Update app.py

app.py

@@ -1,117 +1,591 @@
+"""
+PDF → Summary → Audio → Talk to PDF → Diagram
+- Summarization: Groq (LLaMA 3)
+- TTS: Deepgram (aura-asteria-en)
+- Talk to PDF: Groq chat completions
+- Diagram Generator: Stable Diffusion XL (Hugging Face Inference API)
+"""
+
 import os
-import io
-import pdfplumber
-from huggingface_hub import InferenceClient
-from PIL import Image
+import tempfile
+import traceback
+import time
+from typing import List
+
+import requests
+import fitz  # PyMuPDF
 import gradio as gr
-import base64
-
-# ---------------- CONFIG ----------------
-LLAMA_MODEL = "Groq/Llama-3-Groq-8B-Tool-Use"
-TTS_MODEL = "espnet/kan-bayashi_ljspeech_vits"
-SDXL_MODEL = "stabilityai/stable-diffusion-xl-base-1.0"
-
-HF_TOKEN = os.environ.get("HF_TOKEN")
-GROQ_TOKEN = os.environ.get("GROQ_TOKEN")
-
-client = None
-if GROQ_TOKEN:
-    client = InferenceClient(provider="groq", api_key=GROQ_TOKEN)
-elif HF_TOKEN:
-    client = InferenceClient(api_key=HF_TOKEN)
-
-# ---------------- HELPERS ----------------
-def pdf_to_text(file):
-    text_chunks = []
-    pages = 0
-    with pdfplumber.open(file) as pdf:
-        pages = len(pdf.pages)
-        for page in pdf.pages:
-            ptext = page.extract_text() or ""
-            text_chunks.append(ptext)
-    return "\n\n".join(text_chunks), pages
-
-def llama_summarize(text):
-    messages = [
-        {"role": "system", "content": "You are a concise summarizer. Give 6 short bullet points."},
-        {"role": "user", "content": f"Summarize this document in 6 concise bullet points:\n\n{text}"}
-    ]
-    resp = client.chat.completions.create(model=LLAMA_MODEL, messages=messages)
-    return resp.choices[0].message["content"]
-
-def llama_chat(history, question):
-    messages = history + [{"role": "user", "content": question}]
-    resp = client.chat.completions.create(model=LLAMA_MODEL, messages=messages)
-    return resp.choices[0].message["content"]
-
-def tts_synthesize(text):
-    audio_bytes = client.text_to_speech(model=TTS_MODEL, inputs=text)
-    return audio_bytes
-
-def generate_image(prompt_text):
-    img_bytes = client.text_to_image(prompt_text, model=SDXL_MODEL)
-    return Image.open(io.BytesIO(img_bytes))
-
-def ask_question_and_maybe_diagram(chat_text, question, history):
-    if not history:
-        history = [{"role": "system", "content": f"Document context:\n{chat_text[:4000]}"}]
-    ans = llama_chat(history, question)
-    history.append({"role": "user", "content": question})
-    history.append({"role": "assistant", "content": ans})
-
-    diagram_img = None
-    if question.strip().lower().startswith("!diagram"):
-        prompt = question[len("!diagram"):].strip()
-        if prompt:
-            diagram_img = generate_image(prompt)
-    return ans, diagram_img, history
-
-# ---------------- GRADIO INTERFACE ----------------
-with gr.Blocks() as demo:
-    gr.Markdown("## 📄 PDF Buddy — Summarize • Speak • Chat • Draw")
-    
-    with gr.Row():
-        pdf_file = gr.File(label="Upload PDF", type="file")
-        extract_status = gr.Textbox(label="Status")
-    
-    extracted_text = gr.Textbox(label="Document Preview", lines=10)
-    
-    with gr.Row():
-        summarize_btn = gr.Button("📝 Summarize")
-        summary_output = gr.Textbox(label="Summary", lines=6)
-    
-    tts_btn = gr.Button("🔊 Synthesize Summary to Audio")
-    audio_out = gr.Audio(label="Audio", type="filepath")
-    
-    chat_question = gr.Textbox(label="Ask a question (use !diagram for image)")
-    chat_btn = gr.Button("❓ Ask")
-    chat_output = gr.Textbox(label="Answer")
-    diagram_out = gr.Image(label="Diagram (optional)")
-    chat_history_state = gr.State()
-    
-    # ---------------- CALLBACKS ----------------
-    pdf_file.change(
-        lambda f: pdf_to_text(f) if f else ("No file uploaded", "", None),
-        inputs=pdf_file,
-        outputs=[extract_status, extracted_text]
-    )
-    
-    summarize_btn.click(
-        lambda text: llama_summarize(text[:30000]) if text else "No text to summarize",
-        inputs=extracted_text,
-        outputs=summary_output
-    )
-    
-    tts_btn.click(
-        lambda summary: tts_synthesize(summary) if summary else None,
-        inputs=summary_output,
