update

app.py

@@ -1,5 +1,20 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
+import os
+import torch
+import transformers
+from tensorflow import keras
+from transformers import  AutoTokenizer,  pipeline, AutoModelForSeq2SeqLM,AutoModelForCausalLM
+from langchain.document_loaders import PyPDFLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.vectorstores import Chroma
+from langchain.prompts import ChatPromptTemplate
+from langchain.schema.runnable import RunnablePassthrough
+from langchain.schema.output_parser import StrOutputParser
+from langchain.llms import HuggingFacePipeline
+import gradio as gr
+
 
 """
 For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
@@ -7,58 +22,141 @@ For more information on `huggingface_hub` Inference API support, please check th
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+print("GPU Available:", torch.cuda.is_available())
+print("GPU Name:", torch.cuda.get_device_name(0) if torch.cuda.is_available() else "No GPU Found")
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+pdf_files = ["Apple-10K-2023.pdf", "Apple-10K-2024.pdf"]
 
-    messages.append({"role": "user", "content": message})
+"""### 📌 Step 1: Load Multiple 10-K Financial Report PDFs """
 
-    response = ""
+all_documents = []
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
+def preprocess_text(text):
+    text = text.replace("\n", " ").strip()
+    text = ' '.join(text.split())  # Remove extra spaces
+    return text
 
+for pdf_path in pdf_files:
+    loader = PyPDFLoader(pdf_path)
+    documents = loader.load()
+    for doc in documents:
+        doc.page_content = preprocess_text(doc.page_content)
+    all_documents.extend(documents)
 
+"""### 📌 Step 2: Split Text into Chunks
+  <p> Here each split will also have a metadata defining the location of the chunk in the actual document for citation,also other details.As the pdf text is clean with no html tags etc , we use it as such with no cleaning
 """
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
+
+text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=20)
+all_splits = text_splitter.split_documents(all_documents)
+
+"""### 📌 Step 3: Create Embeddings using Sentence Transformers"""
+
+# Check if CUDA (GPU) is available; otherwise, use CPU
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model_name = "sentence-transformers/all-MiniLM-L6-v2"
+embeddings = HuggingFaceEmbeddings(model_name=model_name, model_kwargs={"device": device})
+
+"""### 📌 Step 4: Store & Retrieve using ChromaDB"""
+
+vectordb = Chroma.from_documents(documents=all_splits, embedding=embeddings, persist_directory="content/drive/MyDrive/RAG_DB/chroma_db")
+retriever = vectordb.as_retriever()
+
+# Choose a smaller T5 model
+model_name = "google/flan-t5-large"
+
+# Load model & tokenizer
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name, device_map="auto")
+
+# Create Hugging Face pipeline
+hf_pipeline = pipeline( "text2text-generation",
+    model=model,
+    tokenizer=tokenizer,
+    truncation=True)
+
+# Integrate with LangChain
+llm = HuggingFacePipeline(pipeline=hf_pipeline)
+
+"""### 📌 Step 6: Define RAG Prompt"""
+
+# Define RAG Prompt
+template = """You are an AI assistant answering financial questions using retrieved financial reports.
+Use the following retrieved context to answer the question concisely.
+Question: {question}
+Context: {context}
+Answer:
 """
+prompt = ChatPromptTemplate.from_template(template)
+
+"""### 📌 Step 7: Create RAG pipeline"""
+
+# Create RAG Pipeline
+conversation_chain = (
+    {"context": retriever, "question": RunnablePassthrough()}
+    | prompt
+    | llm
+    | StrOutputParser()
+)
+
+"""### 📌 Step 8: Create a function to get the confidence score"""
+
+# Function to Get Confidence Score
+def get_confidence_score(question):
+    retrieved_docs_with_scores = vectordb.similarity_search_with_score(question, k=5)
+    max_score = max([doc[1] for doc in retrieved_docs_with_scores]) if retrieved_docs_with_scores else 0
+    print(max_score)
+    return min(1.0, round(max_score, 2))  # Normalize to 0-1 scale
+
+
+"""### 📌 Step 10: Integrate with Gradio UI"""
+
+# Define Chatbot Function
+def chat_with_rag(message, history):
+    try:
+        response = conversation_chain.invoke(message)
+        confidence_score = get_confidence_score(message)
+        formatted_response = f"**Answer:** {response}\n\n**Confidence Score:** {confidence_score:.2f}"
+        return formatted_response
+    except Exception as e:
+        return f"Error: {str(e)}"
+
+# Create Gradio Chatbot UI with Auto-Clearing Input
 demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
+    fn=chat_with_rag,  # Function to generate responses
+    title="📊 Financial RAG Chatbot",
+    description="Ask questions about financial reports and get AI-powered answers!",
+    theme="soft",  # Aesthetic theme
+    examples=[
+        ["What was Apple's total revenue in 2024?"],
+        ["What are the biggest financial risks for Apple?"]
+        ["What are the biggest challenges for Apple?"],
+        ["What is the capital of France?"],
     ],
+    submit_btn="Ask",  # Customize the submit button text
+    stop_btn=None,  # Remove unnecessary stop button
 )
 
 
+"""
+For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
+"""
+# demo = gr.ChatInterface(
+#     respond,
+#     additional_inputs=[
+#         gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+#         gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+#         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+#         gr.Slider(
+#             minimum=0.1,
+#             maximum=1.0,
+#             value=0.95,
+#             step=0.05,
+#             label="Top-p (nucleus sampling)",
+#         ),
+#     ],
+# )
+
+
 if __name__ == "__main__":
     demo.launch()