Become part of our community and unlock immense value in AI Automations! Click here: shorturl.at/lrwyX 🚀 Just tried integrating the custom GPT on your site? Drop a comment and let me know how it went! Any bugs or was it all smooth sailing? 🛠🚀
Nicely done! I haven't tried integrating with the website yet, but all steps leading up to it have worked flawlessly. It's interesting that the code is creating the assistant rather than us creating/managing the assistant on the openai site. I guess it could be done either way, and the current method is a 'one-stop-shop'. Next steps on my end will be to host the python code within my own infrastructure (replacing replit). The python code looks simple enough to host and run, it's basically the template that openAI recommends. Very much appreciate you posting this. I'm looking to integrate a comprehensive chatbot for support as well as integration with actions that point to existing APIs on our back end (allowing users to interact with some of the functions we offer via the bot instead of using the UI on the site). Your tutorial was absolutely perfect in terms of pacing and execution. Thank you! Next step is to play with the Functions section of Assistants. I had actions working in custom GPT, it'll be interesting to see how Functions differ from that.
This is great bro. I added this to my 'how to - Ai & Bots for ultra beginners' playlist. There are very few people out there who show the newest and best ways to easily create ai and chatbots that anyone can do catering to people without technical skills. You are defjnately a legend in my eyes, subscribed. Credit to you and the other legends who provide the code for us. Looking forward to more videos from you
🎯 Key Takeaways for quick navigation: 00:00 📹 *This video provides a guide on integrating a custom GPT into a website, building on a previous tutorial about creating a custom GPT.* 00:13 🌐 *Two methods for integrating a custom GPT into a website are covered: a no-code method and a more advanced method.* 00:40 🤖 *Explains the difference between GPTs and Assistant API, with GPTs being B2C (business-to-consumer) and Assistant API being B2B (business-to-business).* 01:23 🛠️ *The tutorial uses a code-based approach to create an assistant, connect it to a chatbot builder, and integrate it onto a website.* 01:35 💻 *Demonstrates using Chatbase as an example of a platform offering easy integration of custom GPTs into websites.* 02:05 📚 *Shows how to add a knowledge base to a chatbot and connect it to a GPT model.* 02:44 🔄 *Discusses customization options for the chatbot, including selecting GPT models and adding custom instructions.* 03:38 🔗 *Details the process of embedding the chatbot into a website using a script.* 04:09 ⚖️ *Compares the advantages and disadvantages of using third-party services for GPT integration.* 04:52 🏗️ *Introduces the second method involving creating an Assistant API on OpenAI and connecting it to a chatbot for website integration.* 05:31 🧩 *Explains the process of using Replit and Voice Flow for the backend and frontend of the chatbot, respectively.* 06:25 📝 *Guides on customizing the chatbot with specific instructions and knowledge base.* 07:33 💡 *Offers insights into how the backend code works and interacts with OpenAI's platform.* 08:15 🌍 *Demonstrates setting up the frontend with Voice Flow and linking it to the backend.* 09:26 🖥️ *Shows how to embed the chatbot into a website and customize its appearance.* 10:34 🤔 *Reflects on the evolving role of AI in business, particularly in customer support, and encourages leveraging AI for efficiency.* Made with HARPA AI
i gotta leave a comment. this is too well done. straightforward , simple and on point, explain everything necessary for outsider, pre-set files are the best. this is just too good. thank you so much. subscribed and liked
Hi, Great video! Why should we use replit here? Can we just use Vflow only? I think vflow also link to GPTs and we can create the bot directly there. Sorry, I am confused about the role of replit here. Thanks!
Thank you for sharing. If I make changes to replit in the knowledge.docx, how do i update in my openai. I also see an option in openai assistance to upload my file documents. How does replit and openai communicate with these updates?
hey man ! i didn't get the part where dragged and dropped knowledge base , did you make you make your own knowledge bas , if so , how do make one ourselves
Hey man. For the knowledge base it's pretty easy. Just create a word document. And paste all the info you want your assistant to know. It's that easy. Then put the file in the thing
Thanks. I guess I will try this. 2 questions: 1. so actually if you need to modify instructions to the gpt you have to do through the code? Where it says instructions? 2. Can I use google tag manager to install this code on my website?
this is great. now I just need to figure out how to get user input saved as a form. easy million dollars right there because the possibilities are endless from there.
