Create a Customized LLM Chatbot on Your Own Data Using Vertex AI Agent Builder & Dialogflow

[@niteshsawant5754]

Since long I was looking for this and finally got it. Thanks.

[@techtrapture]

Glad it helped

[@changtimwu]

Thank you. This is probably the first demo of Vertex AI Agent Builder.

[@mpchris01]

You're the first to publish a tutorial for Vertex AI, Thank you from the heart. I was reading the instructions and found the agent builder and the Vertex AI console. The console is crazy awesome you can input your code and it will provide the SKULL coding structure for your Python, Back-end and front-end can be built after. The exemples i saw were from the GITHUB and the person used Jupyter Notebook. I heard it can be build in Collab also. Again Thank You

[@MrStroker_]

Fantastic explanation by the way! Can we connect any structured data formats (Databases) directly to the agent?

[@eddy.bruart]

Man, this video is so valuable ! Thank you so much for providing such incredible knowledge with a simple approach 🙏

[@techtrapture]

Thank you mate !

[@Quick_memes_blast-nf3wi]

Nice one, i couldn't find anyother tutorial that is better and reliable as this one

[@SundeepYalamanchili]

Very informative brother. Can we also include uploading the files in the frontend and ask the model to compare with the previous dataset. That would be really helpful.

[@SaurabhBhatt-vx8bq]

Awesome video, very informative !!

[@techtrapture]

Glad it helped you !

[@SaurabhBhatt-vx8bq]

@@techtrapture Actually I had one question after watching this video, if I want to integrate this chatbot in my app like google assistant and try using voice command to navigate through my application and the ai bot should be able to know on which page the user is in and what is the current content of the UI, so that it can answer accrodingly, is it possible to do so with this ? If yes , what will be the flow of that implementation ? Thanks in Advance !

[@techtrapture]

@@SaurabhBhatt-vx8bq this bot has voice feature integration available

[@SaurabhBhatt-vx8bq]

@@techtrapture Would love a tutuorial video on this if my question is relevant 🔥🔥

[@ayushranjan1265]

what kind of documents do I need to upload to get my chatbot give responses? I don't have a website as of now, but I want to keep my chatbot ready so that I can integrate it later. So the only way to provide information to my chatbot is to provide documents, please clarify

[@gamesofmoney3903]

keep it up😊

[@tech_fire_6043]

the information is worth it ..........................

[@JabariThompson-pg8eb]

Great video very helpful, thank you!!

[@hashjaami]

I want to create a chatbot or search engine based on my data stored in BigQuery. Is this possible? I want to get data insights through this chatbot in a quick session.

[@tech_fire_6043]

thank you very much brother .......................................................

[@SahilArifRavani]

can you explain hown to use website as data store?

[@robin2165]

Thank you for the wonderful tutorial. Is there any way to hide the document reference part from the chat? I gone through all the settings, but could not find it. If you know how we can hide it, Please share... Thanks

[@SmartBets24]

did you found solution to this, as it shows up source document link to download

[@robin2165]

@@SmartBets24 I did not get chance to work on it after that, Will resume work in couple of days and will let you know. Thanks

[@cafesoftwarevideos6011]

Nice demo, what if I don't want to link the source document? And additionally to the agent do we have access to the api to ask for example from a backend app ?

[@GoogleAIDemo]

Hi, the Generative Agent setting is not showing up in my account, Google documentation says that it has restricted access, Please let me know how I can gain access to the feature?

[@NeeharikaJha]

How can I additionally enable voice assistant in this? Also is there any functionality regarding multiple language detection? And can i train it using 2 features like : web urls and Cloud both ? Together? Thanks for the video, really helpful

[@techtrapture]

1. Voice assistant setting is in dialogflow ,I will check and let you know exactly where we can do it. 2. We can train on multiple dataset

[@EggseySuii]

Hi, really informative video! May I ask how can I remove/disable the custom payload? or rather, the part where the documents are also shown when given the result response to the user

[@techtrapture]

Even I tried this but did not get any option for this

[@EggseySuii]

​ sorry i think there was misconception on my part, I was talking about this part of the video 16:20. which talks about if the response is generated, the link to the document will also be shown and available to download. can that part be removed?

[@robin2165]

@@EggseySuii Hello... were you able to find the option to remove the document link? I am also looking for the same!

[@fahrisoft]

Thanks for sharing. its great but I cant see generative ai model selection under generative ai section in settings. so chat onluy answev about docs. is there any solutions about that?

