README.md

metadata

title: ChatGPT Plugins UI With Langchain
emoji: 📊
colorFrom: yellow
colorTo: yellow
sdk: gradio
sdk_version: 3.36.0
app_file: app.py
pinned: false
license: mit

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

Steps to add new Plugins to your Langchain-Gradio ChatGPT PLUGIN WebUI

1. Acquire the API Endpoint

   • You need an API which you can query, and for this example let's consider using a The Bored API.
   • API Endpoint: https://www.boredapi.com/api/activity/?type=

2. Create a Function to Query the API

   • You can access any Gradio demo as an API via the Gradio Python Client.

   def bored_api(activity_type) -> str:
     '''
     Get a random activity to do based on the activity type. 
     '''
     activity_type_list = ["education", "recreational", "social", "diy", "charity", "cooking", "relaxation", "music", "busywork"]
     activity_type = activity_type.lower()
     if activity_type not in activity_type_list:
       activity_type = random.choice(activity_type_list)
   
     api_url = "https://www.boredapi.com/api/activity/?type=" + activity_type 
     response = requests.get(
         api_url 
         )
     return response.json()['activity']
   

3. Add Function definitions

   • Add the function definition to the gpt_function_definitions.py file (simply copy and paste). Don't forget to add function description in docstring.
   • Add required imports

   from gpt_function_definitions import generate_image, generate_caption, get_news, bored_api
   

4. Add the function to the Tools list

   • Add a description - describe what your function does. Models like GPT3.5/4 support Function Calling. The OpenAI Functions Agent from Langchain is designed to work with these functions and models.
   • Name - add a name of your function, don't include spaces

   tools = [
       #image generation 
       ...
   
       # Describe an image
       ...
   
       # Get lattest top news
       ...
   
       # Bored Api 
       Tool.from_function(
       func=bored_api,
       name="bored_api",
       description="Get a random activity to do based on the activity type"
       # coroutine= ... <- you can specify an async method if desired as well
       ),  
   ]
   

5. Update the Chatbot Layout

   • Go to the Blocks Chatbot layout and add a new checkbox for your plugin as:

   bored = gr.Checkbox(label="🙄bored", value=False)
   

   • Add the new checkbox component (example - bored) to your submit and click events for your chatbot and to the predict function accordingly.
   • And also to the plugins list in predict

     plugins = [stable_diff, image_cap, top_news, search, bored]
     

Thats it! you have added your own brand new CHATGPT Plugin for yourself. Go PLAY!!