Thanks! I had to do much customization on your Replit template, but now it's up and running! how can we implement it to take control of our Whatsapp conversation as well? I'd like to let it also take control over our Whatsapp during closing hours
also can you show us how you structured your knowledge base?. Should it be in Q&A pairs or paragraphs? Does the structure of the knowledge base and the file type matter in how fast and efficiently the chatbot can retrieve the right answers? If yes, please advise on what is the best way to structure the knowledge base..thank u
Super nice tuto ! Bravo mate ;) Quick Question : Can this chat bot initiate a discussion ? Is it possible to set up this parameter on voiceflow ? also is it possible to remove the voiceflow watermark ? Thank you for your asnwers.
I am having trouble connecting the replit and voice flow. When copying the URL to paste into voice flow, I am getting the wrong URL that does not work. Anyone knows what I am doing wrong?
Thank you for the guide and resources. The response generated by the bot is not streamed, can you please help me with the approach I can use to get a streamed response.
Thanks. The set up went well, but when I run the chat the searching in the knowledgebase is very slow. It takes about 20 seconds for the chatbot to show the answer. Why is that?
@@RomanH91 had to pay for replit core so I could use a more powerful server with more allocation. Luckilly this project is making me money even after paying for replit. If you want to keep it free, make your own home server (old laptop) and run it there
Hello, I used the api key in my free plan. I also changed the model to gpt-3.5 turbo. The get method works successfully, but when I respond, the assistant cannot produce a response and the run status falls into failed. but "api request failed" does not happen. When I try it from my upgraded openai account, gbt-4o still doesn't work, gbt-3.5 turbo works. Likewise, I get the run status failed error. can you help me?
Is it possible to do connect it to a GPT built on ChatGPT? so all the instructions and actions are there on the pre-built GPT and the integration only needs to talk to the GPT
Hi man, thanks for this useful tutorial. Will you be updating this code to support Assistant 2 API and other models (GPT4o-mini)? It seems to use Assistants v1 API which will be phased out end of 2024
Thanks so much for that video. Do you know what happens with the data entered by the user? For example if I ask for name and address, how is the privacy/gdpr handled?
Hi, i have encountered a issue, hoping you can help me understand where it went wrong. After following your steps, i works perfectly until I closed all the windows, the chatbot when it started it says "Failed to create new thread". Do i need to press save button or pay to keep it running?
Hi Bo, thanks very much for the video. One problem though - overnight when Mac went to sleep, when turned the Mac back on in the morning, the code was still running in replit, but it lost connection since the Mac was sleeping, meaning the chatbot was offline. Any solutions? Much Appreciated
@@Sharpthookay, but what is the link then? in tutorial you have to copy link after clicking webview, if you deploy repel there is no such option to check the link. Of course you have the deploy link, where your app is hosted, but it doesn't not work as the link from the tutorial. Any ideas how to do it?
@@Sharptho Thank you for your reply. Okay, I am currently using link in Voiceflow bot and it shows a status of 200. However, after sending a query to the bot, it only displays the welcome information and seems like it doesn't send the request to OpenAI. I tried link itself nad with /start and /chat also
Great video! Just implemented it without any issue. A big thanks! What are others thoughts on the cost of this way of implementing it? Other than the OpenAI balance, no hidden costs are applicable I assume?
@@bo-sar great video - thanks! To check: I take it that a) the calls to openAI originate from replit so that the openAI key only needs to be provided in replit (not voiceflow) and b) - regarding costs - that all token charges originate from openAI directly & the AI token limits in voiceflow plans are irrelevant given this setup. Is that right?
Firstly, thank you so much. does this simply return word by word from the knowledge base or does it use its inference power to make up answers using the knowledge base?.
Wow. Thanks a lot Bo. In a short period, I was able to make things happen. It was my first time using these tools and Python for that matter, but was able to run my Chat Bot successfully. I just need to learn and understand how the code works and debug the constant connection error. Is there any way one can host the code in Replit in my own shared server?
i have been developing ocr past month, the result i get is not 100% accurate so i want to integrate the 'Sider AI' their ocr into my wedsite, additional user input ill get as the output form. if possible can you make a video.
Thank you for the video, really simple and easy 👍🏼👍🏼 Just have a question please, if I want to edit the instructions and the knowledge file, should I edit it inside replit or can I edit them from the Assistants panel in OpenAI?
