Update app.py

app.py

@@ -1,134 +1,122 @@
-import streamlit as st
-import pandas as pd
-import pypdf
-import docx2txt
-from typing import List
-from transformers import pipeline
-from langchain_community.vectorstores import FAISS
-from langchain_community.embeddings import HuggingFaceEmbeddings
-from langchain_text_splitters import RecursiveCharacterTextSplitter
-from langchain_core.documents import Document
-import evaluate
-
-# ======================================
-# 🧩 Streamlit Title
-# ======================================
-st.set_page_config(page_title="RAG Chatbot with Grok 70B", page_icon="📚", layout="centered")
-st.title("📚 RAG Chatbot powered by Grok 70B")
-
-# ======================================
-# 🔹 Load Model (HF Hosted)
-# ======================================
-@st.cache_resource
-def load_model():
-    """
-    Load Grok 70B model via Hugging Face pipeline (hosted inference)
-    """
-    model_name = "GrokAI/llama-3.3-70b-versatile"
-    generator = pipeline(
-        "text-generation",
-        model=model_name,
-        device_map="auto",   # uses GPU if available, else CPU
-        trust_remote_code=True  # required for Grok models
+import gradio as gr
+import os
+from dotenv import load_dotenv
+from groq import Groq
+
+load_dotenv()
+
+# ── Agent Runner ─────────────────────────────────────────────────
+def run_agent(client, role, backstory, task):
+    response = client.chat.completions.create(
+        model="llama-3.3-70b-versatile",
+        messages=[
+            {"role": "system", "content": f"You are a {role}. {backstory}"},
+            {"role": "user",   "content": task}
+        ],
+        temperature=0.7,
+        max_tokens=1024,
     )
-    return generator
-
-with st.spinner("🔄 Loading Grok 70B model (hosted)..."):
-    generator = load_model()
-
-# ======================================
-# 📂 File Upload Section
-# ======================================
-uploaded_file = st.file_uploader("📤 Upload a file (PDF, DOCX, CSV)", type=["pdf", "docx", "csv"])
-
-def extract_text(file):
-    """Extract text from uploaded file"""
-    if file.type == "application/pdf":
-        reader = pypdf.PdfReader(file)
-        return "\n".join([page.extract_text() for page in reader.pages if page.extract_text()])
-    elif file.type == "application/vnd.openxmlformats-officedocument.wordprocessingml.document":
-        return docx2txt.process(file)
-    elif file.type == "text/csv":
-        df = pd.read_csv(file)
-        return df.to_string(index=False)
-    return ""
-
-# ======================================
-# 🧠 Build FAISS Vector Store
-# ======================================
-@st.cache_resource
-def build_faiss(_docs: List[Document]):
-    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
-    return FAISS.from_documents(_docs, embeddings)
-
-docs, db = [], None
-if uploaded_file:
-    text = extract_text(uploaded_file)
-    if text:
-        st.info("🔍 Splitting and indexing document...")
-        splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)  # increased chunk size
-        docs = [Document(page_content=chunk) for chunk in splitter.split_text(text)]
-        db = build_faiss(docs)
-        st.success("✅ Knowledge base indexed successfully!")
-
-# ======================================
-# 💬 Chat Interface
-# ======================================
-query = st.text_input("💬 Ask a question about your document:")
-
-if query and db:
-    retriever = db.as_retriever(search_kwargs={"k": 3})
-    retrieved_docs = retriever.invoke(query)  # get top 3 relevant docs
-    context = "\n".join([doc.page_content for doc in retrieved_docs])
-
-    prompt = f"""
-You are a factual assistant. Use ONLY the provided context to answer.
-If the context doesn’t contain enough info, say:
-"The document does not provide enough information."
-Context:
-{context}
-Question: {query}
-Answer:
-"""
-
-    with st.spinner("🤔 Generating answer..."):
-        result = generator(prompt, max_new_tokens=200, temperature=0.5, top_p=0.9)
-        generated_text = result[0]["generated_text"]
-        answer = generated_text.split("Answer:")[-1].strip()
-
-    st.subheader("📝 Answer")
-    st.write(answer)
-
-    # ======================================
-    # 🔎 Evaluation Section
-    # ======================================
-    st.divider()
-    st.subheader("📊 Evaluate Answer Quality")
-
-    reference = st.text_area("✅ Reference Answer (optional for evaluation):")
-
-    if st.button("Compute BLEU + ROUGE + BERTScore") and reference:
-        # BLEU
-        bleu = evaluate.load("bleu")
-        bleu_score = bleu.compute(predictions=[answer], references=[[reference]])["bleu"]
-
-        # ROUGE
-        rouge = evaluate.load("rouge")
-        rouge_result = rouge.compute(predictions=[answer], references=[reference])
-        rouge1_f = rouge_result["rouge1"]
-        rouge2_f = rouge_result["rouge2"]
-        rougel_f = rouge_result["rougeL"]
-
-        # BERTScore
-        bertscore = evaluate.load("bertscore")
-        bert = bertscore.compute(predictions=[answer], references=[reference], lang="en")
-        bert_f1 = sum(bert["f1"]) / len(bert["f1"])
-
-        st.markdown(f"""
-        **🧩 Evaluation Results:**
-        - **BLEU:** {bleu_score:.4f}
-        - **ROUGE-1 F1:** {rouge1_f:.4f}
-        - **ROUGE-2 F1:** {rouge2_f:.4f}
-        - **ROUGE-L F1:** {rougel_f:.4f}
-        - **BERTScore F1:** {bert_f1:.4f}
-        """)
+    return response.choices[0].message.content.strip()
+
+# ── Three Agents Sequentially ────────────────────────────────────
+def plan_trip(destination, trip_days, budget_range):
+    api_key = os.getenv("GROQ_API_KEY")
+    if not api_key:
+        return "❌ GROQ_API_KEY not set. Add it in Space Settings → Variables and secrets."