-        outputs=audio_out
+from groq import Groq
+
+# ================== Load API Keys ==================
+try:
+    from google.colab import userdata
+    if not os.environ.get("LLAMA"):
+        val = userdata.get("LLAMA")
+        if val: os.environ["LLAMA"] = val.strip()
+    if not os.environ.get("DEEPGRAM"):
+        val = userdata.get("DEEPGRAM")
+        if val: os.environ["DEEPGRAM"] = val.strip()
+    if not os.environ.get("HF_TOKEN"):
+        val = userdata.get("HF_TOKEN")
+        if val: os.environ["HF_TOKEN"] = val.strip()
+except Exception:
+    pass
+
+# ================== Config ==================
+CHUNK_CHARS = 20000
+DEFAULT_GROQ_MODEL = "llama-3.1-8b-instant"
+DEEPGRAM_TTS_MODEL = "aura-asteria-en"
+DEEPGRAM_ENCODING = "mp3"
+HF_IMAGE_MODEL = "runwayml/stable-diffusion-v1-5"
+
+# Global variable to store PDF text for Q&A
+pdf_text_storage = {"text": "", "processed": False}
+
+# ================== Utils ==================
+def extract_text_from_pdf(file_path: str) -> str:
+    doc = fitz.open(file_path)
+    text = "\n\n".join(page.get_text("text") for page in doc)
+    doc.close()
+    return text.strip()
+
+def chunk_text(text: str, max_chars: int) -> List[str]:
+    if not text:
+        return []
+    parts, start, L = [], 0, len(text)
+    while start < L:
+        end = min(start + max_chars, L)
+        if end < L:
+            back = text.rfind("\n", start, end)
+            if back == -1:
+                back = text.rfind(" ", start, end)
+            if back != -1 and back > start:
+                end = back
+        parts.append(text[start:end].strip())
+        start = end
+    return parts
+
+# ================== Groq Summarization ==================
+def summarize_chunk_via_groq(chunk_text: str, groq_client: Groq, model: str) -> str:
+    prompt = f"Summarize this text into a concise paragraph (~180 words max):\n\n{chunk_text}"
+    resp = groq_client.chat.completions.create(
+        model=model,
+        messages=[{"role": "user", "content": prompt}],
+        temperature=0.2,
+        max_tokens=800,
     )
-    
-    chat_btn.click(
-        ask_question_and_maybe_diagram,
-        inputs=[extracted_text, chat_question, chat_history_state],
-        outputs=[chat_output, diagram_out, chat_history_state]
+    return resp.choices[0].message.content.strip()
+
+def summarize_document(extracted_text: str, groq_api_key: str, groq_model: str = DEFAULT_GROQ_MODEL) -> str:
+    client = Groq(api_key=groq_api_key)
+    if len(extracted_text) <= CHUNK_CHARS:
+        return summarize_chunk_via_groq(extracted_text, client, groq_model)
+    chunks = chunk_text(extracted_text, CHUNK_CHARS)
+    summaries = []
+    for ch in chunks:
+        try:
+            summaries.append(summarize_chunk_via_groq(ch, client, groq_model))
+        except Exception as e:
+            summaries.append(f"(error summarizing chunk: {str(e)})")
+    final_prompt = "Combine and refine the following summaries into a single clear summary (200-300 words):\n\n" + " ".join(summaries)
+    resp = client.chat.completions.create(
+        model=groq_model,
+        messages=[{"role": "user", "content": final_prompt}],
+        temperature=0.2,
+        max_tokens=900,
     )
+    return resp.choices[0].message.content.strip()
+
+# ================== Deepgram TTS ==================
+def deepgram_tts(summary_text: str, deepgram_api_key: str, model: str = DEEPGRAM_TTS_MODEL, encoding: str = DEEPGRAM_ENCODING) -> str:
+    url = f"https://api.deepgram.com/v1/speak?model={model}&encoding={encoding}"
+    headers = {"Authorization": f"Token {deepgram_api_key}"}
+    payload = {"text": summary_text}
+    resp = requests.post(url, headers=headers, json=payload, timeout=120)
+    if resp.status_code >= 400:
+        raise RuntimeError(f"Deepgram TTS failed ({resp.status_code}): {resp.text}")
+    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=f".{encoding}")
+    tmp.write(resp.content)
+    tmp.close()
+    return tmp.name
+
+# ================== Talk to PDF (Separate Function) ==================
+def ask_pdf_question(question: str, groq_key: str, model: str = DEFAULT_GROQ_MODEL) -> str:
+    if not pdf_text_storage["processed"]:
+        return "❌ Please process a PDF first before asking questions!"
