I'm following this instructions:

I'm attempting to integrate Typebot with GPT-4V (Vision) through the Webhook Block.
It works seamlessly when the URL points to any image from Google Images, randomly selected. However, it encounters issues when the image is sent/uploaded through Typebot.

The error message provided is too ambiguous, I'm speculating that the problem may be related to the cross-origin policy.

I'm seeking any clues or advice that could help troubleshoot this issue.

For example, it can't handle array of arrays.
It ends up converting the inner arrays to a string.

How feasible is it to develop a custom block using Forge with intricate functionalities, such as multiple outputs or a tailored user input interface similar to those found in a bot engine?

For instance, in the absence of the "Rating" or "AB Test" blocks, could Forge be utilized to create comparable custom blocks?

Is there a method to initiate an action either 'upon user exit' or 'after a specified duration of inactivity,' such as 'x seconds since the last iteration'?

I'm trying to run it locally, following the steps from and using NEXT_PUBLIC_E2E_TEST=trueto bypass the auth.
The site seems to be stuck in a perpetual loading state, and I can't find any errors in the browser console or the terminal where I'm running the pnpm dev, as you can see in the images.

What might I be overlooking?

What may cause this?
It's not loading (otherwise we would see the "...")
I got no error.
Just a premature end.

Add a tool to OpenAI, and then remove it.
In this case, tools will be set to [] instead of undefined, which is not acceptable to OpenAI.

I have this minimal bot, that works fine in the test environment.
I can enable the Whatsapp integration, publish and still works fine.

In certain situations (I'm not sure about the specific conditions yet), when interacting with the bot through WhatsApp, the bot initiates a second flow in the middle of the first one, duplicating all messages. At first glance, it appears that this issue occurs only when there is a prolonged gap between user interactions.

Is there a known bug related to this?

I send an image to Typebot. In Typebot I get an URL like:
https://typebot.myurl.io/api/typebots/<bunchoffstuff>
If I access this url I get the error: {"message":"Not authenticated"}

But if I access my minio, I can see that the original image URL is something like:
https://s3.myurl.io/typebot/public/workspaces/<bunchofstuff>
And that's public accessible.

How to avoid this?
(I need the public available URL to send it to GPT Vision)

While using the Whatsapp (official) integration, looks like the first user messages only starts the flow, but it's not collected. Even if my very first block is a Text Input Block. Is it possible to collect the very first user message from Whatsapp?

Also:
If the user sends a text message I get a text message.
If the user sends an file I get a URL (while using the Text Input Block).
But, if the user sends an image with a text, the text is lost. Any workarround to avoid this situation?

I'm current selfholsting using docker.
But I made some code changes (created a forge block) so I guess docker is not an option anymore, right?

So I'm trying to deploy at Vercel (with a superbase database) following the instructions here:
https://docs.typebot.io/self-hosting/deploy/vercel

I'm getting a "PostgreSQL DATABASE_URL should start with postgres" (it does) warning followed by a "ELIFECYCLE  Command failed with exit code 1." error.

Any advice?

Can a Forge Custom Block create or a set variable?
How?

Within Whatsapp, users have the option to choose between sending a text message, an image, or both. TypeBot, on the other hand, provides separate text and image inputs, lacking a combined option. Is there any workaround to this in TypeBot?
(I'm using GPTVision)

If I do something like the image (URL: https://hub.dummyapis.com/delay?seconds=10)
it takes 20 seconds (10 of each request) of black screen, then display A, B, C.
I was expecting A, 10 seconds, B, 10 seconds, C.
Looks like it run all the sequence till it finds a user input to start displaying the results. Why?
Can I do anything to avoid this?

I'm using the code bellow every time I want to display a monetary variable:

Is there a better way to do this?