I have a pdf that is a scan of handwritten data. I'd like ChatGPT4 Vision to OCR and perform calculations from the data in my tables.. then edit or append the pdf with the results. If I understand correctly, there's 3 ways of doing it 1) Custom GPT using OpenAI Builder, can use plugins (ex. Ai PDF) and Custom Actions. If approved, App can be hosted and shared with other Plus users . Limited to the 40msgs/3hr 2) Using my own app via assistant API/Secret-Key. Can't use plugins and must be hosted elsewhere like Replit. Pay per use, so no message limit 3) 3rd Party tools like MindStudio/AirTable that use assistant APIs to create app. Can't use ChatGPT plugins, but can use other AI models (Gemini, Perplexity etc). Hosted on 3rd party sites. Any recommendations how you would approach the task task of OCR / data analysis?
Wow! Thank you so much for taking the time to put this together. Would it be possible to include TH-cam or Vimeo links in the uploaded documentation? If so, then, when the Chatbot answers if a particular video is relevant, it could give the answer as well as play the video? I'm thinking this might work well for a tutorial site where learning by watching videos is a big part of the process. If it works, one could reference "free videos" most of the time, and if the topic is too complex, it could reference a "full course" that they could subscribe to for more information. (hmmm, this might be a good angle for you!). Again, thank you for this video and thanks in advance for your response. (new sub here). 🙂
Great tutorial, thanks. I changed the functions py so I could use multiple .pdf files as the data source and it works great. Question: for a simple question the cost is about 5 cents. According to the tokenizer, the question was 13 tokens. The response was 255 tokens. But the playground token counter (top right of thread) showed IN: 3950 Out: 294? Why is that? Edit: So, I dug into this and apparently, the reason IN is 3950 tokens is because of 'context' tokens -- and I think the PDF doc(s) are being loaded into the question context? Now I'm really confused -- will the PDF source files be loaded as context for each question -- this will get expensive fast...
@@rodneyschaap I don't remember exactly as that code is long gone, but it was something like this: import json import os def create_assistant(client): assistant_file_path = 'assistant.json' if os.path.exists(assistant_file_path): with open(assistant_file_path, 'r') as file: assistant_data = json.load(file) assistant_id = assistant_data['assistant_id'] print("Loaded existing assistant ID.") else: file_paths = ["pdffile1.pdf", "pdffile2.pdf", "pdffile3.pdf"] # Add additional file paths here file_ids = [] for path in file_paths: file_obj = client.files.create(file=open(path, "rb"), purpose='assistants') file_ids.append(file_obj.id) assistant = client.beta.assistants.create( instructions="""This tool assists with information about ....""", model="gpt-4-1106-preview", tools=[{ "type": "retrieval" }], file_ids=file_ids) with open(assistant_file_path, 'w') as file: json.dump({'assistant_id': assistant.id}, file) print("Created a new assistant and saved the ID.") assistant_id = assistant.id return assistant_id
@bo-sar i'm completely new to replit and voiceflow so forgive me if this is easier then i may know lol. If i want the customer to upload images and ask questions about those images, how would i modify this method?
Everything works great when I use the "run" button. However, for the application to work continuously, it needs to be deployed. I deployed it on Replit, but then there is no way to check the link as you show at (8:33). I tried to use the production link that I have for the deployment, but then ChatGPT does not launch. So, what link should I place in Voiceflow if I deploy on Replit?
Hi, I've gone through all the steps and generally it works, but unfortunately it doesn't work the same as the custom-made GPT in OpenAI. I don't know what's wrong. The instructions and training I developed in my custom GPT worked perfectly. I copied them to the knowledge file where I have described how the chat should describe, what the maximum number of responses it should give, etc. Additionally, I added a database from which the chat is supposed to pull data about available products. Unfortunately, the end result is different than on the OpenAI platform. Is it the same for you?
Hi, sorry for the delay, I took a little break... How different is it? Is it wrong/worse than the one by the custom GPT, or is it just using different wording?
Awesome tutorial! The only thing is if I add new information in knowledge.docx, it is not reflected in the bot. Does replit save automatically the changes in this docx? If so, does the bot get updated with new info automatically? Thanks
Thanks man! Amazing content. But I have an important question, what’s the knowledge base limit? I have a projeto with a lot of text files (like thousands) can I put it all? And also, using this method can I use links as knowledge base like what chatbase provide? And did you have to pay voice flow to use in this tutorial and no having the “created by voiceflow” message? Thank you!
