Update app.py

app.py

@@ -1,60 +1,61 @@
 import gradio as gr
-import faiss
-import torch
-from sentence_transformers import SentenceTransformer
-from transformers import AutoTokenizer, AutoModelForCausalLM, AutoModelForSeq2SeqLM
-
-# ---------- Load models ----------
-embed_model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
-
-gen_tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-2")
-gen_model = AutoModelForCausalLM.from_pretrained("microsoft/phi-2", torch_dtype=torch.float32)
-
-# Example: EN->HI
-trans_tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-hi")
-trans_model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-hi")
-
-# ---------- Sample docs + FAISS index ----------
-documents = [
-    "The Taj Mahal is an ivory-white marble mausoleum in India.",
-    "ChatGPT is a large language model developed by OpenAI.",
-    "RAG combines retrieval-based and generation-based approaches."
-]
-
-doc_embeddings = embed_model.encode(documents, convert_to_tensor=True)
-index = faiss.IndexFlatL2(doc_embeddings.shape[1])
-index.add(doc_embeddings.cpu().numpy())
-
-# ---------- RAG Function ----------
-def rag_translate(query, target_lang='hi'):
-    query_vec = embed_model.encode([query])
-    _, top_indices = index.search(query_vec, k=1)
-    retrieved_doc = documents[top_indices[0][0]]
-
-    prompt = f"Context: {retrieved_doc}\nQuestion: {query}\nAnswer:"
-    inputs = gen_tokenizer(prompt, return_tensors="pt")
-    outputs = gen_model.generate(**inputs, max_new_tokens=64)
-    answer_en = gen_tokenizer.decode(outputs[0], skip_special_tokens=True)
-
-    # Translate if requested
-    if target_lang and target_lang != 'en':
-        trans_inputs = trans_tokenizer(answer_en, return_tensors="pt", truncation=True)
-        trans_output = trans_model.generate(**trans_inputs)
-        translated = trans_tokenizer.decode(trans_output[0], skip_special_tokens=True)
-        return f"🔍 Answer:\n{answer_en}\n\n🌐 Translated:\n{translated}"
+import requests
+
+TOGETHER_API_KEY = "tgp_v1_ZytvDbMu9PMwIlnBZEfYSq9nzJAYwS0MecjY9Kt7RxE"
+SERPER_API_KEY = "75f06519187851ad63486c3012b34c5e0e6501f1"
+
+# Step 1: Use Serper to search the web
+def web_search(query):
+    url = "https://google.serper.dev/search"
+    headers = {"X-API-KEY": SERPER_API_KEY}
+    payload = {"q": query}
+    res = requests.post(url, json=payload, headers=headers)
+    results = res.json().get("organic", [])
+    if not results:
+        return "No relevant search results found."
+    context = "\n".join([f"{item['title']}: {item['link']}\n{item['snippet']}" for item in results[:3]])
+    return context
+
+# Step 2: Use Together API to get a response
+def call_together_llm(context, question):
+    url = "https://api.together.xyz/v1/chat/completions"
+    headers = {
+        "Authorization": f"Bearer {TOGETHER_API_KEY}",
+        "Content-Type": "application/json"
+    }
+    messages = [
+        {"role": "system", "content": "You are a helpful assistant that uses the web context to answer."},
+        {"role": "user", "content": f"Context: {context}\n\nQuestion: {question}"}
+    ]
+    data = {
+        "model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
+        "messages": messages,
+        "temperature": 0.7,
+        "max_tokens": 512
+    }
+    res = requests.post(url, headers=headers, json=data)
+    return res.json()['choices'][0]['message']['content']
+
+# Step 3: Combine Search + LLM for RAG
+def rag_chatbot(question):
+    try:
+        context = web_search(question)
+        response = call_together_llm(context, question)
+        return response
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
+
+# Step 4: UI with Gradio
+with gr.Blocks() as demo:
+    gr.Markdown("# 🤖 Free RAG Chatbot with Together + Serper")
+    gr.Markdown("Ask any question. It will search the web and give a smart answer.")
     
-    return f"🔍 Answer:\n{answer_en}"
-
-# ---------- Gradio UI ----------
-iface = gr.Interface(
-    fn=rag_translate,
-    inputs=[
-        gr.Textbox(label="Ask a Question"),
-        gr.Dropdown(choices=["en", "hi", "fr", "es"], value="hi", label="Target Language")
-    ],
-    outputs=gr.Textbox(label="Answer"),
-    title="🧠 RAG + 🌍 Translator",
-    description="A lightweight RAG system with answer translation. Powered by Phi-2 + MiniLM + Opus MT."
-)
-
-iface.launch()
+    with gr.Row():
+        input_box = gr.Textbox(placeholder="Ask a question...", label="Your Question")
+        output_box = gr.Textbox(label="Answer", lines=10)
+
+    submit_btn = gr.Button("🔍 Search & Answer")
+
+    submit_btn.click(fn=rag_chatbot, inputs=input_box, outputs=output_box)
+
+demo.launch()