HuggingFace Fix PDF Edition

README.md

@@ -1,6 +1,6 @@
 ---
-title: YouTube to LinkedIn Post Converter
-emoji: 🎥
+title: PDF to LinkedIn Post Converter
+emoji: 📄
 colorFrom: blue
 colorTo: purple
 sdk: gradio
@@ -9,11 +9,11 @@ app_file: app.py
 pinned: false
 ---
 
-# YouTube to LinkedIn Post Converter
+# PDF to LinkedIn Post Converter
 
-Transform your YouTube videos into professional LinkedIn posts with AI-powered content enhancement. This application:
+Transform your PDF documents into professional LinkedIn posts with AI-powered content enhancement. This application:
 
-- Extracts transcripts from YouTube videos
+- Extracts content from PDF files
 - Enhances content using AI
 - Formats posts for LinkedIn
 - Verifies content quality
@@ -21,7 +21,7 @@ Transform your YouTube videos into professional LinkedIn posts with AI-powered c
 
 ## Features
 
-- 🎥 YouTube video processing
+- 📄 PDF document processing
 - ✨ AI-powered content enhancement
 - 🔗 LinkedIn post formatting
 - ✓ Content verification
@@ -29,14 +29,22 @@ Transform your YouTube videos into professional LinkedIn posts with AI-powered c
 
 ## How to Use
 
-1. Enter a YouTube video URL
+1. Upload a PDF file
 2. Click "Generate Post"
 3. Review the enhanced content
 4. Copy your LinkedIn-ready post
 
-## Sample Videos
+## Tips for Best Results
 
-Try these videos to test the application:
-- Open AI video: https://www.youtube.com/watch?v=LsMxX86mm2Y
-- Financial News: https://www.youtube.com/watch?v=hvP1UNALZ3g
-- Video About AI: https://www.youtube.com/watch?v=Yq0QkCxoTHM 
+- Use well-formatted PDFs with clear text
+- Optimal length: 2-10 pages
+- Ensure PDFs have readable text (not scanned images)
+- Review and personalize the post before sharing
+- Consider your target audience when selecting content
+
+## Sample PDFs
+
+Try these PDFs to test the application:
+- Open AI PDF: https://example.com/open-ai.pdf
+- Financial News PDF: https://example.com/financial-news.pdf
+- PDF About AI: https://example.com/ai.pdf 

app.py

@@ -1,7 +1,6 @@
 import gradio as gr
 import os
 from dotenv import load_dotenv
-from youtube_transcript_api import YouTubeTranscriptApi
 from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 from langchain.prompts import ChatPromptTemplate
 from langchain_core.output_parsers import StrOutputParser
@@ -12,6 +11,7 @@ from langchain_chroma import Chroma
 from langchain.schema import Document
 from datetime import datetime
 from langchain.text_splitter import RecursiveCharacterTextSplitter
+from pypdf import PdfReader
 
 # Load environment variables
 load_dotenv(verbose=True)
@@ -22,8 +22,8 @@ if not os.getenv("OPENAI_API_KEY"):
 
 # Define state types
 class ProcessState(TypedDict):
-    video_url: str
-    transcript: str
+    pdf_file: str
+    content: str
     enhanced: str
     linkedin_post: str
     verification: dict
@@ -34,29 +34,28 @@ class ProcessState(TypedDict):
     needs_improvement: bool
     research_context: str
 
-def extract_video_id(url: str) -> str:
-    """Extract video ID from YouTube URL."""
-    if "youtu.be" in url:
-        return url.split("/")[-1]
-    return url.split("v=")[-1].split("&")[0]
+def extract_pdf_content(pdf_file: str) -> str:
+    """Extract text content from PDF file."""
+    try:
+        reader = PdfReader(pdf_file)
+        text = ""
+        for page in reader.pages:
+            text += page.extract_text() + "\n"
+        return text.strip()
+    except Exception as e:
+        raise Exception(f"Error extracting PDF content: {str(e)}")
 
