initial commit with chatbot

.gitignore

@@ -0,0 +1 @@
+.env

app.py

@@ -1,4 +1,292 @@
 import streamlit as st
+import os
+import PyPDF2
+from io import BytesIO
+from openai import OpenAI
+from huggingface_hub import InferenceClient
+from dotenv import load_dotenv
 
-x = st.slider('Select a value')
-st.write(x, 'squared is', x * x)
+load_dotenv()
+
+# ---------------------
+# Utility Functions
+# ---------------------
+
+
+def authenticate():
+    """
+    A simple authentication mechanism using a password stored in an environment variable (APP_PASSWORD).
+
+    Returns True if the user is authenticated, otherwise stops the Streamlit execution.
+    """
+    app_password = os.getenv("APP_PASSWORD", None)
+    if not app_password:
+        st.warning("No password set for the app. Please set the 'APP_PASSWORD' environment variable.")
+        return True  # Or return False if you want to block access
+
+    if "authenticated" not in st.session_state:
+        st.session_state["authenticated"] = False
+
+    if not st.session_state["authenticated"]:
+        st.text_input("Enter your access code:", type="password", key="login_password")
+        if st.button("Login"):
+            if st.session_state["login_password"] == app_password:
+                st.session_state["authenticated"] = True
+                st.experimental_rerun()
+            else:
+                st.error("Invalid password. Please try again.")
+                st.stop()
+
+    return st.session_state["authenticated"]
+
+
+def read_pdf(file):
+    """
+    Reads a PDF file using PyPDF2 and returns the extracted text.
+    """
+    pdf_reader = PyPDF2.PdfReader(file)
+    text = []
+    for page_num in range(len(pdf_reader.pages)):
+        page = pdf_reader.pages[page_num]
+        text.append(page.extract_text())
+    return "\n".join(text)
+
+def call_gpt_4o_api(
+    messages,
+    model,
+    temperature,
+    max_tokens,
+    stream
+):
+    """
+    Calls GPT-4o-compatible API (via OpenAI-like client).
+    Expects a list of messages (with "role" and "content" keys),
+    including the system message(s) as the first item(s) and
+    user/assistant messages subsequently.
+
+    Yields partial (streaming) or complete text.
+    """
+    client = OpenAI(api_key=os.getenv("OPENAI_API_KEY", ""))
+
+    # remove the second element from messages 
+    # (likely the "additional PDF context" system message, or your second system message)
+    messages = [messages[0]] + messages[2:]
+
+    if stream:
+        response = client.chat.completions.create(
+            model=model,
+            messages=messages,
+            temperature=temperature,
+            max_tokens=max_tokens,
+            stream=True
+        )
+        partial_text = ""
+        for chunk in response:
+            delta = chunk.choices[0].delta
+            if hasattr(delta, "content") and delta.content:
+                partial_text += delta.content
+                yield partial_text
+    else:
+        response = client.chat.completions.create(
+            model=model,
+            messages=messages,
+            temperature=temperature,
+            max_tokens=max_tokens,
+            stream=False
+        )
+        complete_text = response.choices[0].message.content
+        yield complete_text
+
+def call_hf_inference(
+    messages,
+    model_repo,
+    temperature=0.7,
+    max_tokens=200,
+    stream=False
+):
+    """
+    Calls a Hugging Face open-source LLM via the InferenceClient's chat endpoint.
+    Expects a list of messages (with "role" and "content"), including
+    system and user/assistant roles.
+
+    Yields partial (streaming) or complete text.
+    """
+    HF_TOKEN = os.getenv("HF_TOKEN", None)
+    if not HF_TOKEN:
+        raise ValueError("Please set your HF_TOKEN environment variable.")
+
+    client = InferenceClient(api_key=HF_TOKEN)
+
+    # remove the second element from messages
+    messages = [messages[0]] + messages[2:]
+
+    response = client.chat.completions.create(
+        model=model_repo,
+        messages=messages,
+        max_tokens=max_tokens,
+        temperature=temperature,
+        stream=stream
+    )
+
+    if stream:
+        partial_text = ""
+        for chunk in response:
+            delta = chunk.choices[0].delta
+            if isinstance(delta, dict):
+                chunk_content = delta.get("content", "")
+                partial_text += chunk_content
+                yield partial_text
+    else:
+        complete_text = response.choices[0].message["content"]
+        yield complete_text
+
+# ---------------------
+# Streamlit App
+# ---------------------
+
+def main():
+    if not authenticate():
+        st.stop()  # or just `return` to end early
+        
+    st.set_page_config(page_title="CVI-GPT", layout="centered")
+    st.title("CVI-GPT: Conversational Interface")
+
+    # ---------------------
+    # Sidebar: Model & Params
+    # ---------------------
+    st.sidebar.header("Model & Parameters")
+    
+    # Model selection
+    model_choice = st.sidebar.selectbox(
+        "Select Model",
+        [
+            "gpt-4o",
+            "gpt-4o-mini",