+
+    if not question.strip():
+        return "❌ Please enter a question!"
+
+    if not groq_key.strip():
+        return "❌ Please provide your Groq API key!"
+
+    try:
+        client = Groq(api_key=groq_key)
+        prompt = f"Here is PDF content:\n\n{pdf_text_storage['text'][:15000]}\n\nUser Question: {question}\n\nAnswer strictly based on PDF content. Be concise and specific."
+        resp = client.chat.completions.create(
+            model=model,
+            messages=[{"role": "user", "content": prompt}],
+            temperature=0,
+            max_tokens=500,
+        )
+        return f"🤖 {resp.choices[0].message.content.strip()}"
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
+
+# ================== Diagram via HF (Fixed) ==================
+def generate_diagram(summary: str, hf_token: str, max_retries: int = 3) -> str:
+    headers = {"Authorization": f"Bearer {hf_token}"}
+    url = f"https://api-inference.huggingface.co/models/{HF_IMAGE_MODEL}"
+
+    prompt = f"detailed technical diagram, infographic style, clean illustration of: {summary[:500]}"
+    payload = {"inputs": prompt}
+
+    for attempt in range(max_retries):
+        try:
+            resp = requests.post(url, headers=headers, json=payload, timeout=60)
+
+            if resp.status_code == 503:
+                try:
+                    error_data = resp.json()
+                    if "loading" in error_data.get("error", "").lower():
+                        estimated_time = error_data.get("estimated_time", 20)
+                        time.sleep(estimated_time)
+                        continue
+                except:
+                    pass
+
+            if resp.status_code == 200:
+                content_type = resp.headers.get('content-type', '')
+                if 'image' in content_type or len(resp.content) > 1000:
+                    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".png")
+                    tmp.write(resp.content)
+                    tmp.close()
+                    return tmp.name
+
+            if attempt < max_retries - 1:
+                wait_time = (attempt + 1) * 10
+                time.sleep(wait_time)
+
+        except requests.exceptions.RequestException as e:
+            if attempt < max_retries - 1:
+                time.sleep((attempt + 1) * 5)
+
+    alternative_models = [
+        "stabilityai/stable-diffusion-xl-base-1.0",
+        "CompVis/stable-diffusion-v1-4"
+    ]
+
+    for alt_model in alternative_models:
+        try:
+            alt_url = f"https://api-inference.huggingface.co/models/{alt_model}"
+            resp = requests.post(alt_url, headers=headers, json=payload, timeout=60)
+
+            if resp.status_code == 200:
+                content_type = resp.headers.get('content-type', '')
+                if 'image' in content_type or len(resp.content) > 1000:
+                    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".png")
+                    tmp.write(resp.content)
+                    tmp.close()
+                    return tmp.name
+        except Exception:
+            continue
+
+    return create_text_diagram_placeholder(summary)
+
+def create_text_diagram_placeholder(summary: str) -> str:
+    try:
+        from PIL import Image, ImageDraw, ImageFont
+
+        width, height = 800, 600
+        img = Image.new('RGB', (width, height), color='#0a0a0a')
+        draw = ImageDraw.Draw(img)
+
+        try:
+            font = ImageFont.truetype("arial.ttf", 16)
+            title_font = ImageFont.truetype("arial.ttf", 20)
+        except:
+            font = ImageFont.load_default()
+            title_font = ImageFont.load_default()
+
+        draw.text((50, 50), "📊 Document Summary", fill='#00ff88', font=title_font)
+
+        words = summary.split()
+        lines = []
+        current_line = []
+        max_width = 45
+
+        for word in words:
+            if len(' '.join(current_line + [word])) <= max_width:
+                current_line.append(word)
+            else:
+                if current_line:
+                    lines.append(' '.join(current_line))
+                current_line = [word]
+        if current_line:
+            lines.append(' '.join(current_line))
+
+        y_offset = 100
+        for line in lines[:18]:
+            draw.text((50, y_offset), line, fill='#ccffcc', font=font)
+            y_offset += 25
+
+        draw.rectangle([25, 25, width-25, height-25], outline='#00ff88', width=3)
+
+        tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".png")
+        img.save(tmp.name, "PNG")
+        tmp.close()
+        return tmp.name
+
+    except Exception:
+        tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".txt")
+        tmp.write(f"Diagram generation failed. Summary: {summary[:200]}...".encode())
+        tmp.close()
+        return tmp.name
+
+# ================== Main Pipeline ==================
+def process_pdf_pipeline(pdf_file, groq_key, deepgram_key, hf_token, groq_model):
+    try:
+        if not groq_key.strip():
+            return "❌ Missing Groq API key!", None, None, "Process a PDF first!"