-def get_transcript(state: ProcessState, progress=gr.Progress()) -> ProcessState:
-    """Get transcript from YouTube video."""
+def get_content(state: ProcessState, progress=gr.Progress()) -> ProcessState:
+    """Get content from PDF file."""
     try:
-        progress(0.25, desc="Fetching transcript...")
-        video_id = extract_video_id(state["video_url"])
-        transcript = YouTubeTranscriptApi.get_transcript(video_id)
-        state["transcript"] = " ".join([segment["text"] for segment in transcript])
-        state["status"] = "✅ Transcript fetched"
+        progress(0.25, desc="Extracting PDF content...")
+        content = extract_pdf_content(state["pdf_file"])
+        state["content"] = content
+        state["status"] = "✅ PDF content extracted"
         return state
     except Exception as e:
-        error_message = str(e).lower()
-        if "too many requests" in error_message or "429" in error_message:
-            state["error"] = "⚠️ YouTube API rate limit reached. Please wait a few minutes and try again."
-            state["status"] = "❌ Rate limit exceeded"
-        else:
-            state["error"] = f"⚠️ Error fetching transcript: {str(e)}"
-            state["status"] = "❌ Failed to fetch transcript"
+        state["error"] = f"⚠️ Error extracting PDF content: {str(e)}"
+        state["status"] = "❌ Failed to extract PDF content"
         return state
 
 def get_chroma_collection():
@@ -72,9 +71,9 @@ def get_chroma_collection():
         raise Exception(f"Error creating Chroma collection: {str(e)}")
 
 def enhance_content(state: ProcessState, progress=gr.Progress()) -> ProcessState:
-    """Enhance the transcript content with semantic search and similarity analysis."""
+    """Enhance the PDF content with semantic search and similarity analysis."""
     try:
-        if not state["transcript"]:
+        if not state["content"]:
             return state
         
         progress(0.50, desc="Enhancing content...")
@@ -82,14 +81,14 @@ def enhance_content(state: ProcessState, progress=gr.Progress()) -> ProcessState
         # Get similar content from the vector store
         collection = get_chroma_collection()
         similar_docs = collection.similarity_search(
-            state["transcript"],
+            state["content"],
             k=3
         )
         
         # Initialize LLM for content generation
         llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.7)
         prompt = ChatPromptTemplate.from_messages([
-            ("system", """You are an expert content enhancer. Transform this transcript into engaging content:
+            ("system", """You are an expert content enhancer. Transform this PDF content into engaging content:
 
 1. Identify and emphasize key points
 2. Add context and examples
@@ -97,8 +96,8 @@ def enhance_content(state: ProcessState, progress=gr.Progress()) -> ProcessState
 4. Keep it concise (max 3000 characters)
 5. Maintain factual accuracy
 
-Transcript:
-{transcript}
+Content:
+{content}
 
 Similar Content for Context:
 {similar_content}
@@ -108,7 +107,7 @@ Similar Content for Context:
         
         chain = prompt | llm | StrOutputParser()
         state["enhanced"] = chain.invoke({
-            "transcript": state["transcript"],
+            "content": state["content"],
             "similar_content": "\n".join([doc.page_content for doc in similar_docs])
         })
         state["status"] = "✅ Content enhanced"
@@ -176,7 +175,7 @@ Remember: The goal is to make the content more engaging while keeping ALL the or
 def verify_content(state: ProcessState, progress=gr.Progress()) -> ProcessState:
     """Verify the enhanced content against the original using semantic similarity."""
     try:
-        if not state["enhanced"] or not state["transcript"]:
+        if not state["enhanced"] or not state["content"]:
             return state
         
         progress(1.0, desc="Verifying content...")
@@ -195,7 +194,7 @@ def verify_content(state: ProcessState, progress=gr.Progress()) -> ProcessState:
         if similar_docs:
             # Chroma returns a list of Document objects with a score attribute
             # But the default similarity_search does not return scores, so we just check if content is similar
-            similarity_score = 1.0 if similar_docs[0].page_content == state["transcript"] else 0.0
+            similarity_score = 1.0 if similar_docs[0].page_content == state["content"] else 0.0
         