When I view the URL from web view in replit it shows up different, it look like a bunch of random letters/numbers and ends in .dev, not .co like yours. And doesn't work when I paste it into voiceflow. Am I missing a step in order to use the URL from replit?
Chatbase would be costly for sure, but getting Replit, Voiceflow, and openAI tokens would end up even more expensive. Although, it can give more flexibility.
How can we put a restriction on the chatbot so that it only replies with information about what we want? Because there are some jailbreaks happening with gpt-powered chatbots where users can ask completely unrelated questions and the gpt will still answer them
Hey, amazing video… I have a question though. I currently try to build a discord bot for using my gpt i built on discord and sharing it with only a few friends. I don‘t wanna share a link, cause i wanna be 100% sure, who’s using my gpt bot. Now I try to find videos on yt on how to connect my gpt with a discord bot. Could you do a video about it or at least give any advise? Id highly appreciate it!
Become part of our community and unlock immense value in AI Automations! Click here: shorturl.at/lrwyX 🚀
Just tried integrating the custom GPT on your site? Drop a comment and let me know how it went! Any bugs or was it all smooth sailing? 🛠🚀
Nicely done! I haven't tried integrating with the website yet, but all steps leading up to it have worked flawlessly. It's interesting that the code is creating the assistant rather than us creating/managing the assistant on the openai site. I guess it could be done either way, and the current method is a 'one-stop-shop'. Next steps on my end will be to host the python code within my own infrastructure (replacing replit). The python code looks simple enough to host and run, it's basically the template that openAI recommends.
Very much appreciate you posting this. I'm looking to integrate a comprehensive chatbot for support as well as integration with actions that point to existing APIs on our back end (allowing users to interact with some of the functions we offer via the bot instead of using the UI on the site). Your tutorial was absolutely perfect in terms of pacing and execution. Thank you!
Next step is to play with the Functions section of Assistants. I had actions working in custom GPT, it'll be interesting to see how Functions differ from that.
At replit...is API key visible to everyone?
Can I hire you to do this for my website?
@bo-sar
@@Bengee555 feel free to contact me via email: bogdan.saranchuk4@gmail.com
This is great bro. I added this to my 'how to - Ai & Bots for ultra beginners' playlist. There are very few people out there who show the newest and best ways to easily create ai and chatbots that anyone can do catering to people without technical skills. You are defjnately a legend in my eyes, subscribed. Credit to you and the other legends who provide the code for us. Looking forward to more videos from you
thanks! that's my favourite comment so far. Appreciate it very much!
Thank you so much for this video! I have integrated my own custon GPT into our website!
🎯 Key Takeaways for quick navigation:
00:00 📹 *This video provides a guide on integrating a custom GPT into a website, building on a previous tutorial about creating a custom GPT.*
00:13 🌐 *Two methods for integrating a custom GPT into a website are covered: a no-code method and a more advanced method.*
00:40 🤖 *Explains the difference between GPTs and Assistant API, with GPTs being B2C (business-to-consumer) and Assistant API being B2B (business-to-business).*
01:23 🛠️ *The tutorial uses a code-based approach to create an assistant, connect it to a chatbot builder, and integrate it onto a website.*
01:35 💻 *Demonstrates using Chatbase as an example of a platform offering easy integration of custom GPTs into websites.*
02:05 📚 *Shows how to add a knowledge base to a chatbot and connect it to a GPT model.*
02:44 🔄 *Discusses customization options for the chatbot, including selecting GPT models and adding custom instructions.*
03:38 🔗 *Details the process of embedding the chatbot into a website using a script.*
04:09 ⚖️ *Compares the advantages and disadvantages of using third-party services for GPT integration.*
04:52 🏗️ *Introduces the second method involving creating an Assistant API on OpenAI and connecting it to a chatbot for website integration.*
05:31 🧩 *Explains the process of using Replit and Voice Flow for the backend and frontend of the chatbot, respectively.*
06:25 📝 *Guides on customizing the chatbot with specific instructions and knowledge base.*
07:33 💡 *Offers insights into how the backend code works and interacts with OpenAI's platform.*
08:15 🌍 *Demonstrates setting up the frontend with Voice Flow and linking it to the backend.*
09:26 🖥️ *Shows how to embed the chatbot into a website and customize its appearance.*
10:34 🤔 *Reflects on the evolving role of AI in business, particularly in customer support, and encourages leveraging AI for efficiency.*
Made with HARPA AI
i gotta leave a comment. this is too well done. straightforward , simple and on point, explain everything necessary for outsider, pre-set files are the best. this is just too good. thank you so much. subscribed and liked
Glad you liked it!
