Upload 3 files

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 my_info.pdf filter=lfs diff=lfs merge=lfs -text
+Document2.pdf filter=lfs diff=lfs merge=lfs -text

Document2.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98d708c94e4103d6f5f924f4c1177388eea43b6890968f65a2b0f51ac0e3da35
+size 163499

app.py

@@ -0,0 +1,253 @@
+
+import os
+import json
+from pypdf import PdfReader
+from openai import OpenAI
+import gradio as gr
+
+# --- 1. PDF Data Processing Functions ---
+
+def extract_text_from_pdf(pdf_path):
+    """
+    Extracts text content from a PDF file.
+
+    Args:
+        pdf_path (str): The path to the PDF document.
+
+    Returns:
+        str: The concatenated text content from all pages of the PDF.
+             Returns an empty string if the file is not found or an error occurs.
+    """
+    if not os.path.exists(pdf_path):
+        print(f"Error: PDF file not found at '{pdf_path}'")
+        return ""
+
+    try:
+        reader = PdfReader(pdf_path)
+        text_content = []
+        for page in reader.pages:
+            text_content.append(page.extract_text())
+        return "\n".join(text_content)
+    except Exception as e:
+        print(f"An error occurred while reading the PDF: {e}")
+        return ""
+
+def chunk_text(text, chunk_size=1000, chunk_overlap=200):
+    """
+    Splits a given text into smaller, overlapping chunks.
+
+    Args:
+        text (str): The input text to be chunked.
+        chunk_size (int): The desired size of each chunk.
+        chunk_overlap (int): The number of characters to overlap between consecutive chunks.
+
+    Returns:
+        list: A list of text chunks.
+    """
+    chunks = []
+    if not text:
+        return chunks
+
+    start_index = 0
+    while start_index < len(text):
+        end_index = min(start_index + chunk_size, len(text))
+        chunks.append(text[start_index:end_index])
+        if end_index == len(text):
+            break
+        start_index += chunk_size - chunk_overlap
+    return chunks
+
+# --- 2. OpenRouter API Client Setup ---
+
+# For local testing, get the API key from the environment or a hardcoded value if not found.
+# For Hugging Face deployment, ensure it's set as a Space Secret.
+OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY", "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") # Placeholder from earlier cell
+
+# If no key is found (e.g., if the placeholder was removed and env var isn't set)
+if not OPENROUTER_API_KEY:
+    raise ValueError("OPENROUTER_API_KEY not found. Please set it as an environment variable or provide it in app.py for local testing.")
+
+client = OpenAI(
+    base_url="https://openrouter.ai/api/v1",
+    api_key=OPENROUTER_API_KEY,
+)
+
+OPENROUTER_MODEL_NAME = "google/gemma-2-9b-it"
+
+# --- 3. Response Generation Agent ---
+
+def generate_response(user_query, context_chunks):
+    """
+    Generates a response from the LLM based on a user query and provided context chunks.
+
+    Args:
+        user_query (str): The user's question.
+        context_chunks (list): A list of strings, each being a chunk of text from the PDF.
+
+    Returns:
+        str: The LLM's generated response, or an error message if something goes wrong.
+    """
+    context = "\n---\n".join(context_chunks)
+
+    system_message_content = (
+        "You are a personal avatar chatbot. Your task is to answer the user's questions "
+        "based *only* on the provided context. If the answer cannot be found in the context, "
+        "state that you don't have enough information to answer. Do not make up information."
+        f"\n\nContext:\n{context}"
+    )
+
+    messages = [
+        {"role": "system", "content": system_message_content},
+        {"role": "user", "content": user_query}
+    ]
+
+    try:
+        response = client.chat.completions.create(
+            model=OPENROUTER_MODEL_NAME,
+            messages=messages,
+            temperature=0.5,
+            max_tokens=500
+        )
+        return response.choices[0].message.content
+    except Exception as e:
+        return f"An error occurred while generating response: {e}"
+
+# --- 4. Response Evaluation Agent ---
+
+def evaluate_response(user_query, generated_response, context_chunks):
+    """
+    Evaluates a generated response based on the user query and context,
+    using an OpenRouter-powered LLM.
+
+    Args:
+        user_query (str): The original user's question.
+        generated_response (str): The response generated by the first agent.
+        context_chunks (list): The list of text chunks from the PDF used as context.
+
+    Returns:
+        tuple: A tuple containing (evaluation_pass_fail (bool), reasoning (str)).
+               Returns (False, error_message) if an error occurs.
+    """
+    context = "\n---\n".join(context_chunks)
+
+    evaluation_system_message = (
+        "You are an evaluation agent. Your task is to assess a 'generated_response' "
+        "based on a 'user_query' and provided 'context'.\n\n"
+        "Your assessment should focus on the following criteria:\n"
+        "1. **Accuracy**: Is the 'generated_response' factually correct according to the 'context'?\n"
+        "2. **Relevance**: Does the 'generated_response' directly address the 'user_query'?\n"