         # Initialize LLM for verification
         llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
@@ -225,7 +224,7 @@ Semantic Similarity Score: {similarity_score}"""),
         
         chain = prompt | llm | StrOutputParser()
         verification_result = json.loads(chain.invoke({
-            "original": state["transcript"],
+            "original": state["content"],
             "enhanced": state["enhanced"],
             "similarity_score": similarity_score
         }))
@@ -262,7 +261,7 @@ def create_workflow() -> StateGraph:
     workflow = StateGraph(ProcessState)
     
     # Add nodes
-    workflow.add_node("get_transcript", get_transcript)
+    workflow.add_node("get_content", get_content)
     workflow.add_node("enhance_content", enhance_content)
     workflow.add_node("format_linkedin", format_linkedin_post)
     workflow.add_node("verify_content", verify_content)
@@ -271,10 +270,10 @@ def create_workflow() -> StateGraph:
     workflow.add_node("enhance_again", enhance_again)
     
     # Set entry point
-    workflow.set_entry_point("get_transcript")
+    workflow.set_entry_point("get_content")
     
     # Add edges for main flow
-    workflow.add_edge("get_transcript", "enhance_content")
+    workflow.add_edge("get_content", "enhance_content")
     workflow.add_edge("enhance_content", "format_linkedin")
     workflow.add_edge("format_linkedin", "verify_content")
     workflow.add_edge("verify_content", "agent_decide")
@@ -295,7 +294,7 @@ def create_workflow() -> StateGraph:
     
     # Add conditional edges for error handling
     workflow.add_conditional_edges(
-        "get_transcript",
+        "get_content",
         should_continue,
         {
             True: "enhance_content",
@@ -419,11 +418,11 @@ def process_from_stage(state: ProcessState, start_stage: str, progress=gr.Progre
         # Select appropriate workflow based on stage
         if start_stage == "enhance":
             workflow = create_workflow()
-            if not state["transcript"]:
+            if not state["content"]:
                 return (
-                    "⚠️ No transcript available to enhance",
-                    "❌ Failed: No transcript",
-                    state.get("transcript", ""),
+                    "⚠️ No content available to enhance",
+                    "❌ Failed: No content",
+                    state.get("content", ""),
                     "",
                     "",
                     ""
@@ -434,7 +433,7 @@ def process_from_stage(state: ProcessState, start_stage: str, progress=gr.Progre
                 return (
                     "⚠️ No enhanced content available to format",
                     "❌ Failed: No enhanced content",
-                    state.get("transcript", ""),
+                    state.get("content", ""),
                     state.get("enhanced", ""),
                     "",
                     ""
@@ -457,7 +456,7 @@ def process_from_stage(state: ProcessState, start_stage: str, progress=gr.Progre
         return (
             final_state.get("error", ""),
             final_state.get("status", ""),
-            final_state.get("transcript", ""),
+            final_state.get("content", ""),
             final_state.get("enhanced", ""),
             final_state.get("linkedin_post", ""),
             verification_text
@@ -467,7 +466,7 @@ def process_from_stage(state: ProcessState, start_stage: str, progress=gr.Progre
         return (
             f"⚠️ Error: {str(e)}",
             "❌ Processing failed",
-            state.get("transcript", ""),
+            state.get("content", ""),
             state.get("enhanced", ""),
             state.get("linkedin_post", ""),
             ""
@@ -555,8 +554,8 @@ def format_research_results(research: dict) -> str:
 def create_ui():
     with gr.Blocks(theme='JohnSmith9982/small_and_pretty') as demo:
         current_state = gr.State({
-            "video_url": "",
-            "transcript": "",
+            "pdf_file": "",
+            "content": "",
             "enhanced": "",
             "linkedin_post": "",
             "verification": {},
@@ -570,33 +569,25 @@ def create_ui():
         