[@AIWALABRO]

thanks brother for this video!

[@waqqashanafi]

Excellent explanation.

[@Ad-jq1jb]

Hi, thank you for a very useful tutorial! When I set Polish language, chat cannot answer my questions, but in English it works well. My training data is in Polish. Is it possible to set a language for chatbot?

[@theblueplanet3576]

Hi.thanks for sharing this video. Is there a way to export the trained model to my own vector database so can run locally?

[@mohammadyahya78]

Hello is this gonna be good chatbot that answers most question of my dataset compare to Llama and other LLMs model?

[@LeonardoFreitas-p7c]

hi, could u show or share the unstructured data please?

[@nishantkumar-lw6ce]

How can we integrate guardrails for enterprise? Can you share some ideas here?

[@DeepakGavhane-hl2be]

can i add both at the same time, website and local document data for the single chatbox??

[@techtrapture]

Yes , you can.

[@brandonhe6460]

In the test agent simulator I don't see any paramters?

[@CodeWithBro]

Hey, can I store a pdf in data source and then use dataflow api on it to create a chatbot from its data

[@MoeinIzadi-s3q]

how does it create a link to the data source in the published chatbot output?

[@techtrapture]

These are the file link which we used for custom Training

[@aarthikrishnan6503]

Thank you

[@madsrigby6288]

great video

[@jebbush2964]

google is trying to charge my credit card, is there a way to create an AI agent for free without using any paid software?

[@gowrimatadh6084]

great tutorial but while executing this i had no option of generative ai is it because i'm using it on a trial account

[@techtrapture]

Now you won't see that option to select model but you can select generative fallback with example prompt

[@gowrimatadh6084]

@@techtrapture thank you so much

[@manikantareddy_k]

Thank you so much brother, and I need pull the data, how I get the data

[@techtrapture]

Which data are you looking for?

[@GloryOfGodToTheHighest]

is there is a way to train the AI that can suggest java codes?

[@techtrapture]

Base LLM is pre built and trained on all coding queries

[@parthmodi8792]

hi im trying to import scanned pdf in data of agent builder but it always throws an error, are scanned docs not supported by agent builder?

[@joecapels]

Thank you!

[@sparkyUK]

Great video, thanks :) I want to create a chatbot that lets my customers ask about products from my suppliers. Do you know how I might return links behind the supplier names, or links behind product names? So when a supplier name is shown in the chat, it's actually a url to the supplier profile on my website?

[@robin2165]

Hello... I am looking to build something similar, Were you able to pull it off?

[@sparkyUK]

@@robin2165 Hi, yes, I created another column in my training data (Product Return URL) and as part of the prompt I ask it to return the URL but to use the product name in the link.

[@harshalyallewar]

very helpful

[@hesha912]

Hello sir, i havw created a dialogflow agent , and it has multiple branches /routes, in one branch i want to give a datastore based response, i added the needful datastore to my already created agent , since it was an already in progreds agent i could not use vertex builder. Now when i ask a question in that page whwre i linked the data store it does not give me a response, i request you ti create a solution /video on how can i add a data store to answet a query in the bot that is already created

[@Sizuka-b4h]

Please upload a video on sole tenancy node project

[@VIBEWITHRAJIB]

Can you suggest how to create those training files

[@techtrapture]

These are usual words or pdf files

[@MrAlex0803]

Like it too. Do you know someone who can help me build a Tool that can do an API call? The mission is that if the request can be solved or not it is synced into the ticket system either for further steps and refinement of the agent to make him able to solve or to just post it as solved to keep track of requests better. This is not commercial, I just want to build a prove of concept.

[@TrevorDBEYDAG]

There is something confusing here, you have created a chatbot - not an agent. I'm looking for deploying an agent on a website is it possible?

[@techtrapture]

If you have gone through video till the end , I have shown you how to deploy it on website

[@TrevorDBEYDAG]

@@techtrapture at 8:50 you are creating a ChatBot which is possible to embed on a website Dialogflow CX, but I have created "Agent" and it is not straightforward to integrate with DialogflowCX, it is asking me to create a chat bot rather than using existing one. Sorry to bother you with this matter but your title lead me here "Vertex AI Agent Builder & Dialogflow" I guess "Vertex AI ChatBot & Dialogflow" would be more compatible with the content. Regards.

