app.py

import os

os.environ["ANTHROPIC_API_KEY"] = "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"
import gradio as gr
from langchain.memory import ConversationBufferMemory
from langchain.chains import RetrievalQA
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import Chroma
from langchain.chat_models import ChatAnthropic

model_kwargs = {'trust_remote_code': True}
embeddings = HuggingFaceEmbeddings(model_name="jinaai/jina-embeddings-v2-base-en",
                                   model_kwargs=model_kwargs)
llm = ChatAnthropic(model='claude-2',
                    temperature=0)

# Load the persisted Chroma database
persist_directory='/Electric_Machinerydb'
embeddings = embeddings
chroma_db = Chroma(persist_directory=persist_directory, embedding_function=embeddings)
retriever = chroma_db.as_retriever(search_kwargs={"k": 3}, return_source_documents=True)

# Initialize ConversationBufferMemory
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True, input_key="query", output_key="result")

# Create a question-answering chain using the retriever
qa_chain = RetrievalQA.from_chain_type(
    llm=llm,
    chain_type="stuff",
    retriever=retriever,
    return_source_documents=True,
    memory=memory,
    output_key="result"  # Specify the output key
)

# Gradio app
def ask_question(question, chat_history):
    query = question.strip()
    if query:
        result = qa_chain({"query": query})
        answer = result['result']
        chat_history.append((query, answer))
        metadata = show_metadata(chat_history)
        return "", chat_history, metadata
    else:
        chat_history.append(("", "Please enter a question."))
        return "", chat_history, ""

def show_metadata(chat_history):
    if chat_history:
        query, answer = chat_history[-1]
        result = qa_chain({"query": query})
        metadata = ""
        for doc in result['source_documents']:
            metadata += f"Page: {doc.metadata['page']}\n"
            metadata += f"Source: {doc.metadata['source']}\n"
            metadata += f"Content: {doc.page_content}\n"
            metadata += "---\n"
        return metadata
    return ""

def clean_history():
    memory.clear()
    return [], "", ""

with gr.Blocks() as demo:
    gr.Markdown("# Electric Machinery QA by Tamil")

    with gr.Row():
        with gr.Column(scale=2):
            question = gr.Textbox(label="Question")
            ask_btn = gr.Button("Ask")
            clean_btn = gr.Button("Clean")
            chatbot = gr.Chatbot(label="Conversation")

        with gr.Column(scale=1):
            gr.Markdown("## Metadata")
            metadata_output = gr.Textbox(label="Source Information", lines=10)

    ask_btn.click(ask_question, inputs=[question, chatbot], outputs=[question, chatbot, metadata_output])
    clean_btn.click(clean_history, inputs=[], outputs=[chatbot, question, metadata_output])

demo.launch()