Zain-ul-din
commented
1 year ago You could create a custom model to achieve this functionality.
Creating a Custom Model
To create a custom model, you need to add a new field in models.Custom. The modelName field specifies the name of the model, and the prefix field specifies the prefix that the model uses. The enable field determines whether the model is enabled or not, and the context field specifies the context that the model uses to generate responses. The context can be a string of text, a file path, or a URL.
modelName: a string that represents the name of your custom modelprefix: a string that represents the prefix that messages should have to get a reply from your custom modelenable: a boolean that indicates whether your custom model should be enabled or disabledcontext: a string that represents the context of your custom model. This can be one of the following options:"your_context": a string that represents the context directly"path to file (.md,.txt)": a string that represents the path to a file containing the context"url": a string that represents the URL to a website containing the context.
{
modelName: "your_model_name",
prefix: "!your_prefix",
enable: true,
context: "your_context" | "path to file (.md,.txt)" | "url",
}Test your model
- run server
yarn dev - type a message starting with
!your_prefix.
Demo
- Name you're model let's call it
bot - Add prefix
!bot. - create
bot.mdfile instaticfolder. - add context
context: "./static/whatsapp-ai-bot.md" - set enable to
true - add the following context
Hey GPT, if the provided question seems to be asking for an image then just return the question with the prefix !dalle otherwise return the question with the prefix !chatgpt
Examples:-
question:
Generate an image of a black cat with light green eyes
you should return => !dalle Generate an image of a black cat with light green eyes
question:
what is HTTP?
you should return => !chatgpt what is HTTP?
Example End.
Results
Note! Although this approach will work. But, there is an overhead check this diagram how this process is working.
See more about how the custom model works under the hood
Apart from that, To achieve this kind of functionality we need to use an NLP model on the server side which may slow down the response. Note! All models being used in this bot is not free any mistake may cause a loss of money.
Maybe in the future, I plan to use a custom TensorFlow NLP model
or
Another option is: Google is releasing a new AI model Bard which can respond to messages in latency of seconds.
ZeanArd
You can make a single prompt and let your model choose which model will suit this input text.
Input
!bot Generate an image of a black cat with light green eyesInput
!bot what is HTTP?HTTP stands for Hypertext Transfer Protocol, which is a protocol used to transfer data over the internet. It is a standard application layer protocol that defines how data is transmitted between web servers and web browsers.