         gr.Markdown(
             """
-            # YouTube to LinkedIn Post Converter
-            Transform your YouTube videos into professional LinkedIn posts with AI content enhancement.
+            # PDF to LinkedIn Post Converter
+            Transform your PDF documents into professional LinkedIn posts with AI content enhancement.
             
-            ### 🎬 Sample Videos to Try
-            Copy any of these URLs to test the application:
-            ```
-            1. Open AI video: https://www.youtube.com/watch?v=LsMxX86mm2Y
-               Agent will likely find high quality initial content and not improve
-            
-            2. Financial News: https://www.youtube.com/watch?v=hvP1UNALZ3g
-               Agent will likely decide to not improve this post
-            
-            3. Video About AI: https://www.youtube.com/watch?v=Yq0QkCxoTHM
-               Agent will likely decide to improve this post
-            ```
-            These videos are chosen to show the application's ability to handle different types of professional content.
+            ### 📄 How to Use
+            1. Upload a PDF file
+            2. Click "Generate Post"
+            3. Review the enhanced content
+            4. Copy your LinkedIn-ready post
             """
         )
         
         with gr.Row():
             with gr.Column():
-                video_url = gr.Textbox(
-                    label="YouTube URL",
-                    placeholder="https://www.youtube.com/watch?v=e1GJ5tZePjk",
-                    show_label=True
+                pdf_file = gr.File(
+                    label="PDF File",
+                    file_types=[".pdf"],
+                    type="filepath"
                 )
-                youtube_convert_btn = gr.Button("🚀 Generate from YouTube", variant="primary", size="lg")
+                convert_btn = gr.Button("🚀 Generate from PDF", variant="primary", size="lg")
 
         status = gr.Textbox(
             label="Status",
@@ -614,8 +605,8 @@ def create_ui():
             with gr.TabItem("📝 Content"):
                 with gr.Row():
                     with gr.Column():
-                        transcript = gr.TextArea(
-                            label="📄 Raw Transcript",
+                        content = gr.TextArea(
+                            label="📄 Raw Content",
                             interactive=False,
                             show_copy_button=True,
                             lines=8
@@ -680,7 +671,7 @@ def create_ui():
 
                 # Loading indicators
                 with gr.Row(visible=False) as loading_indicators:
-                    transcript_loading = gr.Markdown("🔄 Fetching transcript...")
+                    content_loading = gr.Markdown("🔄 Extracting content...")
                     enhanced_loading = gr.Markdown("🔄 Enhancing content...")
                     linkedin_loading = gr.Markdown("🔄 Formatting for LinkedIn...")
                     verify_loading = gr.Markdown("🔄 Verifying content...")
@@ -692,9 +683,9 @@ def create_ui():
                 gr.Markdown(
                     """
                     ### How to Use
-                    1. **Input**: Paste a YouTube video URL in the input field
+                    1. **Input**: Upload a PDF file
                     2. **Process**: Click the "Generate Post" button
-                    3. **Wait**: The system will process your video through multiple steps
+                    3. **Wait**: The system will process your PDF through multiple steps
                     4. **Review**: Check the generated content in each tab
                     5. **Copy**: Use the copy button to grab your LinkedIn post
                     
@@ -703,19 +694,18 @@ def create_ui():
                     - Click 🔄 next to "LinkedIn Post" to regenerate from the formatting stage
                     