+        if not deepgram_key.strip():
+            return "❌ Missing Deepgram API key!", None, None, "Process a PDF first!"
+        if not hf_token.strip():
+            return "❌ Missing HuggingFace token!", None, None, "Process a PDF first!"
+        if pdf_file is None:
+            return "❌ Please upload a PDF file!", None, None, "Process a PDF first!"
+
+        pdf_path = pdf_file.name if hasattr(pdf_file, "name") else str(pdf_file)
+
+        # Extract and store text globally
+        text = extract_text_from_pdf(pdf_path)
+        if not text.strip():
+            return "❌ PDF contains no extractable text!", None, None, "Process a PDF first!"
+
+        # Store text for Q&A
+        pdf_text_storage["text"] = text
+        pdf_text_storage["processed"] = True
+
+        # Generate summary
+        summary = summarize_document(text, groq_api_key=groq_key, groq_model=groq_model or DEFAULT_GROQ_MODEL)
+
+        # Generate audio
+        audio_path = deepgram_tts(summary, deepgram_api_key=deepgram_key)
+
+        # Generate diagram
+        diagram_path = generate_diagram(summary, hf_token)
+
+        return summary, audio_path, diagram_path, "✅ PDF processed! You can now ask questions below."
+
+    except Exception as e:
+        pdf_text_storage["processed"] = False
+        return f"❌ Error: {str(e)}", None, None, "Process a PDF first!"
+
+# ================== Gen-Z Dark Theme CSS ==================
+GENZ_CSS = """
+/* Main container styling */
+.gradio-container {
+    background: linear-gradient(135deg, #000000 0%, #0a0a0a 100%) !important;
+    color: #00ff88 !important;
+    font-family: 'Segoe UI', 'Roboto', sans-serif !important;
+}
+
+body {
+    background: #000000 !important;
+    color: #00ff88 !important;
+}
+
+/* Input fields styling */
+input, textarea, .gradio-textbox, .gradio-file, select {
+    background: linear-gradient(145deg, #111111, #1a1a1a) !important;
+    color: #00ff88 !important;
+    border: 2px solid #00ff88 !important;
+    border-radius: 12px !important;
+    box-shadow: 0 4px 15px rgba(0, 255, 136, 0.2) !important;
+    transition: all 0.3s ease !important;
+}
+
+input:focus, textarea:focus, .gradio-textbox:focus {
+    border-color: #00ff00 !important;
+    box-shadow: 0 0 25px rgba(0, 255, 136, 0.5) !important;
+    transform: translateY(-2px) !important;
+}
+
+/* Button styling */
+button {
+    background: linear-gradient(145deg, #00ff88, #00cc66) !important;
+    color: #000000 !important;
+    border: none !important;
+    border-radius: 15px !important;
+    font-weight: bold !important;
+    text-transform: uppercase !important;
+    letter-spacing: 1px !important;
+    box-shadow: 0 6px 20px rgba(0, 255, 136, 0.3) !important;
+    transition: all 0.3s ease !important;
+}
+
+button:hover {
+    background: linear-gradient(145deg, #00cc66, #00ff88) !important;
+    transform: translateY(-3px) !important;
+    box-shadow: 0 8px 25px rgba(0, 255, 136, 0.5) !important;
+}
+
+button:active {
+    transform: translateY(1px) !important;
+}
+
+/* Headers and text */
+h1, h2, h3, h4, .gradio-markdown {
+    color: #00ff88 !important;
+    text-shadow: 0 0 10px rgba(0, 255, 136, 0.3) !important;
+}
+
+h1 {
+    font-size: 2.5em !important;
+    background: linear-gradient(45deg, #00ff88, #00cc66) !important;
+    -webkit-background-clip: text !important;
+    -webkit-text-fill-color: transparent !important;
+}
+