Honestly amazing video man. Straightforward and no issues were ran into!!!!!!
thanks! happy to hear that!
Thanks!
Thank you.. You helped a lot. Besides, you are a real cutie, but this is completely off the record..
Thanks so much for providing this clear step by step guide. I will definitely try this out myself.
Please do!
finally the best video about the topic
bro did you get the code in the description
Hi, Great video! Why should we use replit here? Can we just use Vflow only? I think vflow also link to GPTs and we can create the bot directly there. Sorry, I am confused about the role of replit here.
Thanks!
Thank you for sharing. If I make changes to replit in the knowledge.docx, how do i update in my openai. I also see an option in openai assistance to upload my file documents. How does replit and openai communicate with these updates?
thank yo soooooooo much you have helped me a lot, was such a easy process to create achatbot and integrate it into my website.
Great to hear!
Is it free to integrate
And please let me know the pricing
And can we connect for 5min regarding this
hey man ! i didn't get the part where dragged and dropped knowledge base , did you make you make your own knowledge bas , if so , how do make one ourselves
Hey man. For the knowledge base it's pretty easy. Just create a word document. And paste all the info you want your assistant to know. It's that easy. Then put the file in the thing
Dude thank you so much!! You have no idea how much I appreciate this!!
Any time!
Thank you for amazing tutorial. I wanted to make sure, in order for the chat to work on my site, do I have to have the replit server always up?
hey,can i know if you fixed that?
Thanks. I guess I will try this.
2 questions: 1. so actually if you need to modify instructions to the gpt you have to do through the code? Where it says instructions?
2. Can I use google tag manager to install this code on my website?
you didn't mentioned if you need to pay for it or not - and how you create the file knowledge base ?
what kind of parameters do I need to use . etc?
Pure Gold, great value! NEW SUSCRIBER FOR LIFE!
this is great. now I just need to figure out how to get user input saved as a form. easy million dollars right there because the possibilities are endless from there.
Did you figure that out @ST-actual?
Thank you so much, really second method worked like a magic. Thank you for sharing this video
You saved my weeks man, Thank you..
Great video. I can't seem to find the code and the other templates provided in the video description though. Where are they located? Anyone?
Thank you for sharing 🙏 Straight to the point!
Thanks! I had to do much customization on your Replit template, but now it's up and running! how can we implement it to take control of our Whatsapp conversation as well? I'd like to let it also take control over our Whatsapp during closing hours
Fantastic video and walk thrus!
Many thanks!
Спасибо большое! Понятное и познавательное видео! Объяснение понятней чем у Liam Ottley
also can you show us how you structured your knowledge base?. Should it be in Q&A pairs or paragraphs? Does the structure of the knowledge base and the file type matter in how fast and efficiently the chatbot can retrieve the right answers? If yes, please advise on what is the best way to structure the knowledge base..thank u
Great video! Pretty simple to understand
Glad to hear it!
Super nice tuto ! Bravo mate ;) Quick Question : Can this chat bot initiate a discussion ? Is it possible to set up this parameter on voiceflow ? also is it possible to remove the voiceflow watermark ? Thank you for your asnwers.
I am having trouble connecting the replit and voice flow. When copying the URL to paste into voice flow, I am getting the wrong URL that does not work. Anyone knows what I am doing wrong?
Thanks! Great video, great instructions!!
Glad it was helpful!
can anyone tell where is the code for method 2 i cant find it in the description
Thank you for the guide and resources. The response generated by the bot is not streamed, can you please help me with the approach I can use to get a streamed response.
Thanks. The set up went well, but when I run the chat the searching in the knowledgebase is very slow. It takes about 20 seconds for the chatbot to show the answer. Why is that?
cool.. what size of knowledge base you can upload to the bot?
You can attach a maximum of 20 files per Assistant, and they can be at most 512 MB each
Is there a way for the knowledge base to be dynamic? Like from an airtable etc?
This doesnt let me have multiple users at a time how do I fix that?