+    if not destination.strip():
+        return "⚠️ Please enter a destination."
+
+    client = Groq(api_key=api_key)
+
+    try:
+        # Agent 1 — Researcher
+        research = run_agent(
+            client,
+            role="Travel Researcher",
+            backstory="You have deep knowledge of global destinations, cultures, and hidden gems.",
+            task=(
+                f"Research {destination} and provide:\n"
+                f"- Top 5 tourist attractions (1-2 lines each)\n"
+                f"- 3 hotel recommendations across budget levels\n"
+                f"- 5 must-try local foods\n"
+                f"- Best time to visit and general travel tips\n"
+                f"Be concise and specific."
+            )
+        )
+
+        # Agent 2 — Planner (receives researcher output)
+        itinerary = run_agent(
+            client,
+            role="Travel Planner",
+            backstory="You craft realistic, enjoyable day-by-day itineraries tailored to traveler preferences.",
+            task=(
+                f"Based on this research about {destination}:\n{research}\n\n"
+                f"Create a {trip_days}-day itinerary with Morning, Afternoon, and Evening activities each day. "
+                f"Be specific with place names and timings."
+            )
+        )
+
+        # Agent 3 — Budget Analyst (receives itinerary output)
+        budget = run_agent(
+            client,
+            role="Travel Budget Analyst",
+            backstory="You give accurate, realistic travel cost breakdowns in USD.",
+            task=(
+                f"Based on this {trip_days}-day itinerary for {destination}:\n{itinerary}\n\n"
+                f"Provide a full budget for a {budget_range} traveler (1 person) covering:\n"
+                f"- Flights (round trip estimate)\n"
+                f"- Accommodation (per night × {trip_days} nights)\n"
+                f"- Food & Dining\n"
+                f"- Activities & Entrance Fees\n"
+                f"- Local Transport\n"
+                f"- Miscellaneous\n"
+                f"- TOTAL ESTIMATED COST\n"
+                f"Use USD. Be realistic for the {budget_range} level."
+            )
+        )
+
+        return (
+            f"# ✈️ {destination} — {trip_days}-Day Trip ({budget_range})\n"
+            f"{'='*60}\n\n"
+            f"## 🔍 Agent 1: Research\n{research}\n\n"
+            f"{'='*60}\n\n"
+            f"## 🗓️ Agent 2: Itinerary\n{itinerary}\n\n"
+            f"{'='*60}\n\n"
+            f"## 💰 Agent 3: Budget Breakdown\n{budget}\n"
+        )
+
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
+
+
+# ── Gradio UI ─────────────────────────────────────────────────────
+with gr.Blocks(theme=gr.themes.Soft(), title="Travel Planner AI") as demo:
+
+    gr.Markdown(
+        "# ✈️ Travel Planner AI\n"
+        "Three AI agents work sequentially: **Researcher → Planner → Budget Analyst** "
+        "(powered by Groq + LLaMA3-70b)"
+    )
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            destination  = gr.Textbox(label="Destination", placeholder="e.g. Paris, Tokyo, Istanbul")
+            trip_days    = gr.Slider(1, 14, value=5, step=1, label="Number of Days")
+            budget_range = gr.Radio(["Budget", "Mid-range", "Luxury"], value="Mid-range", label="Budget Style")
+            plan_btn     = gr.Button("🗺️ Plan My Trip", variant="primary")
+
+        with gr.Column(scale=2):
+            output = gr.Textbox(
+                label="Your AI-Generated Trip Plan",
+                lines=35,
+                placeholder="Your trip plan will appear here once the three agents complete their work..."
+            )
+
+    plan_btn.click(fn=plan_trip, inputs=[destination, trip_days, budget_range], outputs=output)
+
+    gr.Markdown(
+        "> 🔑 Powered by [Groq](https://groq.com) (free). "
+        "Add your `GROQ_API_KEY` in **Space Settings → Variables and secrets**."
+    )
+
+demo.launch(server_name="0.0.0.0", server_port=7860, show_error=True)