Thanks a lot for all the value! I have a use case for Bubble x OpenAI involving File Search, Vector Store, and all that pretty gnarly stuff to make a chatbot using a massive knowledge base. Do you have a video on the topic, or could you make one sometime? :)
When your title is about "temporary lists using custom state in Bubble" then why do we have to watch you set up icons and option sets for a couple of minutes first? This type of quick tutorial should go straight to the point, so it helps those that look for a quick answer to their question.
Not checked it out yet, but if if Comboco offers just a way to stream responses from API connector in Bubble, this would be a game changer.Many would agree we would finally not have to rely on 3rd party products for streaming.
Unfortunately we think this is unlikely as connecting Comnoco to Bubble still requires using the API Connector or Backend Workflows (to receive webhooks). Both these are limited by not supporting streaming.
I feel this intro and background would have been better to include at least a cursory overview / quick look at the product. I'm now none the wiser and it's a bit of a missed opportunity to be honest
Comnoco (comnoco.com/) is a visual an advanced function/workflow builder providing features and tools far bound the Bubble workflow builder. Effectively you can build your own APIs to link with your Bubble app.
We teach no code because we are a team of no coders. Simply put we can't code, but no code allows us to create 90% of the app ideas we think of. There will always be a place for traditional coding, but no code the most accessible tool for some people.
It entirely depends on what you're trying to achieve. We found that changing a URL parameter offers a native app-like experience. Changing pages is quick. However, we'd suggest you bear in mind that Bubble will be releasing a native app-building at some point later this year. This will mean that wrappers are no longer necessary. It is also our advice that you approach wrappers with caution. Many Bubble builders have been successful with wrappers. However, it is always worth checking the terms and conditions of each app store to see whether you get into trouble with using a wrapper.
I have a multidropdown, how to i filter a repeating group based on multiple values from the dropdown. "Tags (This is a list field type under products data type) contains multidropdown's value". This is not working.
@no_code hace 0 segundos Thanks for the amazing tutorial. I have a question tho. How scalable is this? If I wanted to scrape 10,20,30 sites, clean all that date, calculate some trends and present the info in charts or RG. Would I need xano or bubble is enough? Would Claude support all of that?
Depends on the sort of features you need. There will likely be plenty of WordPress plugins that will get a website set up in just a matter of minutes. But if you have no pre coding knowledge, you may find this hard to customize. Building it in Bubble in would take longer, but you get to choose and create your own features as you go.
Would love to see you do a deep dive on the Loops side of things and how you integrate Bubble actions with Loops features (doing things like an onboarding email campaign with things like splitting audiences, lists for custom unsubscribes, etc.)
Claude doesn't have threads, but you can submit the conversation history using the main endpoint to learn how to structure this content. Here our video explaining the process with OpenAI, which will be very similar (start around the 13:20 mark): www.planetnocode.com/tutorial/build-your-own-no-code-custom-gpt
Have you figured out how to get the data into these charts using UX connected APIs through accounting software such as quickbooks and go high level for sales and conversion data? How can I get in touch with you?
Yes, you can pull in external data as long as you achieve the right structure required by the chart plugin. This is something we can definitely with. Please reach out to support[AT]planetnocode.com or book a 1 to 1 call direct with Matt www.planetnocode.com/bubble-coaching/
This tutorial might have just salvaged the little belief left in me. Fixing responsiveness on mobile was tough, but following this tutorial helped a bunch. Thank you so much.
Thanks for the recommendation. We've been meaning to check out Buildship. Have you heard of Comnoco (comnoco.com/)? Seems like a little overlap with Buildship. We met the developer behind Comnoco at the No Code UK conference in May. Very impressive tool.
I hadn't heard about conmoco but curious to know more about it@@planetnocode. I'm still trying to figure out what's the best ai-orchestration tool tbh. There are many pontetial options such as stack ai, buildship, relevance ai, flowise, gumloop, among others.
first, thanks for this tutorial video. I was building assistant with open ai Assistants and as you know, its able to search files and save instructions but ngl its too expensive..im curious about something.. is it possible to upload files or make the bot able to search files? like for knowledge bases. so in summary is it possible to build some bot like gpt assistants with claude?
Great question. There are three ways we'd approaching this. 1. The OpenAI Assistant allows you to provide a small number of PDFs as part of the knowledge base. 2. You can use a third-party service like Voiceflow to manage the knowledge base. 3. The easiest option, without incurring much additional cost, would be to use a plugin to extract text from PDFs you upload to your Bubble app. Then, you can put all of that text in the context window of your first message or prompt that you send off to the AI. However, with this method, be aware that token usage can skyrocket.
Hi there! You could extract the CLOSING_DATE by applying "EXTRACT WITH REGEX" to the text using this: "closing_date":\s*"([^"]+) Do it within an ARBITRATY TEXT, so then you can apply (to that result) a FIND AND REPLACE of the following text with nothing (just leave the other empty): "closing_date": " Thanks for sharing by the way!! 🫶🏻
Thanks for the really helpful suggestion. Yes, you could use REGEX to extract the closing date here. However, what may not have been clear in the video is how it may be possible with the right prompt to extract a closing date from multiple job adverts on multiple websites. The AI needs to be able to interpret the position, the placement, the label of the closing date on the web page so that you can have less uniform structure and yet still achieve a reliable extraction of data.
@@planetnocode Glad you found the suggestion helpful, thanks again for sharing your knowledge! I interpret that the AI will be "just as capable as a human" to find the dates within the metadata of the web page, in the sense that if you had to run a javascript to display the data, and when scraping it was not done, the value will surely not appear, or something similar. Then, on the structured response of the assistant, we would apply the regex to extract the values of the keys we need (or we can iterate all of them using the same regex). That would be my idea about what I understand you are trying to do in this video, but I may have misunderstood something haha, cheers!