How did you solve this?
@@RomanH91 had to pay for replit core so I could use a more powerful server with more allocation. Luckilly this project is making me money even after paying for replit. If you want to keep it free, make your own home server (old laptop) and run it there
Hello, I used the api key in my free plan. I also changed the model to gpt-3.5 turbo. The get method works successfully, but when I respond, the assistant cannot produce a response and the run status falls into failed. but "api request failed" does not happen. When I try it from my upgraded openai account, gbt-4o still doesn't work, gbt-3.5 turbo works. Likewise, I get the run status failed error. can you help me?
Thank you so much! I am a total newbie and this worked beautifully.
hi i cant the the code in video it was told that it was in the description
Is it possible to do connect it to a GPT built on ChatGPT? so all the instructions and actions are there on the pre-built GPT and the integration only needs to talk to the GPT
yes, just published a new video showing how to do that kind of integration:
th-cam.com/video/CmFuiiuLpk4/w-d-xo.html
So where can I add the instructions of AI bot, a side of knowledge files, in replit or voiceflow?
Awesome video, helped me a ton! Quick question, Why would we use Replit to generate endpoints rather than using OpenAI api endpoints?
Hi man, thanks for this useful tutorial. Will you be updating this code to support Assistant 2 API and other models (GPT4o-mini)? It seems to use Assistants v1 API which will be phased out end of 2024
Hey great stuff Bo! Does this still work on December 2024?
Hi there. How can I get advanced-gtp-website-template ?
hello I had some ideas and I wonder if I could make this one with GPT customs, could you tell me what you think about it
Thanks so much for that video. Do you know what happens with the data entered by the user? For example if I ask for name and address, how is the privacy/gdpr handled?
both Voiceflow and OpenAI claim to be GDPR-compliant, but I have not personally dived deeper into their policies on this
Bro, you are so smart! I wanna be super smart too. This video saved me days and hours of coding. How can I thank you?
Thank you so much! This is a great tutorial! :)
You're very welcome!
In Liams script you work with the standard 'knowledge.docx' file, which is a bit unhandy. Can you change the docx to a pdf file for example?
Hi, i have encountered a issue, hoping you can help me understand where it went wrong. After following your steps, i works perfectly until I closed all the windows, the chatbot when it started it says "Failed to create new thread". Do i need to press save button or pay to keep it running?
thank you for this nice video. I have a pb: I can't get the URL from my project on replit??? any help pls
Hi Bo, thanks very much for the video.
One problem though - overnight when Mac went to sleep, when turned the Mac back on in the morning, the code was still running in replit, but it lost connection since the Mac was sleeping, meaning the chatbot was offline. Any solutions? Much Appreciated
Hey! To solve this you have to deploy your repel. If not deployed repels will shut down when the program isn't running.
@@Sharpthookay, but what is the link then? in tutorial you have to copy link after clicking webview, if you deploy repel there is no such option to check the link. Of course you have the deploy link, where your app is hosted, but it doesn't not work as the link from the tutorial. Any ideas how to do it?
@@iankhai that deploy URL you're talking about is the one you'll paste into the bot instead
@@Sharptho Thank you for your reply. Okay, I am currently using link in Voiceflow bot and it shows a status of 200. However, after sending a query to the bot, it only displays the welcome information and seems like it doesn't send the request to OpenAI. I tried link itself nad with /start and /chat also
Any update on this?
Great video! Just implemented it without any issue. A big thanks!
What are others thoughts on the cost of this way of implementing it? Other than the OpenAI balance, no hidden costs are applicable I assume?
I think voiceflow also charge for template
check out the voiceflow pricing: www.voiceflow.com/pricing
@@bo-sar great video - thanks! To check: I take it that a) the calls to openAI originate from replit so that the openAI key only needs to be provided in replit (not voiceflow) and b) - regarding costs - that all token charges originate from openAI directly & the AI token limits in voiceflow plans are irrelevant given this setup. Is that right?
