Build a Smart Chatbot in MINUTES! Voiceflow KB + Gemini Pro Integration for Optimized Responses
ฝัง
- เผยแพร่เมื่อ 17 ก.ย. 2024
- Want to build a chatbot that actually understands your users? Look no further! In this video, you'll learn how to create a powerful chat assistant using Voiceflow and integrate it with the cutting-edge Gemini Pro model for optimized responses.
Here's what you'll learn:
No/Low-code chatbot building with Voiceflow: Master the basics of building conversational flows with no or low code.
Fetching data with ease: Leverage Voiceflow's powerful example function to seamlessly retrieve information from your knowledge base.
Boost response quality with Gemini Pro: Supercharge your chatbot's responses by integrating with the AI-powered Gemini Pro model for natural and optimized language. ✨
This video is perfect for:
Developers: Learn a new, low-code approach to chatbot development.
Content creators: Enhance your audience engagement with a smart chatbot.
Business owners: Automate customer service and improve communication with your clients.
Keywords: #chatbot, #Voiceflow, #Gemini Pro, #AI, #natural language processing, #knowledge base, #no-code, #machine learning, #customer service, #content creation, #automation.
*********************************
AI Modles:
• Gemini 1.5 Pro: • Unleash the Power of G...
• Gemini 1.5 Flash: • Real time chat with we...
• Gemini Pro: • Elevate Your Voiceflow...
• Gemini Pro Vision: • Elevate Your Voiceflow...
• Mistral 7B: • No-Code AI Chatbot! Bu...
• Mistral Mixtral 8x7b: • Build a FREE AI Chatbo...
• Mistral Large: • Supercharge Your Chatb...
• Gemini VS Mistral AI: • Gemini vs Mistral: AI ...
• Stable Diffusion XL (Text2Image): • No-Code AI Chatbot! Bu...
• Meta Llama 3: • Dive into the Future o...
Automation:
• Make: • Create an AAA ChatBot:...
• Notion: • Create an AAA in Voice...
• Zapier: • Automate Lead Gen! Cha...
• Google form: • Revolutionize Your Sur...
• Calendly: • Boost Your AI empowere...
Knowledge Base:
• KB: • Build a Smart Chatbot ...
• KB with function: • Mastering Voiceflow Fu...
• KB with {_memory_}: • No-Code AI Chatbot! Bu...
• KB with {vf_memory}: • Build a FREE AI Chatbo...
• KB + {vf_memory} + Llama 3: • Level Up Your Chatbot!...
• KB + {vf_memory} + Gemini 1.5 Pro: • Supercharge Your Chatb...
• KB + conversation history + Mistral Large: • Supercharge Your Chatb...
• KB + Live web content + Mistral Large: • Supercharge Your Voice...
• KB + User defined content + Gemini 1.5 Flash: • Real time chat with we...
Voiceflow Basic:
• Auto popup, proactive massage, form extension, web integration: • Voiceflow chatbot: int...
• Text Block: • Mastering Voiceflow Te...
• Image Block: • Mastering Voiceflow Im...
• Functions Block: • Mastering Voiceflow Fu...
• Save ordering time and date: • Create a ChatBot with ...
• Dynamic Carousal: • Create an AAA with Voi...
• Dynamic cards and lead generation: • Create a Bot in Voicef...
• TH-cam video: • Supercharge Your Voice...
• Product recommendation and Stripe payment link: • Voiceflow Chatbot Prod...
• Display image, audio and video: • Level Up Your Voiceflo...
• Upload files: • Add File Upload to You...
• Custom chatbot CSS style: • Level Up Your Voiceflo...
Data Base:
• Airtable: • Master Voiceflow API C...
• Notion: • Create an AAA in Voice...
• Zapier: • Automate Lead Gen! Cha...
• Google Sheet 1: • Create a Bot in Voicef...
• Google Sheet 2: • Create an AAA with Voi...
• Google Sheet 3: • Create a ChatBot with ...
This is the best channel about chatbost and voiceflow!!! Thank you for all what you do!
Wow, thank you!
please make video for Voiceflow KB + Gemini 1.5 new one will be so helpful
Will make one.
@@FarthinkAI thank you so much 😊
Great video, please tell me Is the Gemini API free?
Yes, it is
I have followed everything exactly and everything gives me 0 results...
Two doubts: it will always continue to be free or will we have to pay at some point...
The other question has to do with me or the fact that I am using it in Spanish, the knowledge base is in Spanish and the questions in Spanish are in Spanish, sorry if they are naive questions, I am learning
The free version should work fine. You may check the chunk data to make sure the opening and closing quotation marks are paired. This has happened for many audiences.
I can market a job done with Gemini 1.5 pro. It is working well for me but I am worried about the error that disconnects when it reaches that quota... can you guide me please?
Currently, there is a rate limit of 60 requests per minute. In your case with heavy usage, my suggestion is that in your chatbot you may use two APIs from two Google accounts. You may monitor the number of requests of one API. When it reaches the rate limit, switch to the other API to return the answer to the user. Hope this helps.
Hey @farthinkai, another question, how did you create a variable?
You may watch our previous videos in the playlist: th-cam.com/video/8fi65X4zRNs/w-d-xo.html&pp=gAQBiAQB
Hey, you've been putting out great contents. Keep up the good work! By the way, where can I send you business inquiry?
Thank you! Will set up a consulting service soon, and put the contact info in the description.
Thanks for this tutorial, I just have a question! If I'm using Open AI API, it's roughly the same steps right?
Yes, it is.
@@FarthinkAI So far, the chunk retrieval and everything is working, but I'm getting a message saying "API call error - {"status":400, ...{ "message":"We could not parse the JSON body of your request. (HINT: This likely means you aren't using your HTTP library correctly). The OpenAPI API expects a JSON payload, but what was sent was not valid JSON0."