                     ### 💡 Tips for Best Results
-                    - Use videos with clear English audio
-                    - Optimal video length: 5-15 minutes
-                    - Ensure videos have accurate captions
+                    - Use well-formatted PDFs with clear text
+                    - Optimal length: 2-10 pages
+                    - Ensure PDFs have readable text (not scanned images)
                     - Review and personalize the post before sharing
-                    - Consider your target audience when selecting videos
-                    
+                    - Consider your target audience when selecting content
                     """
                 )
 
         def update_loading_state(stage: str):
             """Update loading indicators based on current stage."""
             states = {
-                "transcript": [True, False, False, False, False, False, False],
+                "content": [True, False, False, False, False, False, False],
                 "enhance": [False, True, False, False, False, False, False],
                 "format": [False, False, True, False, False, False, False],
                 "verify": [False, False, False, True, False, False, False],
@@ -727,7 +717,7 @@ def create_ui():
             
             # Loading messages for each stage
             loading_messages = {
-                "transcript": "🔄 Fetching transcript...\n⏳ Please wait...",
+                "content": "🔄 Extracting content...\n⏳ Please wait...",
                 "enhance": "✨ Enhancing content...\n⚡ AI is working its magic...",
                 "format": "🎨 Formatting for LinkedIn...\n📝 Creating engaging post...",
                 "verify": "🔍 Verifying content...\n⚖️ Checking accuracy...",
@@ -744,8 +734,8 @@ def create_ui():
                 gr.update(visible=state) for state in states.get(stage, [False] * 7)
             ], current_message
 
-        def process_with_loading(url, state):
-            """Process video with loading indicators."""
+        def process_with_loading(pdf_path, state):
+            """Process PDF with loading indicators."""
             try:
                 # Initialize state if needed
                 if "improvement_plan" not in state:
@@ -758,11 +748,11 @@ def create_ui():
                     state["needs_improvement"] = False
                     
                 # Show loading indicators
-                loading_states, message = update_loading_state("transcript")
+                loading_states, message = update_loading_state("content")
                 yield [
                     "",  # error
                     "Processing...",  # status
-                    message,  # transcript (loading)
+                    message,  # content (loading)
                     "",  # enhanced
                     "",  # linkedin
                     "",  # verification
@@ -773,16 +763,16 @@ def create_ui():
                     *loading_states  # loading indicators
                 ]
                 
-                # Get transcript
-                state["video_url"] = url
-                transcript_text = get_transcript(state)["transcript"]
+                # Get content
+                state["pdf_file"] = pdf_path
+                content_text = get_content(state)["content"]
                 
                 # Show enhancing state
                 loading_states, message = update_loading_state("enhance")
                 yield [
                     "",
                     "Enhancing content...",
-                    transcript_text,
+                    content_text,
                     message,  # enhanced (loading)
                     "",
                     "",
@@ -794,7 +784,7 @@ def create_ui():
                 ]
                 
                 # Enhance content
-                state["transcript"] = transcript_text
+                state["content"] = content_text
                 enhanced_state = enhance_content(state)
                 enhanced_text = enhanced_state["enhanced"]
                 
