of guessing what it could have said, then reopen the issue.
Thank you.

<!-- gh-comment-id:3184516326 --> @b9AcE commented on GitHub (Aug 13, 2025): > LMStudio is NOT Ollama Yes..? That is also NOT claimed in any way. Completely irrelevant response. Please read the actual text instead of guessing what it could have said, then reopen the issue. Thank you.
Author
Owner

@rgaricano commented on GitHub (Aug 13, 2025):

For conect directly to lmstudio (as in documentation https://lmstudio.ai/docs/app/api/endpoints/openai ) you have to add it as openai api compatible connection, e.g. http://localhost:1234/v1)

<!-- gh-comment-id:3184700108 --> @rgaricano commented on GitHub (Aug 13, 2025): For conect directly to lmstudio (as in documentation https://lmstudio.ai/docs/app/api/endpoints/openai ) you have to add it as openai api compatible connection, e.g. `http://localhost:1234/v1`)
Author
Owner

@b9AcE commented on GitHub (Aug 13, 2025):

Yes, @rgaricano, that is indeed what I am trying to do.
When I add the "/v1" as in your response, clicking "Verify Connection" in Open WebUI results in "2025-08-13 19:02:14 [ERROR]
Unexpected endpoint or method. (GET /v1/api/version). Returning 200 anyway" in LM Studio's "Developer Logs" and no models listed in Open WebUI.

Using the "localhost" hostname (or 127.0.0.1) in the "URL" field does not work in this setup, as Open WebUI just responds "Ollama: Network Problem" plausibly because the Open WebUI is on a different host from the LLM host.
That is also true when reverting back to using Ollama, which still works, except only for their GGUF issue but everything else still works,
but when either a LAN IP-address based URL or actual hostname (e.g. http://desktop.network.internal:51234) is used instead, Open WebUI says "Server connection verified" instead but with error message in LM Studio's log (in the case with the added "/v1" it is "2025-08-13 19:13:50 [ERROR]
Unexpected endpoint or method. (GET /v1/api/version). Returning 200 anyway").

So, they do communicate, there is no "Network Problem" when using non-loopback address, but there seems to be some issue with the OpenAI API compatibility.

<!-- gh-comment-id:3184812892 --> @b9AcE commented on GitHub (Aug 13, 2025): Yes, @rgaricano, that is indeed what I am trying to do. When I add the "/v1" as in your response, clicking "Verify Connection" in Open WebUI results in "2025-08-13 19:02:14 [ERROR] Unexpected endpoint or method. (GET /v1/api/version). Returning 200 anyway" in LM Studio's "Developer Logs" and no models listed in Open WebUI. Using the "localhost" hostname (or 127.0.0.1) in the "URL" field does not work in this setup, as Open WebUI just responds "Ollama: Network Problem" plausibly because the Open WebUI is on a different host from the LLM host. That is also true when reverting back to using Ollama, which still works, except only for their GGUF issue but everything else still works, but when either a LAN IP-address based URL or actual hostname (e.g. http://desktop.network.internal:51234) is used instead, Open WebUI says "Server connection verified" instead but with error message in LM Studio's log (in the case with the added "/v1" it is "2025-08-13 19:13:50 [ERROR] Unexpected endpoint or method. (GET /v1/api/version). Returning 200 anyway"). So, they do communicate, there is no "Network Problem" when using non-loopback address, but there seems to be some issue with the OpenAI API compatibility.
Author
Owner

@rgaricano commented on GitHub (Aug 13, 2025):

You have to add it as openAI conection (not ollama)
and in admin settings (in direct connections ip have to be public accesible)

If you can verify the connection and you have errors in lmstudio, that are because it api is not a usable openai api compatible, refer to lmstudio repo and ask there for a working openai api compatible endpoints.

(or make a pipe for the non compatible lmstudio api)

<!-- gh-comment-id:3184900596 --> @rgaricano commented on GitHub (Aug 13, 2025): You have to add it as openAI conection (not ollama) and in admin settings (in direct connections ip have to be public accesible) If you can verify the connection and you have errors in lmstudio, that are because it api is not a usable openai api compatible, refer to lmstudio repo and ask there for a working openai api compatible endpoints. (or make a pipe for the non compatible lmstudio api)
Author
Owner

@b9AcE commented on GitHub (Aug 13, 2025):

Ah. Got it working now.
I believe what confused me was that on the "OpenAI-Compatible Servers" page https://docs.openwebui.com/getting-started/quick-start/starting-with-openai-compatible/#step-1-choose-an-openai-compatible-server Ollama is listed among llama.cpp, LM Studio and Lemonade, and Ollama is used as the example in the following instructions, so maybe that documentation page would be made more clearly understandable by not listing Ollama there and not using it as the example, but instead use llama.cpp as example in those instructions... is my proposal anyway.
Ollama already has its own separate page (https://docs.openwebui.com/getting-started/quick-start/starting-with-ollama) for instructions, so maybe including Ollama in the "OpenAI-Compatible Servers" documentation too seems superfluous and potentially confounding.

What will have subconsciously happened is that I noted Ollama and those others listed as if functional equivalents on that first page and without sufficient evaluation just concluded that since Ollama is on that page and I have Ollama configured working, all I need to do is to change from the Ollama TCP port to the LM Studio TCP port, as they are listed side by side on that page.
Furthermore, I'll personally blame that I am literally "legally blind". ;-P
Thank you, @rgaricano!

<!-- gh-comment-id:3185003377 --> @b9AcE commented on GitHub (Aug 13, 2025): Ah. Got it working now. I believe what confused me was that on the "OpenAI-Compatible Servers" page https://docs.openwebui.com/getting-started/quick-start/starting-with-openai-compatible/#step-1-choose-an-openai-compatible-server Ollama is listed among llama.cpp, LM Studio and Lemonade, and Ollama is used as the example in the following instructions, so maybe that documentation page would be made more clearly understandable by not listing Ollama there and not using it as the example, but instead use llama.cpp as example in those instructions... is my proposal anyway. Ollama already has its own separate page (https://docs.openwebui.com/getting-started/quick-start/starting-with-ollama) for instructions, so maybe including Ollama in the "OpenAI-Compatible Servers" documentation too seems superfluous and potentially confounding. What will have subconsciously happened is that I noted Ollama and those others listed as if functional equivalents on that first page and without sufficient evaluation just concluded that since Ollama is on that page and I have Ollama configured working, all I need to do is to change from the Ollama TCP port to the LM Studio TCP port, as they are listed side by side on that page. Furthermore, I'll personally blame that I am literally "legally blind". ;-P Thank you, @rgaricano!
Author
Owner

@rgaricano commented on GitHub (Aug 13, 2025):

Just for reference,
Ollama have it own API endpoints and also have openAI API compatible endpoints (you can connect ollama as openAI API compatible with e.g. http://localOllama:11434/v1 )

<!-- gh-comment-id:3185058081 --> @rgaricano commented on GitHub (Aug 13, 2025): Just for reference, Ollama have it own API endpoints and also have openAI API compatible endpoints (you can connect ollama as openAI API compatible with e.g. `http://localOllama:11434/v1` )
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/open-webui#17965