+        "3. **Context Adherence**: Does the 'generated_response' *only* use information present "
+        "in the 'context'? If it brings in outside information or makes up facts, it fails this criterion.\n\n"
+        "Based on these criteria, determine if the 'generated_response' is acceptable. "
+        "If the response explicitly states it cannot answer based on the context, and it's true, consider it acceptable."
+        "Return your evaluation as a JSON object with two keys: 'pass' (boolean: true if acceptable, false otherwise) "
+        "and 'reasoning' (string: a brief explanation for your decision).\n\n"
+        f"Context:\n{context}"
+    )
+
+    evaluation_user_message = (
+        f"User Query: {user_query}\n\n"
+        f"Generated Response: {generated_response}\n\n"
+        "Please evaluate this generated response according to the instructions." 
+    )
+
+    messages = [
+        {"role": "system", "content": evaluation_system_message},
+        {"role": "user", "content": evaluation_user_message}
+    ]
+
+    try:
+        response = client.chat.completions.create(
+            model=OPENROUTER_MODEL_NAME,
+            messages=messages,
+            temperature=0.1,
+            max_tokens=300,
+            response_format={ "type": "json_object" }
+        )
+
+        evaluation_output = response.choices[0].message.content
+
+        try:
+            eval_result = json.loads(evaluation_output)
+            return eval_result.get('pass', False), eval_result.get('reasoning', 'No reasoning provided.')
+        except json.JSONDecodeError:
+            print(f"Warning: Could not decode JSON from evaluator: {evaluation_output}")
+            return False, f"Evaluator returned malformed JSON: {evaluation_output}"
+
+    except Exception as e:
+        return False, f"An error occurred while evaluating response: {e}"
+
+# --- 5. Chatbot Orchestration Logic ---
+
+def chat_with_avatar(user_query, context_chunks, max_retries=3):
+    """
+    Orchestrates the interaction between the generator and evaluator agents.
+    Attempts to generate a response and evaluates it, retrying if necessary.
+
+    Args:
+        user_query (str): The user's question.
+        context_chunks (list): A list of strings, each being a chunk of text from the PDF.
+        max_retries (int): The maximum number of attempts to generate an acceptable response.
+
+    Returns:
+        str: The final, approved response or a message indicating failure.
+    """
+    for attempt in range(max_retries):
+        # print(f"\n--- Attempt {attempt + 1}/{max_retries} ---") # Commented for Gradio output
+
+        generated_response = generate_response(user_query, context_chunks)
+        # print(f"Generated Response (Attempt {attempt + 1}):\n{generated_response[:200]}...")
+
+        pass_fail, reasoning = evaluate_response(user_query, generated_response, context_chunks)
+        # print(f"Evaluation Result (Attempt {attempt + 1}): Pass = {pass_fail}, Reasoning: {reasoning}")
+
+        if pass_fail:
+            return generated_response
+        else:
+            # print(f"Response failed evaluation. Retrying... Reason: {reasoning}")
+            pass # Just retry with the same prompt for now
+
+    return "I'm sorry, I couldn't generate an acceptable response after several attempts. Please try rephrasing your question."
+
+# --- Initial Setup (Load PDF and Chunk) ---
+
+# Path to your personal PDF document. 
+# For Hugging Face Spaces, upload your PDF to the 'data' folder in your Space repository.
+PDF_PATH = "Document2.pdf" # Make sure this matches the filename you upload
+
+# Create a dummy PDF file for testing if it doesn't exist, this part is mainly for local dev/testing
+# In Hugging Face spaces, the PDF should be present.
+if not os.path.exists(PDF_PATH):
+    # This block would only execute if the PDF is missing locally for testing.
+    # In a deployed HF Space, the PDF should be present.
+    print(f"Warning: PDF file not found at '{PDF_PATH}'. Using empty content.")
+    PDF_CONTENT = ""
+else:
+    print(f"Extracting text from {PDF_PATH}...")
+    PDF_CONTENT = extract_text_from_pdf(PDF_PATH)
+    print("PDF content extracted.")
+
+TEXT_CHUNKS = chunk_text(PDF_CONTENT)
+
+if not TEXT_CHUNKS:
+    print("Warning: No text chunks were created from the PDF. The chatbot will not have context.")
+
+# --- Gradio Interface ---
+
+def respond(message, history):
+    global TEXT_CHUNKS # Use the globally loaded text chunks
+    if not TEXT_CHUNKS:
+        return "I am unable to answer questions as my knowledge base (PDF) could not be loaded or processed. Please check the PDF file."
+    return chat_with_avatar(message, TEXT_CHUNKS)
+
+gr.ChatInterface(
+    respond,
+    title="Personal Avatar Chatbot",
+    description="Ask me anything about Pavan Thakkallapalli! (Information based on provided PDF)",
+    examples=[
+        "What is Pavan Thakkallapalli's primary role and education?",
+        "Tell me about Pavan's experience with MLOps and Machine Learning.",
+        "What is Pavan's favorite movie?"
+    ],
+    theme="soft"
+).launch(debug=True)

requirements.txt

@@ -0,0 +1,3 @@
+pypdf==6.6.2
+openai==2.15.0
+gradio==4.20.0