+/* Tabs styling */
+.gradio-tab {
+    background: linear-gradient(145deg, #111111, #1a1a1a) !important;
+    color: #00ff88 !important;
+    border: 2px solid #00ff88 !important;
+    border-radius: 10px !important;
+}
+
+.gradio-tab.selected {
+    background: linear-gradient(145deg, #00ff88, #00cc66) !important;
+    color: #000000 !important;
+}
+
+/* Slider styling */
+.gradio-slider input[type="range"] {
+    background: #00ff88 !important;
+}
+
+.gradio-slider .gradio-slider-track {
+    background: #333333 !important;
+}
+
+.gradio-slider .gradio-slider-thumb {
+    background: #00ff88 !important;
+    border: 2px solid #00cc66 !important;
+}
+
+/* File upload area */
+.gradio-file {
+    border: 3px dashed #00ff88 !important;
+    background: rgba(0, 255, 136, 0.1) !important;
+    border-radius: 15px !important;
+}
+
+/* Progress bar */
+.progress-bar {
+    background: linear-gradient(90deg, #00ff88, #00cc66) !important;
+    border-radius: 10px !important;
+}
+
+/* Accordion styling */
+.gradio-accordion {
+    background: linear-gradient(145deg, #111111, #1a1a1a) !important;
+    border: 2px solid #00ff88 !important;
+    border-radius: 12px !important;
+}
+
+/* Scrollbar */
+::-webkit-scrollbar {
+    width: 12px !important;
+}
+
+::-webkit-scrollbar-track {
+    background: #111111 !important;
+}
+
+::-webkit-scrollbar-thumb {
+    background: linear-gradient(145deg, #00ff88, #00cc66) !important;
+    border-radius: 6px !important;
+}
+
+/* Glowing effects */
+.glow {
+    box-shadow: 0 0 20px rgba(0, 255, 136, 0.5) !important;
+}
+
+/* Custom animations */
+@keyframes pulse {
+    0% { box-shadow: 0 0 20px rgba(0, 255, 136, 0.3); }
+    50% { box-shadow: 0 0 30px rgba(0, 255, 136, 0.6); }
+    100% { box-shadow: 0 0 20px rgba(0, 255, 136, 0.3); }
+}
+
+.pulse-effect {
+    animation: pulse 2s infinite !important;
+}
+"""
+
+# ================== UI Build Function ==================
+def build_ui():
+    env_groq = os.environ.get("LLAMA", "")
+    env_deepgram = os.environ.get("DEEPGRAM", "")
+    env_hf = os.environ.get("HF_TOKEN", "")
+
+    with gr.Blocks(css=GENZ_CSS, title="🔥 PDF AI Pipeline", theme=gr.themes.Base()) as demo:
+
+        # Header - Centered
+        gr.Markdown("""
+        <div style="text-align: center; margin: 20px 0;">
+            <h1 style="font-size: 3.5em; margin-bottom: 10px;">🔥 AI PDF PROCESSOR</h1>
+            <h2 style="font-size: 1.8em; margin-bottom: 10px;">Transform PDFs into Audio, Summaries & Interactive Q&A</h2>
+            <h3 style="font-size: 1.2em; font-style: italic; opacity: 0.9;"> PEC COHORT 3</h3>
+        </div>
+        """, elem_classes=["pulse-effect"])
+
+        with gr.Row():
+            # Left Column - Upload & API Settings
+            with gr.Column(scale=1):
+                with gr.Accordion("📁 UPLOAD PDF", open=True):
+                    pdf_input = gr.File(
+                        label="Drop your PDF here",
+                        file_types=[".pdf"],
+                        height=150
+                    )
+
+                with gr.Accordion("🔑 API KEYS", open=False):
+                    gr.Markdown("*Keep your keys secure • Use env vars in production*")
+                    groq_key = gr.Textbox(
+                        label="🤖 Groq API Key",
+                        value=env_groq,
+                        type="password",
+                        placeholder="sk-..."
