Adding configuration silently fails

[ejgroene]

I installed CWC in VScodium and tried to configure Perplexity for use with API.

Pressing the cog (settings) in CWC sidebar and the 'API Providers' and then '+ Add API provider' allows me to fill in:
Name: Perplexity
Base URL: https://api.perplexity.ai
API Key: ...

Then, switching to 'API call' in the sidebar.
Pressing 'Manage Configurations' brings me to 'Edit Context Configuration' and tells me "No configurations found, add one to use this API tool'. I click '+ Add configuration', it brings me to 'Configured API providers' and it lists my previously added API configuration.

I click on it and it jumps back to 'Edit Context Configuration' with nothing changed. I can repeat it forever. No luck.

[robertpiosik]

Hi, as I cannot reproduce, can you try any other provider and see if it works? You can add OpenRouter with invalid key (their /models endpoint is public).

[ejgroene]

Thank you for your quick reply. With OpenRouter, it seems to work, and a configuration appears under "MY CONFIGURATIONS", though I cannot select it, but that is perhaps because the key is not valid. But at least it adds it.

BTW: the logs are empty.

[robertpiosik]

Can you reach me on telegram @robertpiosik? We will sort it out

[ejgroene]

Sorry, only Signal or iMessage, or any ad-hoc solution if you have any preference.

[ejgroene]

Google Meet also works, via ieee.

[robertpiosik]

Sent you an email

[robertpiosik]

@ejgroene thank you for this bug report. The cause is that the provider doesn't expose /models endpoint (as should if conforming to the OpenAI-API spec) we can read to assist in model selection. A fallback mechanism has been implemented for manual model name input.

In case of perplexity, model ids can be found here: https://docs.perplexity.ai/getting-started/overview

Fix has landed in 1.209.0.

[ejgroene]

Hi Robert, Thank you very much for you quick reply. It is now possible to add configurations for both Perplexity and Ollama. However, pressing enter always does 'select config'. The same happens when I click on either 'Select config' or 'Send request' button. Also, although 'Manage Configurations' allows me to select a default (which it remembers), this is not reflected under "MY CONFIGURATIONS" when selecting API call. Hence it is not possible to actually chat. Best regards, Erik

…

On Tuesday, 29 July 2025 at 14:09, Robert Piosik ***@***.***> wrote: ***@***.***(https://github.com/ejgroene) thank you for this bug report. The cause is that the provider doesn't expose /models endpoint (as should if conforming to OpenAI-API spec) we can read to assist in model selection. A fallback mechanism has been implemented for manual model name input. In case of perplexity, model ids can be found here: https://docs.perplexity.ai/getting-started/overview — Reply to this email directly, [view it on GitHub](#381 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/AANOORY3ICF4J5ZW75WN4ZD3K5P7TAVCNFSM6AAAAACCQOJP3SVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTGOJSGE4DSOI). You are receiving this because you were mentioned.Message ID: ***@***.***>

[robertpiosik]

However, pressing enter always does 'select config'. The same happens when I click on either 'Select config' or 'Send request' button.

I know it's not currently documented and causes confusion... CWC has 2 modes of operation, it can initialize chats (New chat switch position) or make API calls. In both modes, messages are identical. It is currently not possible to use API provider (and Edit Context tool) for chats, as we're missing our own web chat implementation but this is planned.

The solution is to host your own instance of Open WebUI, set up API provider there, and set model in a new chat preset. Open WebUI is fully supported (model, system instructions, temperature, etc.)

[robertpiosik]

However, pressing enter always does 'select config'. The same happens when I click on either 'Select config' or 'Send request' button.

I'll add default config selection in this picker that is shown when no default is chosen.

[robertpiosik]

However, pressing enter always does 'select config'. The same happens when I click on either 'Select config' or 'Send request' button.

I'll add default config selection in this picker that is shown when no default is chosen.

This is now implemented in 1.211.0.

[ejgroene]

In addition, it now gives error messages in the lower right corner: API request failed. Check console for details. Where would that console be?

…

On Tuesday, 29 July 2025 at 15:05, Erik J. Groeneveld - LPV ***@***.***> wrote: Hi Robert, Thank you very much for you quick reply. It is now possible to add configurations for both Perplexity and Ollama. However, pressing enter always does 'select config'. The same happens when I click on either 'Select config' or 'Send request' button. Also, although 'Manage Configurations' allows me to select a default (which it remembers), this is not reflected under "MY CONFIGURATIONS" when selecting API call. Hence it is not possible to actually chat. Best regards, Erik On Tuesday, 29 July 2025 at 14:09, Robert Piosik ***@***.***> wrote: > ***@***.***(https://github.com/ejgroene) thank you for this bug report. The cause is that the provider doesn't expose /models endpoint (as should if conforming to OpenAI-API spec) we can read to assist in model selection. A fallback mechanism has been implemented for manual model name input. > > In case of perplexity, model ids can be found here: https://docs.perplexity.ai/getting-started/overview > > — > Reply to this email directly, [view it on GitHub](#381 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/AANOORY3ICF4J5ZW75WN4ZD3K5P7TAVCNFSM6AAAAACCQOJP3SVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTGOJSGE4DSOI). > You are receiving this because you were mentioned.Message ID: ***@***.***>

[robertpiosik]

On what action?

[ejgroene]

Ah, that kind of console... thanks!

…

On Tuesday, 29 July 2025 at 16:03, Robert Piosik ***@***.***> wrote: On what action? [image.png (view on web)](https://github.com/user-attachments/assets/fdf0472a-fce3-468d-972b-2a2ea61feae1) — Reply to this email directly, [view it on GitHub](#381 (reply in thread)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/AANOORYOBCQXN5OXMHYRG7T3K55LHAVCNFSM6AAAAACCQOJP3SVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTGOJSGMYTANI). You are receiving this because you were mentioned.Message ID: ***@***.***>