+            "meta-llama/Llama-3.3-70B-Instruct",
+            "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"
+        ]
+    )
+
+    # Temperature & max_tokens
+    temperature = st.sidebar.slider("Temperature", 0.0, 1.5, 0.7, 0.1)
+    max_tokens = st.sidebar.slider("Max Tokens", 50, 2000, 500, 50)
+
+    # We store the selected model in session_state to detect changes
+    if "selected_model" not in st.session_state:
+        st.session_state.selected_model = model_choice
+
+    # If the user changes the model, reset the conversation
+    if model_choice != st.session_state.selected_model:
+        st.session_state.selected_model = model_choice
+        st.session_state["messages"] = [
+            {"role": "assistant", "content": f"Model changed to `{model_choice}`. How can I help you?"}
+        ]
+
+    # System / Instruction Message
+    base_instructions = st.sidebar.text_area(
+        "System / Instruction Message",
+        value=(
+            "You are a Helpful Assistant. Respond in a concise, helpful, and markdown-friendly format.\n\n"
+            "Formatting Instructions:\n"
+            "- Responses should be in markdown.\n"
+            "- Use headings, bullet points, bold, italics, etc. for clarity.\n"
+            "- Use triple backticks for code blocks.\n"
+            "- Provide references or disclaimers when needed."
+        ),
+        height=200
+    )
+
+    # Clear Chat Button
+    if st.sidebar.button("Clear Chat"):
+        st.session_state["messages"] = [
+            {"role": "assistant", "content": "Chat cleared. How can I help you now?"}
+        ]
+
+    # ---------------------
+    # PDF Upload
+    # ---------------------
+    st.sidebar.header("Optional: PDF Upload")
+    uploaded_file = st.sidebar.file_uploader("Upload a PDF", type=["pdf"])
+    pdf_text = ""
+    if uploaded_file is not None:
+        pdf_text = read_pdf(uploaded_file)
+        # We do NOT print the PDF content. Just let user know it's loaded.
+        st.sidebar.write("PDF content loaded (not displayed).")
+        
+    st.sidebar.divider()
+    with st.sidebar:
+        st.subheader("👨‍💻 Author: *Adrish Maity*", anchor=False)
+
+    # ---------------------
+    # Initialize conversation if not present
+    # ---------------------
+    if "messages" not in st.session_state:
+        st.session_state["messages"] = [
+            {"role": "assistant", "content": "Hello! How can I help you today?"}
+        ]
+
+    # ---------------------
+    # Display Conversation
+    # ---------------------
+    for msg in st.session_state["messages"]:
+        with st.chat_message(msg["role"]):
+            st.markdown(msg["content"])
+
+    # ---------------------
+    # Chat Input
+    # ---------------------
+    if user_input := st.chat_input("Type your question..."):
+        # Just store the user's typed text
+        user_text = user_input
+
+        st.session_state["messages"].append({"role": "user", "content": user_text})
+
+        # Display user's message
+        with st.chat_message("user"):
+            st.markdown(user_text)
+
+        # Now build the full conversation:
+        # 1) A system message (instructions)
+        # 2) A second system message with PDF context if present (kept hidden from UI)
+        # 3) All prior conversation
+        full_conversation = [{"role": "system", "content": base_instructions}]
+
+        if pdf_text:
+            full_conversation[0]["content"] += "\n\n" + "Additional PDF context (user provided):\n" + pdf_text
+
+        full_conversation.extend(st.session_state["messages"])
+
+        # Placeholder for assistant's streaming response
+        with st.chat_message("assistant"):
+            response_placeholder = st.empty()
+            streamed_text = ""
+
+            # Decide how to call the model
+            if model_choice in ["gpt-4o", "gpt-4o-mini"]:
+                stream_response = call_gpt_4o_api(
+                    messages=full_conversation,
+                    model=model_choice,
+                    temperature=temperature,
+                    max_tokens=max_tokens,
+                    stream=True
+                )
+                for partial_output in stream_response:
+                    streamed_text = partial_output
+                    response_placeholder.markdown(streamed_text)
+            else:
+                hf_stream = call_hf_inference(
+                    messages=full_conversation,
+                    model_repo=model_choice,
+                    temperature=temperature,
+                    max_tokens=max_tokens,
+                    stream=True
+                )
+                for partial_output in hf_stream:
+                    streamed_text = partial_output
+                    response_placeholder.markdown(streamed_text)
+
+            # Once done, store the final assistant message
+            st.session_state["messages"].append({"role": "assistant", "content": streamed_text})
+
+if __name__ == "__main__":
+    main()

requirements.txt

@@ -0,0 +1,5 @@
+streamlit==1.38.0
+PyPDF2==3.0.1
+requests==2.31.0
+huggingface_hub==0.24.5
+openai==1.60.0