[@SENG-lq4kt]

Hi I am currently doing my fyp now, I actually planning to build a web application by using spring boot, but I want put an Ai chatbot feature in my website, after watching alot of video I am planning by using the dialogflow to do this feature, and I saw ur video u put ur own data to train itself right? Can you show me the example of data? ( Like the format of data, how it should be written in the doc, how categorized the data). And when we upload the data into the dialogflow, meanwhile we are waiting, can you close the window? Or we must open it to wait until it is completely train? How can getting ur response, really appreciate thank you

[@techtrapture]

You can schedule a call with me to explain process.

[@SENG-lq4kt]

@@techtrapture how can i contact with you, by linkIn? But I didnt purchase the linkIn premium, or may i have your email? Appreciate, need your help

[@techtrapture]

You connect me on linkedin. Else book a call from here: topmate.io/vishal_bulbule/

[@ah87479]

Hi thank you for the video, but it looks like a search engine based on documents that pops out the exact answer from the input document. It doesn't look like a real generative AI that will interpret the inputs and display a constructed answer...

[@techtrapture]

Here foundational model already having generative capabilities, we are feeding external data for additional context related to our own data and expectations is to get exact answers. If you ground the model on lowest confidence level it started a constructive answer based on guess as well.

[@ah87479]

@@techtrapture Thanks for the answer, could you please give an example of it, and how to perform this kind of task ? Specially, to show case how efficient the model is becoming with external data ? I am asking this because Vertex AI (+Dialogflow + Gemini API) billing can rapidly become expensive. Thanks

[@techtrapture]

@@ah87479 even I thought it would be expensive but pricing is 12$ per 1000 queries asked. I would request you to test it out with your own data. I showed demo only with couple of files and it's far better with little data as well.

[@ah87479]

@@techtrapture Ok thanks for the demo

[@闲者-u7r]

HI my agent cant run in more local like it is not reply is it possible that sir you can provide the source code so i can check thnxxxx

[@DEVPandey-n5r]

As of now GPT is also able to web scrape the data to provide latest information.

[@DEVPandey-n5r]

Also will the sensitive data will interact with corpus of LLM when we select gemini-1.0-pro-001 in generative model selection, compromising the data?

[@DEVPandey-n5r]

BTW great Video !

[@techtrapture]

Glad to see such comments! Difficult to explain in details , but this use "RAG" mechanism to get reference to external data and not actually " retrain" the Gemini model on our data. When we working with our own ChatBot it checked reference in attached datastore and then provide results combining datastore result + LLM results. You can explore more on "RAG( Retrieval Augmentated Generation)"

[@DEVPandey-n5r]

@@techtrapture Thank you for the reply, So I referred to RAG Documentation . So we are passing the query and similar vectors from vector database which is created from our data. Again and again if we interact with the chatbot with different context of our documentation a healthy amount of our information is interacting with the corpus of the LLM. So ain't our data is still vulnerable to be compromised ?

[@techtrapture]

@@DEVPandey-n5r only that request and session data will access your data .

[@lokeshkcse]

Can you share your pdf data set for reference

[@shortjobupdate4589]

Hi sir is it enough free credit for creating the app?

[@techtrapture]

1 year you can create App using this. You will have 80000 INR credits for vertex AI agent builder

[@siddheshphaple342]

How to integrate it with local streamlit

[@techtrapture]

Same way , you need to add html code only

[@akshaydevadiga4916]

@@techtrapture can you please elaborate how to integrate with streamlit app please

[@prashlovessamosa]

Thanks Anna

[@moisesmansilla-l8y]

Hi, do you have some video or course that use RAG or similar?, we have about 1 million documents (word and pdf) and we want to ask them like you show in your video, if you can email me Thanks for share your knowledge and time. Great video

[@andresmelo7631]

can you avoid linking the document to the answer??

[@bocilmillenium7698]

how to make chatbot use local language? in welcoming message and in reply.

[@techtrapture]

You can refer cloud.google.com/dialogflow/cx/docs/concept/agent-multilingual

[@ShanawazAnwar]

how did you get 80,000 credits/rupees in your account ? when I created the free tier account, I only got 25,000

[@techtrapture]

You can verify here document mention 1000$ credits and i received same cloud.google.com/generative-ai-app-builder/pricing

[@ShanawazAnwar]

@@techtrapture thank you, I will try this

[@ShanawazAnwar]

@@techtrapture sorry for asking again, but what kind of document do I have to submit ?

[@techtrapture]

No documents required just enable vertex AI agent builder API

[@JohnDoeDude]

what is tech trapture? Are you trapping tech?

[@techtrapture]

Haha funny 😂 but it's for something different

[@rishiraj2548]

🎉