app.py

from llama_index import LLMPredictor, PromptHelper, ServiceContext, GPTKeywordTableIndex
import gradio as gr
from llama_index.indices.knowledge_graph.base import GPTKnowledgeGraphIndex
import os
from langchain.chat_models import ChatOpenAI
from langchain import PromptTemplate


api_key = os.environ['tau_api_key']
os.environ["OPENAI_API_KEY"] = api_key


template = """
I want you to act as a document that I am having a conversation with. Your name is "AI Assistant" from Vegetable NZ.
You will provide me with answers from the given info. If the answer is not included, say exactly 
"Unfortunately, I do not know the answer to your question." and stop after that.
Refuse to answer any question not about the info. Never break character.

User: What is the capital of France?
AI Assistant: The capital of France is Paris.

User: Who is the author of 'Pride and Prejudice'?
AI Assistant: The author of 'Pride and Prejudice' is Jane Austen.

User: {query}
AI Assistant: """

prompt_template = PromptTemplate(input_variables=["query"], template=template)


def chat(indexfile, chat_history, user_input):
    
    max_input_size = 4096
    num_outputs = 512
    max_chunk_overlap = 20
    chunk_size_limit = 600
    
    prompt_helper = PromptHelper( max_input_size, num_outputs, max_chunk_overlap, chunk_size_limit=chunk_size_limit)
    llm_predictor = LLMPredictor(llm=ChatOpenAI(temperature=0.0, model_name="gpt-3.5-turbo", max_tokens=num_outputs))
    service_context = ServiceContext.from_defaults(llm_predictor=llm_predictor, prompt_helper=prompt_helper)
    
    index_filename = "index/"+ indexfile + ".json"

    index = GPTKnowledgeGraphIndex.load_from_disk(index_filename, service_context=service_context)
    
    bot_response = index.query(prompt_template.format(query=user_input), response_mode="compact")  
   
    response = ""
    
    for letter in ''.join(bot_response.response):
        response += letter + ""
    yield chat_history + [(user_input, response)]
      

index_files = ["Crop Protection", "Environmental Guidance", "Good Management Practice Guides"]

with gr.Blocks() as demo:
    gr.Markdown('Vegetable Expert Advisor')
        
    with gr.Tab("Ask away"):
        indexfile = gr.Radio(choices=list(index_files))
        chatbot = gr.Chatbot()
        message = gr.Textbox ()
        message.submit(chat, [indexfile, chatbot, message], chatbot)

demo.queue().launch(debug=True)