@@ -803,7 +793,7 @@ def create_ui():
                 yield [
                     "",
                     "Formatting for LinkedIn...",
-                    transcript_text,
+                    content_text,
                     enhanced_text,
                     message,  # linkedin (loading)
                     "",
@@ -824,7 +814,7 @@ def create_ui():
                 yield [
                     "",
                     "Verifying content...",
-                    transcript_text,
+                    content_text,
                     enhanced_text,
                     linkedin_text,
                     "🔍 Verifying...\n⚖️ Analyzing accuracy...",  # verification (loading)
@@ -851,7 +841,7 @@ def create_ui():
                     yield [
                         "",
                         f"Creating improvement plan (Attempt {final_state.get('enhancement_attempts', 1)}/3)...",
-                        transcript_text,
+                        content_text,
                         enhanced_text,
                         linkedin_text,
                         verification_text,
@@ -867,7 +857,7 @@ def create_ui():
                     yield [
                         "",
                         f"Researching content (Attempt {final_state.get('enhancement_attempts', 1)}/3)...",
-                        transcript_text,
+                        content_text,
                         enhanced_text,
                         linkedin_text,
                         verification_text,
@@ -887,7 +877,7 @@ def create_ui():
                     yield [
                         "",
                         f"Enhancing content again (Attempt {final_state.get('enhancement_attempts', 1)}/3)...",
-                        transcript_text,
+                        content_text,
                         enhanced_text,
                         linkedin_text,
                         verification_text,
@@ -921,7 +911,7 @@ def create_ui():
                     yield [
                         "",
                         f"Creating improved LinkedIn post (Attempt {final_state.get('enhancement_attempts', 1)}/3)...",
-                        transcript_text,
+                        content_text,
                         enhanced_text,
                         linkedin_text,
                         verification_text,
@@ -944,7 +934,7 @@ def create_ui():
                 yield [
                     "",
                     "✅ Processing complete!",
-                    transcript_text,
+                    content_text,
                     enhanced_text,
                     linkedin_text,
                     verification_text,
@@ -960,7 +950,7 @@ def create_ui():
                 yield [
                     f"⚠️ Error: {str(e)}",
                     "❌ Processing failed",
-                    state.get("transcript", ""),
+                    state.get("content", ""),
                     state.get("enhanced", ""),
                     state.get("linkedin_post", ""),
                     "",
@@ -972,13 +962,13 @@ def create_ui():
                 ]
 
         # Set up event handlers
-        youtube_convert_btn.click(
+        convert_btn.click(
             fn=process_with_loading,
-            inputs=[video_url, current_state],
+            inputs=[pdf_file, current_state],
             outputs=[
                 error,
                 status,
-                transcript,
+                content,
                 enhanced,
                 linkedin,
                 verification,
@@ -986,7 +976,7 @@ def create_ui():
                 research_results,
                 improved_linkedin,
                 current_state,
-                transcript_loading,
+                content_loading,
                 enhanced_loading,
                 linkedin_loading,
                 verify_loading,
@@ -1011,7 +1001,7 @@ def create_ui():
             return {
                 loading: gr.update(visible=is_loading)
                 for loading in [
-                    transcript_loading,
+                    content_loading,
                     enhanced_loading,
                     linkedin_loading,
                     verify_loading,
@@ -1021,10 +1011,10 @@ def create_ui():
                 ]
             }
 
-        youtube_convert_btn.click(
+        convert_btn.click(
             lambda: update_loading_visibility(True),
             None,
-            [transcript_loading, enhanced_loading, linkedin_loading, 
+            [content_loading, enhanced_loading, linkedin_loading, 
              verify_loading, plan_loading, research_loading, improved_loading],
             queue=False
         )
@@ -1239,11 +1229,4 @@ Important:
 
 if __name__ == "__main__":
     demo = create_ui()
-    demo.queue()  # Enable queuing for better handling of concurrent requests
-    demo.launch(
-        server_name="0.0.0.0",  # Required for Hugging Face Spaces
-        server_port=7860,       # Standard port for Hugging Face Spaces
-        show_error=True,
-        share=False,            # Disable sharing for production
-        show_api=False
-    ) 
+    demo.launch(server_name="0.0.0.0", server_port=7860, share=False) 

requirements.txt

@@ -1,8 +1,9 @@
-gradio>=4.19.2
-python-dotenv>=1.0.1
+gradio>=4.0.0
+python-dotenv>=1.0.0
 youtube-transcript-api>=0.6.2
-langchain-openai>=0.0.8
-langchain>=0.1.9
-langgraph>=0.0.27
-langchain-community>=0.0.27
-langchain-chroma>=0.1.4 
+langchain-openai>=0.0.5
+langchain>=0.1.0
+langgraph>=0.0.20
+langchain-community>=0.0.13
+langchain-chroma>=0.0.5
+pypdf>=3.17.1