Update app.py

app.py

@@ -1,347 +1,102 @@
-import logging
-import sys
-import os
-from agno.agent import Agent
-from agno.models.openai import OpenAIChat
-from agno.knowledge.embedder.openai import OpenAIEmbedder
-from agno.tools.duckduckgo import DuckDuckGoTools
-from agno.knowledge.knowledge import Knowledge
-from agno.vectordb.lancedb import LanceDb, SearchType
-import gradio as gr
-import fitz  # PyMuPDF
-from PIL import Image
-import io
-import requests
-import re
-import time
-
-# --- Logging ---
-logging.basicConfig(stream=sys.stdout, level=logging.INFO)
-logger = logging.getLogger(__name__)
-
-# --- Secrets ---
-OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
-
-if not OPENAI_API_KEY:
-    raise ValueError("Missing OPENAI_API_KEY")
-
-# Create a knowledge base with PDF documents
-knowledge = Knowledge(
-    vector_db=LanceDb(
-        uri="tmp/lancedb",
-        table_name="pdf_documents",
-        search_type=SearchType.vector,
-        embedder=OpenAIEmbedder(id="text-embedding-3-small"),
+===== Application Startup at 2026-04-20 10:40:26 =====
+
+INFO Creating table: pdf_documents                                              
+[2026-04-20T10:41:03Z WARN  lance::dataset::write::insert] No existing dataset at /app/tmp/lancedb/pdf_documents.lance, it will be created
+INFO:__main__:Downloading https://media.datacamp.com/cms/working-with-hugging-face.pdf...
+INFO:__main__:Downloaded pdf_cache/file_0.pdf (1445572 bytes)
+INFO:__main__:Prepared PDF 1: https://media.datacamp.com/cms/working-with-hugging-face.pdf
+INFO:__main__:Downloading https://media.datacamp.com/cms/ai-agents-cheat-sheet.pdf...
+INFO:__main__:Downloaded pdf_cache/file_1.pdf (2837543 bytes)
+INFO:__main__:Prepared PDF 2: https://media.datacamp.com/cms/ai-agents-cheat-sheet.pdf
+INFO:__main__:Downloading https://media.datacamp.com/cms/introduction-to-sql-with-ai-1.pdf...
+INFO:__main__:Downloaded pdf_cache/file_2.pdf (2641274 bytes)
+INFO:__main__:Prepared PDF 3: https://media.datacamp.com/cms/introduction-to-sql-with-ai-1.pdf
+INFO:__main__:Downloading https://media.datacamp.com/legacy/image/upload/v1719844709/Marketing/Blog/Azure_CLI_Cheat_Sheet.pdf...
+INFO:__main__:Downloaded pdf_cache/file_3.pdf (3420887 bytes)
+INFO:__main__:Prepared PDF 4: https://media.datacamp.com/legacy/image/upload/v1719844709/Marketing/Blog/Azure_CLI_Cheat_Sheet.pdf
+INFO Adding content from path, 8cd3ebb7-9be0-58ca-a79b-f679d9a51ee8, None,      
+     pdf_cache/file_0.pdf, None                                                 
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO Adding content from path, 4eba8a35-09bf-5fb9-b323-0de8d692bd46, None,      
+     pdf_cache/file_1.pdf, None                                                 
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO Adding content from path, f134cc6f-57ae-5d08-8ad3-e32df428c7c2, None,      
+     pdf_cache/file_2.pdf, None                                                 
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO Adding content from path, 57aa3a67-8e67-5cd0-9a48-6d347d69435e, None,      
+     pdf_cache/file_3.pdf, None                                                 
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO:__main__:✅ Successfully added 4 PDFs using add_content
+INFO:__main__:Agent initialized successfully
+INFO:httpx:HTTP Request: HEAD https://huggingface.co/api/telemetry/https%3A/api.gradio.app/gradio-initiated-analytics "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: GET https://api.gradio.app/pkg-version "HTTP/1.1 200 OK"
+INFO:__main__:Starting Gradio interface...
+* Running on local URL:  http://0.0.0.0:7860, with SSR ⚡ (experimental, to disable set `ssr_mode=False` in `launch()`)
+INFO:httpx:HTTP Request: GET http://localhost:7860/gradio_api/startup-events "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: HEAD http://0.0.0.0:7861/ "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: HEAD http://localhost:7860/ "HTTP/1.1 200 OK"
+* To create a public link, set `share=True` in `launch()`.
+INFO:httpx:HTTP Request: HEAD https://huggingface.co/api/telemetry/https%3A/api.gradio.app/gradio-launched-telemetry "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: GET http://0.0.0.0:7861/?__theme=system "HTTP/1.1 200 OK"
+INFO:__main__:Question asked: How do AI agents work?...
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/embeddings "HTTP/1.1 200 OK"
+INFO Found 10 documents                                                         
+INFO:httpx:HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
+INFO:httpx:HTTP Request: POST https://os-api.agno.com/telemetry/runs "HTTP/2 201 Created"
+INFO:__main__:Displaying PDF from: https://media.datacamp.com/cms/ai-agents-cheat-sheet.pdf
+Traceback (most recent call last):
+  File "/usr/local/lib/python3.13/site-packages/gradio/queueing.py", line 856, in process_events
+    response = await route_utils.call_process_api(
+               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+    ...<5 lines>...
     )
-)
-
-# Download and add PDFs to knowledge base
-pdf_urls = [
-    "https://media.datacamp.com/cms/working-with-hugging-face.pdf",
-    "https://media.datacamp.com/cms/ai-agents-cheat-sheet.pdf",
-    "https://media.datacamp.com/cms/introduction-to-sql-with-ai-1.pdf",
-    "https://media.datacamp.com/legacy/image/upload/v1719844709/Marketing/Blog/Azure_CLI_Cheat_Sheet.pdf"
-]
-
-def download_if_needed(url, filename):
-    if not os.path.exists(filename):
-        logger.info(f"Downloading {url}...")
-        response = requests.get(url)
-        with open(filename, "wb") as f:
-            f.write(response.content)
-        logger.info(f"Downloaded {filename} ({len(response.content)} bytes)")
-
-# Create a directory for PDFs if it doesn't exist
-os.makedirs("pdf_cache", exist_ok=True)
-
-def add_pdfs_to_knowledge():
-    """Add PDFs to knowledge base using the correct method for the installed agno version"""
-    contents_to_add = []
-    
-    for i, url in enumerate(pdf_urls):
-        filename = f"pdf_cache/file_{i}.pdf"
-        try:
-            download_if_needed(url, filename)
-            contents_to_add.append({
-                "path": filename,
-                "metadata": {"source": url}
-            })
-            logger.info(f"Prepared PDF {i+1}: {url}")
-        except Exception as e:
-            logger.error(f"Failed to prepare PDF {i+1}: {str(e)}")
-    
-    if contents_to_add:
-        try:
-            if hasattr(knowledge, 'add_contents'):
-                knowledge.add_contents(contents_to_add)
-                logger.info(f"✅ Successfully added {len(contents_to_add)} PDFs using add_contents")
-            elif hasattr(knowledge, 'add_content'):
-                for item in contents_to_add:
-                    knowledge.add_content(**item)
-                logger.info(f"✅ Successfully added {len(contents_to_add)} PDFs using add_content")
-            else:
-                from agno.document.reader.pdf_reader import PDFReader
-                reader = PDFReader()
-                all_docs = []
-                for item in contents_to_add:
-                    docs = reader.read(item["path"])
-                    for doc in docs:
-                        doc.metadata = item["metadata"]
-                        all_docs.append(doc)
-                knowledge.vector_db.insert(documents=all_docs)
-                logger.info(f"✅ Successfully added {len(all_docs)} document chunks from {len(contents_to_add)} PDFs")
-            
-            time.sleep(2)
-        except Exception as e:
-            logger.error(f"Failed to add PDFs: {str(e)}")
-            raise
-    else:
-        logger.warning("No PDFs were prepared to add")
-
-# Add PDFs to knowledge base
-add_pdfs_to_knowledge()
-
-# Define the agent
-agent = Agent(
-    model=OpenAIChat(id="gpt-4.1-mini", temperature=0.2),
-    description="You are Dox a data expert!",
-    instructions="""
-    You are a data professional's assistant named Dox.
-
-    Your primary goal is to answer questions about data, programming, cloud computing, AI/ML, and technology topics.
-
-    Here are your operating procedures:
-
-    1.  **Information Gathering Strategy**:
-        *   **Prioritize Knowledge Base**: First, search your internal knowledge base for the answer.
-        *   **Supplement with Web Search**: If the knowledge base information is outdated, insufficient, or the question is better suited for current web information, use the DuckDuckGo tool to perform web searches to fill in gaps or find the most up-to-date data.
-        *   For general technology questions not in your knowledge base, use web search to provide accurate answers.
-        *   If the question is NOT data-related, you MUST respond with: "Please ask relevant data questions only." and terminate.
-
-    2.  **Response Length Guidelines**:
-        *   For basic questions, keep your answer to a maximum of 300 words.
-        *   For complex questions, extend your answer to a maximum of 500 words.
-
-    3.  **Citation Rules (CRITICAL)**:
-        *   **Knowledge Base Citation**: For any information sourced from your internal knowledge base, you MUST include a citation on a NEW LINE after the answer, starting with "Source: ", followed by the metadata field 'source' to get the hyperlink.
-        *   **Web Search Citation**: For any information obtained from the web using the DuckDuckGo tool, you MUST include a citation on a NEW LINE after the answer, starting with "Online Source: ", followed by the full hyperlink.
-        *   **Final Rule for Citations**: Always end your answers with the appropriate citations, ensuring they are on separate lines as specified. Do NOT mix or combine citation types on a single line.
-        *   ALWAYS cite with links NOT text like "from internal knowledge base"
-
-    4.  **Accuracy and Non-Hallucination**:
-        *   Provide factual and relevant answers based ONLY on the information found in your knowledge base or through web searches.
-        *   NEVER invent or hallucinate information. If an answer cannot be found, state that directly.
-
-    Make sure to follow these instructions precisely.
-    """,
-    knowledge=knowledge,
-    add_datetime_to_context=True,
-    add_location_to_context=True,
-    search_knowledge=True,
-    tools=[DuckDuckGoTools()],
-    markdown=True
-)
-
-logger.info("Agent initialized successfully")
-
-# -----------------------------
-# Helper Functions
-# -----------------------------
-def extract_pdf_url_from_text(text):
-    """Extract PDF URL from text content"""
-    patterns = [
-        r'Source:\s*(https?://[^\s]+\.pdf)',
-        r'Online Source:\s*(https?://[^\s]+\.pdf)',
-        r'(https?://[^\s]+\.pdf)'
-    ]
-    
-    for pattern in patterns:
-        match = re.search(pattern, text, re.IGNORECASE)
-        if match:
-            return match.group(1)
-    return None
-
-def download_pdf_from_url(url):
-    """Download PDF from URL"""
-    response = requests.get(url, timeout=30)
-    response.raise_for_status()
-    return response.content
-
-def display_pdf(pdf_url):
-    """Convert first page of PDF to image for display"""
-    if not pdf_url:
-        return None
-    
-    try:
-        logger.info(f"Displaying PDF from: {pdf_url}")
-        pdf_bytes = download_pdf_from_url(pdf_url)
-        doc = fitz.open(stream=pdf_bytes, filetype="pdf")
-        page = doc[0]
-        
-        zoom = 1.5
-        mat = fitz.Matrix(zoom, zoom)
-        pix = page.get_pixmap(matrix=mat)
-        
-        img_data = pix.tobytes("png")
-        img = Image.open(io.BytesIO(img_data))
-        doc.close()
-        
-        return img
-    except Exception as e:
-        logger.error(f"Error displaying PDF: {str(e)}")
-        return None
-
-# -----------------------------
-# Main Chat Function
-# -----------------------------
-def respond(message, history):
-    """Process user message and return response with automatic PDF display"""
-    logger.info(f"Question asked: {message[:100]}...")
-    
-    # Get response from agent
-    response = agent.run(message, use_knowledge=True)
-    full_content = response.get_content_as_string()
-    
-    # Extract PDF URL from the response
-    pdf_url = extract_pdf_url_from_text(full_content)
-    
-    # Update history - Gradio 6.0 expects a list of lists/tuples
-    history = history or []
-    history.append([message, full_content])
-    
-    return history, pdf_url
-
-# -----------------------------
-# Example Questions
-# -----------------------------
-examples = [
-    "What is a transformer model in NLP?",
-    "How do AI agents work?",
-    "Explain SQL joins with examples",
-    "What are the key Azure CLI commands?",
-    "How to use Hugging Face for text classification?",
-    "What's the difference between supervised and unsupervised learning?"
-]
-
-# -----------------------------
-# Create Gradio Interface (Gradio 6.0 compatible)
-# -----------------------------
-with gr.Blocks(title="Dox - Data Expert Assistant") as demo:
-    
-    gr.Markdown("""
-    # 🤖 Dox - Your Data Expert Assistant
-    
-    Ask me anything about data science, programming, AI/ML, cloud computing, and technology topics!
-    I'll search my knowledge base and the web to provide accurate answers with citations.
