How to Build an AI Document Chatbot in 10 Minutes
HTML-код
- Опубликовано: 26 сен 2024
- 🚀 Kick-start your freelance career in data: www.datalumina...
Easily Build LLMs Apps - In this video, we are going to explore Flowise, an open-source UI visual tool to build your customized LLM flow using LangchainJS, written in Node Typescript/Javascript.
🔗 Links
flowiseai.com/
github.com/Flo...
github.com/dav...
www.docker.com/
docs.npmjs.com...
👋🏻 About Me
Hey there, my name is @daveebbelaar and I work as a freelance data scientist and run a company called Datalumina. You've stumbled upon my RUclips channel, where I give away all my secrets when it comes to working with data. I'm not here to sell you any data course - everything you need is right here on RUclips. Making videos is my passion, and I've been doing it for 18 years.
While I don't sell any data courses, I do offer a coaching program for data professionals looking to start their own freelance business. If that sounds like you, head over to www.datalumina... to learn more about working with me and kick-starting your freelance career.
👉🏻Join my FREE community if you want to learn more about AI: bit.ly/data-alchemy
Awesome stuff, Dave! Loved how you delved into Flowise and LangchainJS. Hoping to see some comparisons on AI Agents, especially long-running and autonomous ones in Langchain. Keep it up! 🚀
Can I integrate this into Flutter APP as a chat FAQ?
Feel free to reach out any time! @@SmartWizzard
Any solution to build a CPU-based AI Document Chatbot which utilizes CPU and does not need API ?
Hi Dave, thanks for sharing. Flowise is pretty cool as a rapid-prototyping tool, you can visually chain complex objects in just a matter of minutes. I found some cons though, which I think the devs should address.
- Debugging is tough. Each API-based component should have a test button (does the key work? is the deployment name correct?). Those aren't things you want to guess from pieces of text in the chat box.
- Coding. It's nice to have a visual interface, but at the end of the day this is just chaining function calls. It would be nice to have the Python code corresponding to a specific saved flow. So that one can customize it.
I wonder if they are interested in getting precise feedback from users to improve and prioritize features. I'd be really kin to help them build a great product.
Agreed. "Error: Request failed with status code 401" in the chatbox dialog isn't particularly revealing...
great content. Reminds me of the same concept as node-red, but specifically built for AI apps. I think this can go a lot further than just a prototyping tool, seeing as how far node-red has come from where it started. Really neat.
❤ I love all of your videos, and especially this one. As a no coder entrepreneur, I really appreciate this kind of step by step tutorials. Many RUclipsrs are making similar ones, but yours exceeds. Only two points. To your surprise, one of the hardest part of this kind of tutorials is, setting up the apps. The process is no brainer for coders and data scientist, but not for no coders. Two. I always wanted to see a tutorial covering end to end process, meaning designing GUI to (e.g., Figma) to connecting to algo (e.g., Flowise) so that I can build a completely functioning app or MVP. I’d appreciate if you could do this type of video. 🎉
Yes! setting it all up for we beginners is tricky. It takes a while to get familiar with Git etc
Yes, I agree. I also looking on it
+1 For many videos I can't even go pass the point of the setup because they go so fast over it thinking we know about it, but we don't. Guess what, the people who get that, are not here trying to watch the video tutorial. Come on guys.
Hi Dave, my problem is most companies won’t be happy uploading internal documents to a external server. Do you think you could make a video where a chat bot is created but the company can keep there internal documents on their end 🔥 🙏
I already made that before chatgpt even existed
@@airo_robot what’s it called?
Yes please, this!!! Jake and I have the same problem.
Hey, there are ways to do this using the stack explained in this video. For example, you could use the Confluence API document loader to query documents stored in your Confluence site
Hire an app architect if you don't know how to solve this basic problem.
As a non-coder, I am flabbergasted. This looks so easy! Thanks for sharing this. How can I turn this into a function Q&A widget on my webpage? 😳
ask ChatGPT
Can you make a video talking about the cost involved with building AI apps with langchain and OpenAI? Are there strategies that can reduce costs? How many tokens does it take to parse company data and return insights? Great content. Best tutorials in this space that I've found. Thank you and keep up the excellent work!
