Dialog not working
Is it possible that dialog is no longer working? I created a very simple flow following https://www.loom.com/share/df5d64dd01ca47daa5b7acd18b05a725?sid=8dab03d1-f4c2-45eb-8dd6-aaf5b08ef16b and it doesn't seem to be working. I am however churning through a ton of tokens.
One thing I noticed is that the logs are not showing up in OpenAI.
This is what OpenAI says about that.... I checked Typebot’s documentation and changelog, and here’s what I found: • The OpenAI block in Typebot uses the standard chat completion endpoint (/v1/chat/completions) and doesn’t support the newer Responses API endpoint (/v1/responses)  . • The Typebot changelog notes a fix related to custom base URLs for model fetching, but it doesn’t mention support for the Responses API . ⸻ Can you simply change the Base URL? Unlikely. While Typebot lets you specify a custom base URL, it generally expects to build requests using the old API path. Changing the base URL to /v1/responses is unlikely to work unless Typebot also adapts the request payload format — which it currently doesn’t. ⸻ What are your options? 1. Wait for Typebot to officially support the Responses API — keep an eye on their changelog or documentation updates. 2. Use a proxy server — you can route your Typebot requests through a small intermediary service that: • Receives requests from Typebot to /chat/completions • Forwards them to OpenAI’s /v1/responses endpoint • Returns the response back to Typebot This setup lets you populate the Logs tab via the Responses API without waiting for Typebot to natively support it. 3. Track requests yourself — log request and response data at your application level if that meets your needs.
This is what OpenAI says about that.... I checked Typebot’s documentation and changelog, and here’s what I found: • The OpenAI block in Typebot uses the standard chat completion endpoint (/v1/chat/completions) and doesn’t support the newer Responses API endpoint (/v1/responses)  . • The Typebot changelog notes a fix related to custom base URLs for model fetching, but it doesn’t mention support for the Responses API . ⸻ Can you simply change the Base URL? Unlikely. While Typebot lets you specify a custom base URL, it generally expects to build requests using the old API path. Changing the base URL to /v1/responses is unlikely to work unless Typebot also adapts the request payload format — which it currently doesn’t. ⸻ What are your options? 1. Wait for Typebot to officially support the Responses API — keep an eye on their changelog or documentation updates. 2. Use a proxy server — you can route your Typebot requests through a small intermediary service that: • Receives requests from Typebot to /chat/completions • Forwards them to OpenAI’s /v1/responses endpoint • Returns the response back to Typebot This setup lets you populate the Logs tab via the Responses API without waiting for Typebot to natively support it. 3. Track requests yourself — log request and response data at your application level if that meets your needs.
Loom
Understanding the New Dialogue Option for the OpenAI integration
In this video, I explain the deprecation of the message sequence option in the Open AI block and introduce the new dialogue option in Typebot. I discuss how the dialogue option is more flexible and less magical than the previous option. I provide a step-by-step explanation of how to use the new dialogue option, including creating a chat history ...
9 Replies
Mmmh should work though
Can you record a video that shows your entire config and a test?
Baptiste, Please take a quick look at this and tell me what I might be doing wrong
https://www.loom.com/share/ce1d9f38ddb4484384989fe8c8732015?sid=63f86b17-cf16-442b-a255-001cc8e61308
Loom
Debugging Chat History and AI Responses 🤖
In this video, I discuss an issue I'm encountering with trying to set up a dialog with AI. Neither Anthropic or OpenAI seem to be working. I am using api keys for both that I successfully use in other applications.
Right, I see the issue! You forgot to select the item to save in your variable
Sorry that the user interface feedback is not more prominent on this
I was pretty sure this was newbie user error. Thanks.
I ran some interesting tests this morning using Dialog with and without append to see how well gpt could retain info within a conversation. Without append it sucked, which frankly kind of surprised me.
I am not clear on when to use Dialog vrs setting a system prompt and dropping my {{chat_history}} into the user prompt. Seems to me that it works the same. Guidance?
Is there a special dialog varible type in Typebot that automatically turns interactions into a JSON format that looks like this:
[
{"role":"system","content":"…"},
{"role":"user","content":"…"},
{"role":"assistant","content":"…"}
]
And if so, I am missing how it get's created. It seems my chat_history is just a simple text field at this point.
As you can see I attempted to replicate this JSON format, however the Append Values always treats the content as text which means that there is an extra set of quotes. It would be ideal if there were a way to generate this as proper JSON. Not critical, but it sure would be nice (api would be happier!).
You can combine system message and dialogue afterwards:
