03 How to create a custom Gemini chatbot trained on your website content

LLM Powered Search

Just like you can use GPT to create a custom chatbot trained on your website content, you can also use Google’s Gemini to create a similar chatbot.

However, it is not quite as easy for non-programmers as of this writing.

Here are the basic steps to create a custom Gemini chatbot trained on your website content

  • upload the HTML contents of each webpage to a Google Cloud Storage (GCS) bucket
  • create a metadata JSONL file which describes all the files uploaded to the GCS bucket in the previous step
  • create a Vertex AI datastore based on the metadata JSONL file
  • add this datastore to your Dialogflow CX chatbot

Now we can select a suitable generative model for our data store response.

This is the response when I select Gemini 1 Pro as the generative model:

And this is the response when I select Gemini 1.5 Flash as the generative model:

As you can see, for this particular use case, Gemini 1.5 Flash provides much better responses.


About this website

I created this website to provide training and tools for non-programmers who are building Dialogflow chatbots.

I have now changed my focus to Vertex AI Search, which I think is a natural evolution from chatbots.

Note

BotFlo was previously called MiningBusinessData. That is why you see that watermark in many of my previous videos.

Leave a Reply