Great question and fully agree Dave is amazing
i mean those tools were made with using that AI tech as the cornerstone for. a business. so i imagine it is somewhat affordable
For anyone setting this up now and experiencing a "TypeError" when they send messages in the chatbot- removing the document module and adding it again worked for me.
Hi, many thanks for the video. How did you prepare the txt input file? Do you have a video or link to understand the contend and how to prepare it?
Thanks
Copy/pasting the "connect credential"s is different now than how the video shows. In the Flowise UI, you can no longer copy/paste the credentials. You need to click on the Connect Credential field and select "create new" which opens a popup window. In that window, put in your username and then finally you can copy/paste the API key.
Really good video. For beginners a video on how to use Git etc would help just to get grounded in the concepts of how to use VS code, download repositories and so forth would really help. Thanks for all your efforts, much appreciated.
You know your stuff. But I am struggling to follow this to replicate its a lot of detail.
LOVED this tutorial- thanks for sharing!!
Great video! I couldn't find the video you mentioned at 6:07, would be nice if you could link to it.
I was looking for this in your channel yesterday. Todays youtube algorithm rewarded me with this suggestion haha. Thanks for this valuable content. Greetings from Argentina! New suscriber =)
That's awesome haha - thanks!
I'm getting the error TypeError: Cannot read properties of undefined (reading 'startsWith'), anyone knows a fix.
damn flowiseal put me on a waiting list.
Issue I'm having with these document loaders and vector stores is they get soooo slowwwww once you start loading actual data into them. At around 500 MB of pdfs and vectors, these vector stores take so long to query the whole app breaks down. Not to mention cost issues if you go down the openai api / pinecone route.
Thanks for the videos Dave! Keep them comming. Also noticed you doubled your subscribers since you made the video, congrats. Your conversational agent snitched :)
Hi, you are very clear, but you use and refer to so many things I have never heard of. Where can I learn what I need to understand your tutorial?
Lars, Norway 😊
I'm also on the Pinecone waiting list. Do we have any other options other than Pinecone? Thanks
I would like to train a large language model to evaluate scientific publications, specifically regarding research, based on its adherence to the scientific method.
You are a great friend, I really like the information and knowledge you give us.
Great information, here from Nash's channel 🙌🙏
Thanks man!! 🙏🏻
Can one link this to a local LLM like Ollama to reduce the costs?
Hey Dave, I just watched your video and I must say that it was really informative, very useful and well-made. I loved your videos. I was wondering if I could help you edit your videos and also make highly engaging shorts for you?
What would you recommend using: Stack-AI, Flowise or Langflow?
We have so many videos like this
Can't you try showing how to get both video and text as answer for questions based on pdf document
Hello, how are you?, very good video. I ask you a question, how did you create the embedded chat, what did you use to generate it?
Love it. Thanks for making these videos. This seems great to slap some components together then go in and have a lot of the code roughed in so it could be worked on from there. Is that an accurate assumption?
Excatly
Great video thanks. As a no coder myself I have been looking at the best ways to utilise chatbots for my clients. flowise still requires a number of steps and some low code know how, but now we are seeing chatbot solutions requiring no technical ability ie no code become available, Orimon, chatbase to name but a few. Maybe not a flexible as flowise in terms of functionality. Your thoughts?? BTW Subscribed.
Dave, this was great. Thanks. I now have my own flowise agent that can read the thousands of pages of planning regulations and make sense of them. Would you have any guidance on best practice for document managment, e.g. how to ensure the bot has the latest version of a document, how to audit the currency of the docs in a document store or even be reminded when a document is out of date.
How much do you have to pay per search to the GPT provider? One metric i saw had chatgpt-4 at 30 cents per search.
Is it safe in terms of the companies data or will another model use the data to train? For exmaple over the open ai API?
An addition to the setup if (a) you are running on Apple silicon (i.e. M1 or M2) and (b) you want to use Docker compose.
When I first went to the docker setup and issued the command
$ docker-compose up -d
It the docker container for flowise failed to start because the platform was not specified.
