Create app.py

app.py

@@ -0,0 +1,66 @@
+import gradio as gr
+from langchain.vectorstores import FAISS
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.chains import RetrievalQA
+from langchain.llms import HuggingFacePipeline
+from transformers import pipeline
+
+# ---------------- LOAD EMBEDDINGS ----------------
+embeddings = HuggingFaceEmbeddings(
+    model_name="sentence-transformers/all-MiniLM-L6-v2"
+)
+
+# ---------------- LOAD VECTOR STORE ----------------
+db = FAISS.load_local(
+    "vectorstore/faiss_index",
+    embeddings,
+    allow_dangerous_deserialization=True
+)
+
+# ---------------- LOAD LLM ----------------
+generator = pipeline(
+    "text-generation",
+    model="mistralai/Mistral-7B-Instruct-v0.2",
+    max_new_tokens=512,
+    temperature=0.2,
+)
+
+llm = HuggingFacePipeline(pipeline=generator)
+
+# ---------------- RAG CHAIN ----------------
+qa = RetrievalQA.from_chain_type(
+    llm=llm,
+    retriever=db.as_retriever(search_kwargs={"k": 3}),
+    chain_type="stuff",
+)
+
+# ---------------- CHAT FUNCTION ----------------
+def chat(query, history):
+    if not query.strip():
+        return history
+
+    answer = qa.run(query)
+
+    history.append((query, answer))
+    return history
+
+# ---------------- GRADIO UI ----------------
+with gr.Blocks(title="RAG Document Chatbot") as demo:
+    gr.Markdown(
+        """
+        # 📚 RAG Document Chatbot
+        Answers are **strictly based on the provided documents**.
+        """
+    )
+
+    chatbot = gr.Chatbot(height=400)
+    query_box = gr.Textbox(
+        placeholder="Ask a question from the documents...",
+        label="Your Question",
+    )
+    clear = gr.Button("Clear Chat")
+
+    query_box.submit(chat, [query_box, chatbot], chatbot)
+    clear.click(lambda: [], None, chatbot)
+
+demo.launch()