Great info. I mainly wonder how the payed solutions scrape the website content. Cause it seems not so straight forward.
well, they've built a scraper... I guess the only way to find out how well it works is by testing it
Firstly, thank you so much. does this simply return word by word from the knowledge base or does it use its inference power to make up answers using the knowledge base?.
it uses its inference power to make up answers using the knowledge base
Wow. Thanks a lot Bo. In a short period, I was able to make things happen. It was my first time using these tools and Python for that matter, but was able to run my Chat Bot successfully. I just need to learn and understand how the code works and debug the constant connection error. Is there any way one can host the code in Replit in my own shared server?
where is vf template file, i can't find it?. Thanks
"copy the code provided in the video desciption". i cant see any code in the description,....
i have been developing ocr past month, the result i get is not 100% accurate so i want to integrate the 'Sider AI' their ocr into my wedsite, additional user input ill get as the output form. if possible can you make a video.
Thank you so much! Could you do an other video, how to implement function calling ? :)
Thank you for the video, really simple and easy 👍🏼👍🏼
Just have a question please, if I want to edit the instructions and the knowledge file, should I edit it inside replit or can I edit them from the Assistants panel in OpenAI?
I have a pdf that is a scan of handwritten data. I'd like ChatGPT4 Vision to OCR and perform calculations from the data in my tables.. then edit or append the pdf with the results.
If I understand correctly, there's 3 ways of doing it
1) Custom GPT using OpenAI Builder, can use plugins (ex. Ai PDF) and Custom Actions. If approved, App can be hosted and shared with other Plus users . Limited to the 40msgs/3hr
2) Using my own app via assistant API/Secret-Key. Can't use plugins and must be hosted elsewhere like Replit. Pay per use, so no message limit
3) 3rd Party tools like MindStudio/AirTable that use assistant APIs to create app. Can't use ChatGPT plugins, but can use other AI models (Gemini, Perplexity etc). Hosted on 3rd party sites.
Any recommendations how you would approach the task task of OCR / data analysis?
While starting a new chat conversation an error is appearing that failed to create a new thread. Please tell me how to rectify it.
Hey this video is very informative, can you show us how to integrate set up a live agent to this chatbot please.
Wow! Thank you so much for taking the time to put this together. Would it be possible to include TH-cam or Vimeo links in the uploaded documentation? If so, then, when the Chatbot answers if a particular video is relevant, it could give the answer as well as play the video? I'm thinking this might work well for a tutorial site where learning by watching videos is a big part of the process. If it works, one could reference "free videos" most of the time, and if the topic is too complex, it could reference a "full course" that they could subscribe to for more information. (hmmm, this might be a good angle for you!). Again, thank you for this video and thanks in advance for your response. (new sub here). 🙂
justo lo que buscaba, gracias!
i was just seen your video this was amazing.. thank you for good content, i also have a query can we do api calls with botpress also?
Can we embed it as a dedicated tool to a subdomain just like chatgpt dashboard ? and not as chatbot widget
Yes, need to modify the code
Did you remove the voice flow file?
Great tutorial, thanks. I changed the functions py so I could use multiple .pdf files as the data source and it works great. Question: for a simple question the cost is about 5 cents. According to the tokenizer, the question was 13 tokens. The response was 255 tokens. But the playground token counter (top right of thread) showed IN: 3950 Out: 294? Why is that?
Edit: So, I dug into this and apparently, the reason IN is 3950 tokens is because of 'context' tokens -- and I think the PDF doc(s) are being loaded into the question context? Now I'm really confused -- will the PDF source files be loaded as context for each question -- this will get expensive fast...
Hi I was wondering, can you share the code for multiple pdf's? Can you not use the pdf's loaded in the chatgpt 4 bot?
@@rodneyschaap I don't remember exactly as that code is long gone, but it was something like this:
import json
import os
def create_assistant(client):
assistant_file_path = 'assistant.json'
if os.path.exists(assistant_file_path):
with open(assistant_file_path, 'r') as file:
assistant_data = json.load(file)
assistant_id = assistant_data['assistant_id']
print("Loaded existing assistant ID.")
else:
file_paths = ["pdffile1.pdf", "pdffile2.pdf", "pdffile3.pdf"] # Add additional file paths here
file_ids = []
for path in file_paths:
file_obj = client.files.create(file=open(path, "rb"),
purpose='assistants')
file_ids.append(file_obj.id)
assistant = client.beta.assistants.create(
instructions="""This tool assists with information about ....""",
model="gpt-4-1106-preview",
tools=[{
"type": "retrieval"
}],
file_ids=file_ids)
with open(assistant_file_path, 'w') as file:
json.dump({'assistant_id': assistant.id}, file)
print("Created a new assistant and saved the ID.")
assistant_id = assistant.id
return assistant_id
Is this for including chat bot in your website or to make a website????