The solution is to add the platform specification line to `docker-compose.yml` immediately after the line `image: flowiseai/flowise`
My file now looks like this:
```
version: '3.1'
services:
flowise:
image: flowiseai/flowise
platform: linux/amd64
…
```
i had to comment it out to work on my m2 machine
Thank you for this amazing video. Would you be kind to build an app that would refer to a given website and reply based on that site's data?
hi tnx for the guide , i tried to bulid the doucument chatbot , i uplade txt file , when i asked "what this document about " he answer "i dont know "why its happened?
I clicked on the docker folder and ran 'npx flowise start' in the terminal and it worked fine. Already made a chatbot
What are the costs to use it?
Its actual 16 minutes. For starters probably 45 min. Still amazing.
Hey there. Love the content. One comment if I may. Maybe show us in the begining a glimpse of 10 sec how the end looks like. then continue with the video😊
Thanks for the tip!
After watching several times it looks like I might need to be a developer or software writer to complete this. Are any others in this thread totally new to A.I. and have made a complete chat bot ? Does The material loaded become public ?
Thank you ❤
I just bought the AI writing assistant chat bot. I can not do folders. II can not do that stuff on the left siide. I'm paying $14.99 a month. Which cat bot should I be using for this.
Great video :) Keep up the great work.
Thanks 🙏🏻
Very cool! Many thanks!
Very cool video. Content and video quality
Unfortunately doesn´t work for me. It always responds with "Hmm, I´m not sure".
Good video Dave, thanks!
Thanks Hans!
multiple pdf learning also available?
This is like selling the tools to make the shovels in a gold rush
What about the data privacy? Where does your document go?
I need troubleshooting assistance. I csn get everything connected...independently openai will respond. When i link pinecone...it will receive the vectors from the pdf when connected...but he dumb ass assistant when asked about the document says hmmm i dont know! 😅 is ir the splitter overlap...its a larger pdf...i do legal work so its case law in trying to load and ask questions
Like your videos! Subscribed! Just wish you'd ditch the background music.
Hi Dave, thanks for this amazing video. I have a few question, is it possible to teach to our chat bot to ask some specific questions to make a 2 way conversation? I want to use this chatbot as a sales rep in our website, talking with our clients. Also is it possible to teach the chatbot to not saying some specific things. For example I am afraid it recommend our competitors to our clients :)) I guess this chatbot is linked to Chatgpt and uses the same dataset, so it can use its own dataset to say something and offering our clients to the competitors
tnx for info.... can u help me understand the differencebetween there two Artificial Intelligence & artificial intelligence solution !!??
cuz i want to choose a university bachelor's degree???
and what do you think i should learn?? ML /// deep learning/// python //data analyst //data science...... ???? TNX
hi. please help me. how to create custom model from many pdfs in Persian language? tank you.
Hello, 🎉 great content! I would like to request a video on a solution for chatting on PDFs using an open-source model from Hugging Face that utilizes GPU and ensures data privacy.
Very cool, thanks!
I tried to do the same with some (large-ish) PDF's but I'm getting `InsufficientQuotaError: 429` from OpenAI ;-(
Thanks for the great content! If I have big pdf files isn't it will get pricy?
Hello Dave, I am asking here as this is your latest video, I have seen your videos regarding Langchain, I am curious how to optimize inference while using LLMs in Langchain using TensorRT or Onnx Runtime... As in Industries, It's very obvious, to save the time as well as computation cost... In TensorRT while using open source models, we have techniques like quantization and a few more more for that... So In Langchain is there any way to do this?
multiple pdf, docs at same time ?
Flowise is not stable. as I testing, put API keys, however can't work
This is pretty cool. How could we use this to chat with whole databases full of records instead of just a single file at a time?
Are you uploading your data to the cloud or into the docket container?
How to use without Open AI API? I see some Chrome plugins doing this with ChatGPT 3.5. The API is very expensive and not worth it. More thanks for showing this tool, Success!
I tried your exact workflow on windows and on mac, but never been able to get it going. On mac all i got was 'Error: Error: Error: ENOENT: no such file or directory, open'
i have an idea for a video series for you
"gpt-4-code interpreter vs gpt-4 with combo of plugins"
Thanks, this is amaizing!! do you know if there is an alternative to Pinecone since it has the waitlist in place?