-    """)
-    
-    with gr.Row():
-        # Main chat column
-        with gr.Column(scale=2):
-            # Chatbot with minimal parameters for Gradio 6.0
-            chatbot = gr.Chatbot(label="Conversation", height=500)
-            
-            with gr.Row():
-                msg = gr.Textbox(
-                    label="Ask your question",
-                    placeholder="e.g., What is a transformer model in NLP?",
-                    lines=2,
-                    scale=9
-                )
-                send_btn = gr.Button("Send 📤", scale=1, variant="primary")
-            
-            # Examples section
-            gr.Markdown("### 💡 Example Questions")
-            gr.Examples(
-                examples=examples,
-                inputs=msg,
-                label="Click any example to ask"
-            )
-        
-        # PDF preview column
-        with gr.Column(scale=1):
-            gr.Markdown("### 📄 Referenced Document")
-            pdf_display = gr.Image(
-                label="PDF Preview",
-                type="pil",
-                height=450,
-                visible=False,
-                show_label=True,
-                interactive=False
-            )
-            
-            gr.Markdown("""
-            ---
-            ### 📌 Tips:
-            - **PDFs automatically appear** when I cite them
-            - I prioritize my knowledge base then search the web
-            - Citations include direct links to sources
-            - Keep questions data/tech related
-            """)
-    
-    # Hidden state to track current PDF URL
-    current_pdf_url = gr.State(value=None)
-    
-    # Function to handle sending message and updating UI
-    def send_message(message, history):
-        if not message or not message.strip():
-            return history, gr.update(), None
-        
-        # Get response and PDF URL
-        new_history, pdf_url = respond(message, history)
-        
-        # Update PDF display
-        if pdf_url:
-            pdf_image = display_pdf(pdf_url)
-            if pdf_image:
-                return new_history, gr.update(value=pdf_image, visible=True), pdf_url
-            else:
-                return new_history, gr.update(visible=False), None
-        else:
-            return new_history, gr.update(visible=False), None
-    
-    # Function to clear chat
-    def clear_chat():
-        return [], gr.update(visible=False), None
-    
-    # Wire up events
-    send_btn.click(
-        send_message,
-        inputs=[msg, chatbot],
-        outputs=[chatbot, pdf_display, current_pdf_url]
-    ).then(
-        lambda: "",  # Clear input
-        None,
-        msg
+    ^
+  File "/usr/local/lib/python3.13/site-packages/gradio/route_utils.py", line 358, in call_process_api
+    output = await app.get_blocks().process_api(
+             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+    ...<11 lines>...
     )
-    
-    msg.submit(
-        send_message,
-        inputs=[msg, chatbot],
-        outputs=[chatbot, pdf_display, current_pdf_url]
-    ).then(
-        lambda: "",
-        None,
-        msg
+    ^
+  File "/usr/local/lib/python3.13/site-packages/gradio/blocks.py", line 2192, in process_api
+    data = await self.postprocess_data(
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+        block_fn, result["prediction"], state
+        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
     )
-    
-    clear_btn = gr.Button("🗑️ Clear Chat", variant="secondary")
-    clear_btn.click(clear_chat, None, [chatbot, pdf_display, current_pdf_url])
-    
-    # Add footer
-    gr.Markdown("""
-    ---
-    *Powered by Agno Framework, OpenAI, and LanceDB | Responses include citations from knowledge base and web*
-    """)
-
-# -----------------------------
-# Run the app
-# -----------------------------
-if __name__ == "__main__":
-    logger.info("Starting Gradio interface...")
-    demo.launch(
-        share=False,
-        server_port=7860,
-        server_name="0.0.0.0",
-        theme=gr.themes.Soft()
-    )
+    ^
+  File "/usr/local/lib/python3.13/site-packages/gradio/blocks.py", line 1955, in postprocess_data
+    prediction_value = await anyio.to_thread.run_sync(
+                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+        block.postprocess, prediction_value, limiter=self.limiter
+        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+    )
+    ^
+  File "/usr/local/lib/python3.13/site-packages/anyio/to_thread.py", line 63, in run_sync
+    return await get_async_backend().run_sync_in_worker_thread(
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+        func, args, abandon_on_cancel=abandon_on_cancel, limiter=limiter
+        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+    )
+    ^
+  File "/usr/local/lib/python3.13/site-packages/anyio/_backends/_asyncio.py", line 2518, in run_sync_in_worker_thread
+    return await future
+           ^^^^^^^^^^^^
+  File "/usr/local/lib/python3.13/site-packages/anyio/_backends/_asyncio.py", line 1002, in run
+    result = context.run(func, *args)
+  File "/usr/local/lib/python3.13/site-packages/gradio/components/chatbot.py", line 704, in postprocess
+    self._check_format(value)
+    ~~~~~~~~~~~~~~~~~~^^^^^^^
+  File "/usr/local/lib/python3.13/site-packages/gradio/components/chatbot.py", line 402, in _check_format
+    raise Error(
+        "Data incompatible with messages format. Each message should be a dictionary with 'role' and 'content' keys or a ChatMessage object."
+    )
+gradio.exceptions.Error: "Data incompatible with messages format. Each message should be a dictionary with 'role' and 'content' keys or a ChatMessage object."