Add option to add open router api

[sudip358]

Please consider the inclusion of an option to implement an open router API.

[marginal23326]

In the WebUI's LLM Configuration tab, you can put the Base URL as https://openrouter.ai/api/v1 with your OpenRouter API key. Make sure to choose the provider as openai from the LLM Provider selector (which should be the default).
Then you can type any model OpenRouter hosts in the Model Name section.

This should hopefully do the job for now.

[Sandv]

But how to choose a model then?

…

On Thu, Feb 13, 2025, 9:06 AM marginal23326 ***@***.***> wrote: In the WebUI, you can put the base url/endpoint as https://openrouter.ai/api/v1 with your OpenRouter API. Make sure to choose the provider as openai from the selector. — Reply to this email directly, view it on GitHub <#289 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADYOUQM77CGVBE2BOIYOUMD2PTGIJAVCNFSM6AAAAABXCT4VBCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNJXGIZTGOBXGA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> [image: marginal23326]*marginal23326* left a comment (browser-use/web-ui#289) <#289 (comment)> In the WebUI, you can put the base url/endpoint as https://openrouter.ai/api/v1 with your OpenRouter API. Make sure to choose the provider as openai from the selector. — Reply to this email directly, view it on GitHub <#289 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADYOUQM77CGVBE2BOIYOUMD2PTGIJAVCNFSM6AAAAABXCT4VBCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNJXGIZTGOBXGA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[marginal23326]

But how to choose a model then?

You can type the name of the model in the Model Name input field inside the LLM Configuration tab. It has to match whatever OpenRouter uses.

I haven't tested it, but it should work.

[xxxJustSomeBoringLadxxx]

I second this request for openrouter.ai compatibility. I've put the model in after pointing to openrouter url, but sometimes some of it works, most of it doesn't. Deepseek R1 models is erroring. V3 sometimes works. Sometimes nothing. Llama models give errors. Qwen seemed to work usually but certainly isn't as smart as some of the others.

Marginal's suggestion works, but sometimes it seems and only with some models. I tried selecting deepseek and using the openrouter api and it still didn't work with R1.

I just started messing with this today. Is there any way to log the output and input to and from the LLM to see exactly what we're sending and what we're getting back? I'm happy to test some stuff if I can help.

[marginal23326]

Deepseek R1 models is erroring.

Make sure that the "Use Vision" checkbox is turned off when using the DeepSeek model, and others as well, that do not support image data. Refer to this answer:

#243 (comment)

Is there any way to log the output and input to and from the LLM to see exactly what we're sending and what we're getting back?

I think the logs you are looking for are already stored under \tmp\agent_history in JSON format.

Also, it would be helpful if you could share the errors you encountered in the terminal while trying to use some of the models.