@bo-sar i'm completely new to replit and voiceflow so forgive me if this is easier then i may know lol. If i want the customer to upload images and ask questions about those images, how would i modify this method?
I'd use an additional step with Make.com automation.
which kind of business do you run?
Everything works great when I use the "run" button.
However, for the application to work continuously, it needs to be deployed. I deployed it on Replit, but then there is no way to check the link as you show at (8:33). I tried to use the production link that I have for the deployment, but then ChatGPT does not launch.
So, what link should I place in Voiceflow if I deploy on Replit?
BTW - great tutorial! ;)
Any news?
Hi, I've gone through all the steps and generally it works, but unfortunately it doesn't work the same as the custom-made GPT in OpenAI. I don't know what's wrong. The instructions and training I developed in my custom GPT worked perfectly. I copied them to the knowledge file where I have described how the chat should describe, what the maximum number of responses it should give, etc. Additionally, I added a database from which the chat is supposed to pull data about available products. Unfortunately, the end result is different than on the OpenAI platform. Is it the same for you?
Hi, sorry for the delay, I took a little break...
How different is it? Is it wrong/worse than the one by the custom GPT, or is it just using different wording?
Yeah it’s not the same. Looking for a decent ui setup to connect with my assistant or custom gpt. So far nothing
i have already had 'credentials' from my custom GPT in GPT store. how can i connect it to my separate app with assistant API
Awesome tutorial! The only thing is if I add new information in knowledge.docx, it is not reflected in the bot. Does replit save automatically the changes in this docx? If so, does the bot get updated with new info automatically? Thanks
that's a great question!
I think so, but I'd test it to be safe... just add a piece of unique knowledge and check it out
You can make changes to the knowledge base in the actual assistant on openAI's site
I have pasted that code in html but it is not showing me logo
Thank you for this!!
Any time!
Thanks man! Amazing content.
But I have an important question, what’s the knowledge base limit? I have a projeto with a lot of text files (like thousands) can I put it all? And also, using this method can I use links as knowledge base like what chatbase provide? And did you have to pay voice flow to use in this tutorial and no having the “created by voiceflow” message?
Thank you!
Thank you!
Voiceflow is wicked expensive; what would be the point? Was it less costly when the video was made?
How much is? do you think its more expensive than the method 1?
When I view the URL from web view in replit it shows up different, it look like a bunch of random letters/numbers and ends in .dev, not .co like yours. And doesn't work when I paste it into voiceflow. Am I missing a step in order to use the URL from replit?
no it should still work, same thing happens for me
can you do a video on this idea but for using dalle or midjourney?
tried to add my knowledge, but your example keeps coming up in th ebot. do i need to save/deploy or something?
Is this okay for a thesis project?
What if we already created our assistant using a vector?
Chatbase would be costly for sure, but getting Replit, Voiceflow, and openAI tokens would end up even more expensive. Although, it can give more flexibility.
Now if I keep training my bot on OpenAI will it continue to update the information on my website?
Great video, i just need how to call my API assistant with an R code without passing through the chatbot?can you help?
Hi,
I am trying to deploy my custom GPT to botpress but unable to do this. I am on the last step of deploying my GPT to bot press.
this is amazing
How can we put a restriction on the chatbot so that it only replies with information about what we want? Because there are some jailbreaks happening with gpt-powered chatbots where users can ask completely unrelated questions and the gpt will still answer them
good question, I'd try specifying the restrictions within custom instructions and then testing it
How do I add web urls as part of knowledge base instead of pdfs in replit
Need to build a scraping solution for that. I am planning on publishing a video on how to do it. Subscribe and stay tuned
Thank u! Could u tell me how to change the default. Hi , I am neogym’s customer support stuff.
Thanks. How to send and get responses with other app? No Voiceflow. Like PHP
I am planning to do more tutorials on this
If i use gpt 3.5, does the api key can be used?
yes
Hey, amazing video… I have a question though. I currently try to build a discord bot for using my gpt i built on discord and sharing it with only a few friends. I don‘t wanna share a link, cause i wanna be 100% sure, who’s using my gpt bot. Now I try to find videos on yt on how to connect my gpt with a discord bot. Could you do a video about it or at least give any advise? Id highly appreciate it!
How can I make it function in background?