app.py

import gradio as gr
import requests
from langchain.embeddings import SentenceTransformerEmbeddings
from langchain.vectorstores import FAISS
from langchain_community.chat_models.huggingface import ChatHuggingFace
from langchain.schema import SystemMessage, HumanMessage, AIMessage
from langchain_community.llms import HuggingFaceEndpoint

model_name = "sentence-transformers/all-mpnet-base-v2"
embedding_llm = SentenceTransformerEmbeddings(model_name=model_name)

db = FAISS.load_local("faiss_index", embedding_llm, allow_dangerous_deserialization=True)

# Set up Hugging Face model
llm = HuggingFaceEndpoint(
    repo_id="HuggingFaceH4/starchat2-15b-v0.1",
    task="text-generation",
    max_new_tokens=4096,
    temperature=0.6,
    top_p=0.9,
    top_k=40,
    repetition_penalty=1.2,
    do_sample=True,
)
chat_model = ChatHuggingFace(llm=llm)

messages = [
    SystemMessage(content="You are a helpful assistant."),
    HumanMessage(content="Hi AI, how are you today?"),
    AIMessage(content="I'm great thank you. How can I help you?")
]

def handle_message(message: str, mode: str):
    result_text, result_image = "", None
    
    if not message.strip():
        return "Enter a valid message.", None
    
    if mode == "Chat-Message":
        result_text = chat_message(message)
    elif mode == "Web-Search":
        result_text = web_search(message)
    elif mode == "Chart-Generator":
        result_text, result_image = chart_generator(message)
    else:
        result_text = "Select a valid mode."
    
    return result_text, result_image

def chat_message(message: str):
    global messages
    
    prompt = HumanMessage(content=message)
    messages.append(prompt)
    
    response = chat_model.invoke(messages)
    messages.append(response.content)
    
    if len(messages) >= 6:
        messages = messages[-6:]
        
    return f"IT-Assistant: {response.content}"

def web_search(message: str):
    global messages
    
    similar_docs = db.similarity_search(message, k=3)
    
    if similar_docs:
        source_knowledge = "\n".join([x.page_content for x in similar_docs])
    else:
        source_knowledge = ""
        
    augmented_prompt = f"""
    If the answer to the next query is not contained in the Web Search, say 'No Answer Is Available' and then just give guidance for the query.

    Query: {message}

    Web Search:
    {source_knowledge}
    """
    
    prompt = HumanMessage(content=augmented_prompt)
    messages.append(prompt)
    
    response = chat_model.invoke(messages)
    messages.append(response.content)
    
    if len(messages) >= 6:
        messages = messages[-6:]
        
    return f"IT-Assistant: {response.content}"

def chart_generator(message: str):
    global messages

    chart_url = f"https://quickchart.io/natural/{message}"
    response = requests.get(chart_url)

    if response.status_code == 200:
        message_with_description = f"Describe and analyse the content of this chart: {message}"

        prompt = HumanMessage(content=message_with_description)
        messages.append(prompt)

        response = chat_model.invoke(messages)
        messages.append(response.content)
        
        if len(messages) >= 6:
            messages = messages[-6:]

        return f"IT-Assistant: {response.content}", chart_url
    else:
        return f"Can't generate this image. Please provide valid chart details.", None
        
demo = gr.Interface(
    fn=handle_message,
    inputs=["text", gr.Radio(["Chat-Message", "Web-Search", "Chart-Generator"], label="mode", info="Choose a mode and enter your message, then click submit to interact.")],
    outputs=[gr.Textbox(label="Response"), gr.Image(label="Chart", type="auto")],
    title="IT Assistant")

demo.launch()