The thing with no code tools, is they are a decent learning curve that still takes a good chunk of time. I wonder if its just worth learning normal JavaScript in that time, at least then our future proofing and the skillset is way more diverse
We know people who have time and time again tried to learn traditional coding but it is a constant struggle. No code gives them the chance to launch an MVP, when otherwise they have no other choice by hiring a developer.
I’ve been using Coda for managing projects with my clients, in the end I did a web app. But Coda isn’t really for that, so I’m looking for the next step to really build a web app with a solid platform, ideally connected to a true database. What would you recommend? It’s a project management tool that needs to connect via API to other internal tools we have. Good UI is fundamental, my clients are designers and I need to keep a good visual.
Bubble.io is perfect if you're wanting to start with a blank page and really customize how you display lists of data and every step that you take your clients through when they use the app. If you want something a bit quicker, which has more prebuilt components but less flexibility, Glide sounds appropriate.
It would be a strong assumption to think that recurring workflows are just a backend workflow scheduled to repeat itself. Suggesting that as long as the recurring workflow is not canceled, you could change the actions in that workflow and it would continue to run.
@@planetnocode thank you. and do you also have a movie how to solve the issue that the same elements get imported again? and wat if a post gets updated? at this moment if you push the button again you get 5 (additional but the same) posts in the database
You'll likely need a series of checks by using a consistent unique ID shared between your Bubble database and your imported data. You can filter out importing existing records. Checking for updates is certainly more tricky. It would take a lot of workload units to check every field and update a specific field if only that field has changed. A simpler method would be to overwrite existing entries. This means when you find the match with the unique ID, delete the existing record, and create a new one in its place.
Amazing! Is that much faster than Whisper? How would you protect the confidentiality of what s being recorded and sent to Grow? Would it be running this backend?
We've not conducted like-for-like tests, but overall we're very impressed with the speed of Groq both with text generation and using Whisper for speech-to-text. Security is the largest consideration. Your MP3 file needs to be public in order send it to Groq so that Groq can read the file. If this has security implications for your application, you may need to consider a different transcription service like Assembly AI, where you can upload the file for transcription rather than having it in a public directory.
Hi, returning as a json format seems to be different for Claude compared to chatgpt api. I could never get the json manipulator plugin to find the array
We've just recorded a video looking into this. That will be published today or tomorrow. Claude does not have a JSON mode like OpenAI. It relies on you instructing in your prompt and asking for a JSON mode, and asking for a JSON response. In our testing, this JSON response is marked as code within text, like a code block. As a result of this, it won't be treated as JSON. We'd suggest using a custom state or the debugger to check the exact output that you receive to ensure that it is clean JSON.
Totally great. Easy to follow and it worked first time. Thank you very much. There are other tutorials on RU-vid about doing the same thing, this is the one that worked for me.
The question is: How do you sort by a field that is not in the main table you are setting for the table. For example, I have one data type for say "Subscriptions", and one data type for "quotes". My table's type of content is "Subscriptions", but I want to sort by the amount in the "quotes" that I am also displaying in the table. How would you do that?
You're correct, you can only sort by text and number fields directly on the data type you're using as your table data. A workaround for this is to use a database trigger create a text field for any dynamic data that is attached to your main data type. This way, when your secondary data is updated, your database trigger updates the text field in your primary data.
@@planetnocode Thank you! I really hope they improve the table function. It has been released for something like a year now and I don't see many improvements...
How can we make a AI-chat application with Claude API and bubble? In the sense, it should include the context of the previous messages before generating the new response.
Yes, you can do this 100%. We may not have any cloud-specific videos, but the way that you generate a list of messages in JSON is the same with OpenAI as it is with Cloud. Here's a link to an OpenAI video of ours demonstrating this: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Oqs9xgR-MBM.htmlsi=72vvbOiNkYl286PA&t=1146
@@planetnocode yep but here's the thing- even though I followed these tutorials to a T, it doesn't work. Well- it does work for the first 2 inputs from the user, we get the contextual response from Claude, but then suddenly the order in which it received these messages changes. For eg:- role: user, content: 5 countries in Asia Role: assistant, content: Here are 5 countries in Asia (proceeds to list them with 1. Russia, 2. China, 3....) Role: user, content: 2nd option Role: assistant, content: Great! Russia is the largest country by size in... would you like to know more about it's culture, tourism, flora and fauna.... Role: assistant, content: The tourism in Russia has been... Role: user, content: tourism And then, since the order of the last two messages have been switched, it gives an HTTP 400 error since it expected the last message on the list to be that of an Assistant's, not user's. Any way you can guide me on solving this at the earliest? Thanks a bunch!
@@planetnocode Thanks for your response- but here's the thing: If I follow these tutorials, the order of the JSON messages gets switched in the conversation and raises an HTTP 400 error, even though the first part is able to contain the context. It goes like:- User: 5 countries in Asia Assistant: Here are 5 countries in Asia... 1. India, 2. Russia, 3. China... User: 2nd option Assistant: Great! Russia is the largest ... would you like to know more about it's population, tourism, flora? Assistant: The population of Russia.. User: population The above happens with it's JSON script format. Is there any way to tackle this issue?
Thanks for flagging this! Part 2 is exclusively for our members. Here's a direct link: www.planetnocode.com/tutorial/how-to-split-and-save-openai-response We'll also post the link in the description of this video.