app.py

import gradio as gr
from langchain_text_splitters import CharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from transformers import pipeline

# -------------------------
# Load Data
# -------------------------
with open("data.txt", "r") as f:
    raw_text = f.read()

# -------------------------
# Split Text
# -------------------------
text_splitter = CharacterTextSplitter(
    chunk_size=300,
    chunk_overlap=50
)
texts = text_splitter.split_text(raw_text)

# -------------------------
# Embeddings + Vector Store
# -------------------------
embeddings = HuggingFaceEmbeddings(
    model_name="sentence-transformers/all-MiniLM-L6-v2"
)

vectorstore = FAISS.from_texts(texts, embeddings)

# -------------------------
# LLM (Stable HF Model)
# -------------------------
generator = pipeline(
    "text-generation",
    model="microsoft/phi-2",
    max_new_tokens=120,
    temperature=0.2,
    do_sample=False
)

# -------------------------
# Chat Function (RAG + Memory)
# -------------------------
def rag_chat(message, history):

    # Retrieve relevant docs
    docs = vectorstore.similarity_search(message, k=3)
    context = " ".join([doc.page_content for doc in docs])

    # Build chat memory (last 3 messages)
    history_text = ""
    for user, bot in history[-3:]:
        history_text += f"User: {user}\nBot: {bot}\n"

    # Prompt
    prompt = f"""
You are a professional AI assistant.

Rules:
- Answer ONLY from the context
- If answer not found, say "I don't know"
- Keep answer short and clear

Context:
{context}

Conversation:
{history_text}

Question:
{message}

Answer:
"""

    # Generate response
    result = generator(prompt)[0]["generated_text"]

    # Clean output
    answer = result.replace(prompt, "").strip()
    answer = answer.split("\n")[0]

    return answer

# -------------------------
# UI
# -------------------------
demo = gr.ChatInterface(
    fn=rag_chat,
    title="RAG Chatbot .txt base data Trained"
)

# -------------------------
# Run App
# -------------------------
if __name__ == "__main__":
    demo.launch()