I'd like to be able to bring my own key for the app gen functionality so I have introduced DeepSeek, which has an open API-style endpoint.
Issue:
Even though I can select this new endpoint in the AI section of settings, the app gen doesn't work. Even though I still have a Tropic key re-tool managed, it won't default to this one when the other one doesn't function
Can I make it such that the DeepSeek open AI style endpoint works with DeepSeek reasoning?
If not, can I make Retool default to the anthropic?
@ChiEn could you have a look? I think it's very hard to find a way to control the actual model that is used by App Assist
I consulted documentation and went to a resource to configure this in Retool AI
It is nowhere to be found anymore
Then upon checking settings, AI there is no way I can specify the model. I can set an OpenAI key and an Anthropic one but I have no way to control which one is used
And also I provided an endpoint in the OpenAI style with DeepSeek. Also it didn't work
Thanks a lot. I'm looking for a way to control the model that my app gen uses, not the agent.
Also I have included two screenshots where you are able to see how I configured this. Also in the AI section it is selectable if I wanted to select the OpenAI DeepSeek because they offer an OpenAI-style endpoint to hit
Hey @mascaritas - unless you're on an Enterprise plan, app generation via Assist currently requires that you use Retool-managed keys, as described here. The fact that you even have access to this dropdown is probably a bug - I'll follow up internally!
Thanks a lot for your reply and explanation. Now what I have done is reset the API keys in the resources to be retool managed for Anthropic and for OpenAI.
Now I opened a new app and use App Assist to create new elements but constantly this dialogue would show up. I would assume if I have Retool managed for OpenAI and Anthropic, the models should be all right
Also I can then begin to build but it's painfully slow and it lets me wait so much longer than it used to