app.py

import os
import gradio as gr

from huggingface_hub import InferenceClient

from scraper import scrape

from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS


# =====================================================
# 0. Config
# =====================================================

HF_TOKEN = os.environ.get("HF_API_KEY")

MODEL_NAME = "HuggingFaceH4/zephyr-7b-beta"  # Much better for RAG


# =====================================================
# 1. Load + Build Knowledge Base
# =====================================================

print("🔄 Scraping website...")

raw_docs = scrape()

texts = []
metas = []

for d in raw_docs:
    texts.append(d["text"])
    metas.append({"source": d["source"]})


print("✂️ Splitting documents...")

splitter = RecursiveCharacterTextSplitter(
    chunk_size=800,
    chunk_overlap=150,
)

documents = splitter.create_documents(texts, metas)


print("🧠 Building embeddings...")

embeddings = HuggingFaceEmbeddings(
    model_name="sentence-transformers/all-mpnet-base-v2"
)


print("📦 Building vector store...")

db = FAISS.from_documents(documents, embeddings)

retriever = db.as_retriever(search_kwargs={"k": 4})


print("✅ Knowledge base ready!")


# =====================================================
# 2. Prompt Builder
# =====================================================

def build_prompt(question, docs):

    context = "\n\n".join(
        [
            f"[Source: {d.metadata['source']}]\n{d.page_content}"
            for d in docs
        ]
    )

    prompt = f"""
You are an academic assistant for SPJIMR.

Answer ONLY using the context below.
If information is missing, say "I don't know."

---------------------
CONTEXT:
{context}
---------------------

QUESTION:
{question}

ANSWER:
"""

    return prompt.strip()


# =====================================================
# 3. LLM Client
# =====================================================

client = InferenceClient(
    model=MODEL_NAME,
    token=HF_TOKEN
)


# =====================================================
# 4. Chat Function (Fixed Retriever API)
# =====================================================

def chat(message, history):

    # New LangChain API
    docs = retriever.invoke(message)

    prompt = build_prompt(message, docs)

    messages = [
        {"role": "user", "content": prompt}
    ]

    response = ""

    for chunk in client.chat_completion(
        messages=messages,
        max_tokens=700,
        temperature=0.3,
        stream=True,
    ):

        if chunk.choices[0].delta.content:
            token = chunk.choices[0].delta.content
            response += token
            yield response


# =====================================================
# 5. Minimal Dark UI
# =====================================================

custom_css = """
body {
    background: #0f172a !important;
}

.gradio-container {
    max-width: 900px !important;
    margin: auto !important;
}

h1 {
    color: #e5e7eb;
    text-align: center;
}

.subtitle {
    text-align: center;
    color: #9ca3af;
    margin-bottom: 20px;
}

footer {
    display: none !important;
}
"""


# =====================================================
# 6. App
# =====================================================

with gr.Blocks(
    css=custom_css,
    theme=gr.themes.Base(
        primary_hue="indigo",
        neutral_hue="slate",
    ),
) as demo:

    gr.Markdown(
        """
        # 🎓 SPJIMR AI Assistant
        <div class="subtitle">
        Ask questions based on official SPJIMR website
        </div>
        """,
        elem_id="title"
    )

    chatbot = gr.Chatbot(
        height=520,
        bubble_full_width=False,
    )

    msg = gr.Textbox(
        placeholder="Ask about programs, admissions, faculty...",
        show_label=False,
    )

    clear = gr.Button("Clear Chat")

    def user(user_message, history):
        return "", history + [[user_message, None]]

    def bot(history):
        user_message = history[-1][0]

        history[-1][1] = ""

        for chunk in chat(user_message, history):
            history[-1][1] = chunk
            yield history


    msg.submit(
        user,
        [msg, chatbot],
        [msg, chatbot],
        queue=False,
    ).then(
        bot,
        chatbot,
        chatbot,
    )

    clear.click(lambda: [], None, chatbot)


# =====================================================
# 7. Launch
# =====================================================

if __name__ == "__main__":
    demo.launch()