I'm also on the Pinecone waiting list. Do we have any other options other than Pinecone? Thanks
Hi Dave, thanks so much for this very good video. One question. just wondering about privacy and if information written in our documents are passing to OpenAI while using its APIs. Good part of data are also going to a third party on cloud service such as Pinecone. How can we ensure the privacy of our personal data, while using Flowise or similar services that make use of external API?
Curious too
it keeps saying this for me: * error decoding 'ports': No port specified: : how do i fix that?
This is way too good wtf
🙏🏻
Can you add custom prompts?
What happens when you ask unrelated questions ?
Nob question: How is this different from building it on Botpress?
This is the hardest tutorial I've ever seen. It's not a step by step.
Thank you!
I'm getting this error everytime: ReferenceError: Blob is not defined
Great content!
I am getting an error while using flowise - ErrorWithoutStackTrace: PineconeClient: Error getting project name: SyntaxError: Unexpected token A in JSON at position 0. Tried to cross check but not sure what is going wrong. Could yo please help?
10 minutes if you have quite a bit of experience and already have accounts set up. Definitely not 10 minutes for a newb.
Lol.. I love that this is the part you are unhappy about. 😂 It's a visual interface to code langchain etc and natural language models without needing to learn python and other languages.
TypeError: Cannot read properties of undefined (reading 'startsWith')
thanks for this great tutorial, am frontend developer so how can i build SAAS on this?
i want to give one like 3000 academic papers and make it able to be queried, is that too big?
Do Flowise have any intellectual property over your data if you make a chatbot with them and train the model yourself?
How does it deal with the structure of a PDF? For example, if I have a Table with CAS numbers in the format 1234-56-7. Usually, I can use Regex to extract these patterns however what if the text is contained in a table column and broken by a carriage return or just generally by the structure of the document?
e.g.
123-
45
-6
In a single column.
It seems that the text file generated will not keep this text together. This is a frequent issue I run into when extracting data.
Error: Error: Error: ENOENT: no such file or directory, open , any idea how can I solve this error ?
Thanks!
Appreciate that!! 🙏🏻
i wonder if you could get flowwise to show its generated python code
once ive made my bot, how can i get the bot online so its not just accessible from my laptop using the embed function to connect it to my site but then it only works on my laptop
Don't know why but my pdf was of 818 pages and it wasn't working accurately as I was expecting it to work
Hi, Can you please advise how to use the Folder Path in "Folder with Files" document loader. I've deployed the Flowise app on GCP as docker image. Thanks
any way to get around the Azure OpenAI key requirement
I followed all the steps to the end. I uploaded the text file, ran to the chat, and asked it questions. All I received back were error messages such as "Error: Request failed with status code 429"
Me too. It's coming from Open AI and it says I'm exceeding my current request quota. I'm on the OpenAI free tier which allows you only 3 requests a minute. So I'm wondering if that's the problem. But when I look at the analytics for my account it doesn't show any requests.
@bwp2bruce did you figure out what the problem is?
@Curious.Gamer. Did you figure out what the problem is?
@@lancerkind Eventually yes, but I'm not going back to that to recall the steps.
I get the following error message when I try to chat to bot: Error: Error: Error: ENOENT: no such file or directory, open ''
Is there any idea why?
Awesome video. After watching I'm trying to go a step further and use pinecone with ConversationalRetrievalQA with langchan Js directly. I'm getting a response for the query, however for some reason the buffer only always stores the latest messages. The history of messages is never stored so I cant build on prev messages. Do you have any example for successful implementation?
I build with flowise and I believe it was aware of the whole conversation. Maybe something to do with the setup
Thanks for the video, it's very useful. Is it possible to integrate a voice assistant that receives a question as input and answers via voice, using the information in the pdfs? It would be very useful. What do you think about it?
Yep, that's possible. Look into the speech services from Google or Microsoft.
can we add system message in query
Impressive ♥
Thanks!
Thanks for showing this. Can you videos on how to deploy using other things besides Azure?
Maybe in future videos, but I am an Azure guy ;)
Can show it with chromadb as its local and free for most people? Thanks for great work.
I keep getting the following error: anyone know how to solve?
PineconeError: Failed getting project name. Error: Forbidden - Invalid API key