You may check the chunk data to make sure the opening and closing quotation marks are paired. This has happened for many audiences.
You may check the chunk data to make sure the opening and closing quotation marks are correctly paired. This has happened for many audiences.
It's excellent but I'm already stuck, in the integrations part where you have to take the knowledge base api, it's not there, I only get qwebchat and api keys... I don't get the knowledge base api
Now the primary key is the knowledge base api key
hi very informative !!
can u ellaborate in detail , the consumption of tokens used in this tutorial of Gemini Pro (this) VS Openai 3.5 turbo (if used) ??
how much is the difference ??
Other Pros & Cons
Thanks
The tokens used depends on the AI models, and also maybe the prompt used. I have not looked into it yet. Will need to check. Thanks for the discussion.
Great again. I am following these videos. thanks a lot. but I have a problem. With gemini, how can I add memory. So the agent can establish context. without memory, It is not useful unfortunately
Please watch these videos on memory: th-cam.com/video/5oX5TepJMHA/w-d-xo.html
th-cam.com/video/xQwCePHeBRA/w-d-xo.html
@@FarthinkAI Can I apply it Gemini?
Yes.
Also, I can not find your VF API key video. Do you have one for that?
Now the only API key can be used as the KB API key.
@@FarthinkAI I have been through this video a few times and am getting an error that I was hoping you might have some insight on. It says Unable to resolve code - "ReferenceError: prompt is not defined" Any ideas? I am sure it's something minor.
So I eliminated the concatenating step, and the bot worked. Did you have specific instructions for that concatenating step?
You may need to create the prompt variable before using it.
I'm getting error. I think its due to the substring issue. Where do you put the substring?
I can't find place to remove the square brackets on the firsts and last position as mentioned on the video on 11:25
{status":400,"data":{"error": {code":400,"message":"Invalid JSON payload received. Expected or} after key:value pair.
sed on the employer's provisions. New ZE
status":"h3=":443"; ma=2592000,h3-29="443";ma=2592000","cache-control":"private","connection":"close","content-encoding":"gzip","content-type":"application/json;charset=utf.....
You may check your chunk data. Some times the data from the KB have double quotation marks, which can cause wrongly paired.
I can only see the standard api key in integration settings not the knowledge base api key
Click the integration icon, and click the API keys, and you will see the primary key, which is the knowledge base api key
Great tutorials! I am just testing this all out currently and want to try and stay on the free version of Voiceflow. Is there an alternative to the javascript prompt to concatenate the prompt and the chunk data? Javascript blocks are pro-user only feature. Thank you.
Yes. You may add the prompt in the function, and return it together with the chunk data after concatenating the prompt with the chunk.
Hey @@FarthinkAI hope you are well. Could you please tell us where to add it in the function?
Yes. Right before you return the chunk, you may use: chunk = prompt + chunk.
please share the JSON structure in the comments
response.candidates[0].content.parts[0].text
While testing the function I ak getting empty chunks
chunks [ ]
Please help me. I am not using URLs but using .docx as my database
On the VF docs, it says: "add as many PDFs, URLs, or Text files as you want", which does not include docx. learn.voiceflow.com/hc/en-us/articles/15965626139149-Step-2-Add-Data
@@FarthinkAI you can upload docx as database. docx is working fine with inbuilt system but when I try it with Gemini it doesn't work but after hosting same database on a website and providing it's url to train it starts working 😞
why it returns 0 with the answer and the chunks []
It looks like the chunks are empty, and the answer variable is also empty.
@@FarthinkAI do you have any idea why i got this error, because i get the same error with the mistral 7b and mixtral 8x 7b too
If so, you may debug, and make sure the chunks variable has correct data.
i can't find body tab in api. What to do?
You may change the API type to POST instead of GET.
Please send templates
Will set up a link.
drive.google.com/file/d/1Tx0SPibtImJTXRVp8Rm83vPmazv-zTIM/view?usp=sharing
Tried to follow along and received an error at the end. Says "ReferenceError: prompt is not defined." Not sure what I did wrong here.
Code: prompt += chunks.substring(1, chunks.length - 1);
Body
{
"contents": [
{"role":"user",
"parts":[{
"text": '{question}'}]},
{"role": "model",
"parts":[{
"text": '{prompt}'}]},
{"role": "user",
"parts":[{
"text": 'Please find {question} using the provided information.'}]},
]
}
Capture Response:
response.candidates[0].contents.parts.[0].text
Have you created the variable "prompt"? You may check the list of variables to make sure it is there.
@@FarthinkAI Just added it as that was part of the problem. Now it is saying API Call error status 400. Invalid JSON payload received.
You may check the chunks, and make sure the opening and closing quotation marks are paired. This has happened for many audience.
@@FarthinkAI I'm not sure how to check that. Is it possible for you to send me your template?
hello, good work ;) I'm hoping to make it work. I keep getting this API call error with invalid JSON payload received. any idea ?? [API call error -
{"status":400,"data":{"error":{"code":400,"message":"Invalid JSON payload received. Expected , or } after key:value pair.
positions Marc's offerings as innov
^","status":"INVALID_ARGUMENT"},"VF_STATUS_CODE":400,"VF_HEADERS":{"alt-svc":"h3=":443"; ma=2592000,h3-29=":443"; ma=2592000","cache-control":"private","connection":"close","content-encoding":"gzip","content-type":"application/json; charset=UTF-8","date":"Tue, 30 Jan 2024 13:27:01 GMT","se...]
As shown in the video, you may use single quotation marks (') instead of double quotation marks (") to wrap up the variable, and check the returned chunk data to ensure the opening and closing quotation marks are paired.