+                    )
+                    deepgram_key = gr.Textbox(
+                        label="🎤 Deepgram API Key",
+                        value=env_deepgram,
+                        type="password",
+                        placeholder="Enter Deepgram key"
+                    )
+                    hf_key = gr.Textbox(
+                        label="🤗 HuggingFace Token",
+                        value=env_hf,
+                        type="password",
+                        placeholder="hf_..."
+                    )
+
+                with gr.Accordion("⚙️ SETTINGS", open=False):
+                    groq_model = gr.Dropdown(
+                        label="🧠 AI Model",
+                        choices=[
+                            "llama-3.1-8b-instant",
+                            "llama-3.1-70b-versatile",
+                            "mixtral-8x7b-32768",
+                            "gemma2-9b-it"
+                        ],
+                        value=DEFAULT_GROQ_MODEL
+                    )
+
+                # Main Process Button
+                process_btn = gr.Button(
+                    "🚀 PROCESS PDF",
+                    variant="primary",
+                    size="lg",
+                    elem_classes=["pulse-effect"]
+                )
+
+            # Right Column - Results
+            with gr.Column(scale=2):
+                with gr.Tabs():
+                    with gr.Tab("📝 SUMMARY"):
+                        summary_output = gr.Textbox(
+                            label="AI Generated Summary",
+                            lines=12,
+                            placeholder="Your PDF summary will appear here...",
+                            interactive=False
+                        )
+
+                    with gr.Tab("🔊 AUDIO"):
+                        audio_output = gr.Audio(
+                            label="Listen to Summary",
+                            type="filepath",
+                            interactive=False
+                        )
+
+                    with gr.Tab("🎨 DIAGRAM"):
+                        diagram_output = gr.Image(
+                            label="Visual Representation",
+                            interactive=False,
+                            height=400
+                        )
+
+        # Separate Q&A Section
+        gr.Markdown("---")
+        gr.Markdown("## 💬 CHAT WITH YOUR PDF")
+
+        with gr.Row():
+            with gr.Column(scale=3):
+                question_input = gr.Textbox(
+                    label="Ask anything about your PDF",
+                    placeholder="What are the main findings? • Who are the key people mentioned? • Summarize chapter 2...",
+                    lines=2
+                )
+            with gr.Column(scale=1):
+                ask_btn = gr.Button("📨 SEND", variant="secondary", size="lg")
+
+        chat_output = gr.Textbox(
+            label="🤖 AI Response",
+            lines=8,
+            placeholder="Upload and process a PDF first, then ask your questions!",
+            interactive=False
+        )
+
+        # Status indicator
+        status_output = gr.Textbox(
+            label="📊 Status",
+            value="Ready to process PDF...",
+            interactive=False
+        )
+
+        # Footer
+        gr.Markdown("""
+        ---
+        **🔥 Pro Tips:**
+        • Upload PDFs with extractable text (not image-only)
+        • Questions work only after processing
+        • Audio generation takes ~30-60 seconds
+        • Diagrams may take longer depending on HF API load
+
+        *Built with ❤️ for the AI generation*
+        """)
+
+        # Event handlers
+        process_btn.click(
+            fn=process_pdf_pipeline,
+            inputs=[pdf_input, groq_key, deepgram_key, hf_key, groq_model],
+            outputs=[summary_output, audio_output, diagram_output, status_output],
+            show_progress=True
+        )
+
+        ask_btn.click(
+            fn=ask_pdf_question,
+            inputs=[question_input, groq_key, groq_model],
+            outputs=[chat_output],
+            show_progress=False
+        )
+
+        # Enter key support for questions
+        question_input.submit(
+            fn=ask_pdf_question,
+            inputs=[question_input, groq_key, groq_model],
+            outputs=[chat_output]
+        )
+
+    return demo
 
-demo.launch()
+if __name__ == "__main__":
+    demo = build_ui()
+    demo.launch(
+        share=True,
+        debug=True,
+        show_error=True
+    )