Update app.py

app.py

@@ -7,126 +7,172 @@ from fpdf import FPDF
 from youtube_comment_downloader import YoutubeCommentDownloader
 import re
 import os
-
-GROQ_API_KEY = os.getenv("translapikey") or os.getenv("GROQ_API_KEY")
-
+import warnings
+
+warnings.filterwarnings("ignore")
+
+# ====================== CONFIG ======================
+# On Hugging Face, set 'GROQ_API_KEY' in the "Variables and Secrets" settings tab
+GROQ_API_KEY = os.getenv("GROQ_API_KEY")
+
+# ====================== SYSTEM PROMPT ======================
+SYSTEM_PROMPT = """
+You are an expert social media sentiment and poll analysis AI.
+Focus on Yes/No, Agree/Disagree, Support/Oppose, and sentiment.
+
+Handle English + Urdu + Hindi + other languages well.
+Return ONLY valid JSON in this exact format:
+{
+  "main_poll": {
+    "question": "Suggested poll question",
+    "yes_count": int,
+    "no_count": int,
+    "agree_count": int,
+    "disagree_count": int,
+    "support_count": int,
+    "oppose_count": int,
+    "neutral_count": int
+  },
+  "sentiment": {
+    "positive": float,
+    "negative": float,
+    "neutral": float
+  },
+  "top_themes": ["theme1", "theme2"],
+  "summary": "Short professional summary",
+  "labeled_comments": [
+    {"comment": "...", "opinion": "Yes|No|Agree|Disagree|Positive|Negative|Neutral|Mixed"}
+  ]
+}
+"""
+
+# ====================== HELPERS ======================
 def clean_text(text):
     if not text: return ""
-    text = re.sub(r'[\u2022\u2023\u25CF]', '-', text)
-    return text.strip()
-
-def fetch_youtube_comments(url):
+    text = re.sub(r'[\u2022\u2023\u25CF\u25BA\u25C4]', '-', text)
+    text = re.sub(r'[\u2018\u2019\u201C\u201D]', '"', text)
+    text = re.sub(r'[\u2013\u2014]', '-', text)
+    text = re.sub(r'[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]', '', text)
+    return text.encode('latin-1', 'ignore').decode('latin-1') # Ensure PDF compatibility
+
+def extract_youtube_id(url):
+    patterns = [r'youtu\.be/([a-zA-Z0-9_-]+)', r'v=([a-zA-Z0-9_-]+)', r'/embed/([a-zA-Z0-9_-]+)', r'/shorts/([a-zA-Z0-9_-]+)']
+    for p in patterns:
+        match = re.search(p, url)
+        if match: return match.group(1)
+    return None
+
+def fetch_youtube_comments(url, limit=100):
     try:
-        print(f"Trying to fetch: {url}")
-        
-        # Try to extract video ID
-        match = re.search(r'(?:v=|youtu\.be/|embed/|shorts/)([a-zA-Z0-9_-]+)', url)
-        if match:
-            video_id = match.group(1)
-            url = f"https://www.youtube.com/watch?v={video_id}"
-        
+        video_id = extract_youtube_id(url)
+        if not video_id: return []
         downloader = YoutubeCommentDownloader()
         comments = []
-        for comment in downloader.get_comments_from_url(url, sort_by=0):
+        # sort_by=0 is "Newest", 1 is "Top"
+        gen = downloader.get_comments(video_id, sort_by=1) 
+        for comment in gen:
             comments.append(comment['text'])
-            if len(comments) >= 80:  
-                break
-        print(f"✅ Fetched {len(comments)} comments")
+            if len(comments) >= limit: break
         return comments
     except Exception as e:
-        print(f"❌ Fetch Error: {str(e)}")
+        print(f"Fetch error: {e}")
         return []
 
-def analyze_comments_with_groq(comments):
+def analyze_comments_with_groq(comments, post_context=""):
+    if not GROQ_API_KEY: return None
     try:
         client = Groq(api_key=GROQ_API_KEY)
-        text = "\n\n".join([f"Comment: {clean_text(c)[:200]}" for c in comments[:100]])
+        comments_text = "\n\n".join([f"C{i+1}: {clean_text(c)[:200]}" for i, c in enumerate(comments)])
         
         response = client.chat.completions.create(
             model="llama-3.3-70b-versatile",
             messages=[
-                {"role": "system", "content": "Analyze these YouTube comments and create a poll with sentiment."},
-                {"role": "user", "content": text}
+                {"role": "system", "content": SYSTEM_PROMPT},
+                {"role": "user", "content": f"Context: {post_context}\n\nComments:\n{comments_text}"}
             ],
-            temperature=0.3,
+            temperature=0.2,
             response_format={"type": "json_object"}
         )
         return json.loads(response.choices[0].message.content)
     except Exception as e:
-        print("Groq Error:", str(e))
+        print(f"Groq Error: {e}")
         return None
 
-def create_pdf_report(result):
+def create_pdf_report(analysis_result, poll_question):
     pdf = FPDF()
     pdf.add_page()
     pdf.set_font('Arial', 'B', 16)
     pdf.cell(0, 10, 'CommentSurvey AI Report', 0, 1, 'C')
     pdf.ln(10)
-    pdf.set_font('Arial', '', 12)
-    pdf.multi_cell(0, 8, clean_text(result.get('summary', 'Analysis Completed')))
-    pdf.output("report.pdf")
-    return "report.pdf"
+    
+    pdf.set_font('Arial', 'B', 12)
+    pdf.cell(0, 10, f"Question: {poll_question[:60]}", 0, 1, 'L')
+    
+    pdf.set_font('Arial', '', 11)
+    summary = analysis_result.get('summary', 'N/A')
+    pdf.multi_cell(0, 7, clean_text(summary))
+    
+    path = "report.pdf"
+    pdf.output(path)
+    return path
 
+# ====================== LOGIC ======================
 def analyze(url):
     if not GROQ_API_KEY:
-        return None, "**❌ API Key missing**", None, None, None, None
+        return None, "❌ API Key Missing in Hugging Face Secrets", None, None, None, None
     
     comments = fetch_youtube_comments(url)
+    if not comments:
+        return None, "❌ Failed to fetch comments.", None, None, None, None
     
-    if len(comments) < 5:
-        return None, """**⚠️ Could not fetch comments automatically.**
-
-**Reasons & Solutions:**
-• Video may have few comments
-• Video is age-restricted or private
-• YouTube blocked our server
-
-**Try these public videos:**
-• https://youtu.be/dQw4w9WgXcQ
-• https://youtu.be/3JZ_3t5fQ8M
-• Or paste comments manually (temporary)""", None, None, None, None
-
     result = analyze_comments_with_groq(comments)
-    
     if not result:
-        return None, "**❌ AI Analysis failed**", None, None, None, None
-
-    fig_poll = px.pie(names=['Yes', 'No', 'Neutral'], values=[45, 35, 20], hole=0.4)
-    fig_sent = px.bar(x=['Positive','Negative','Neutral'], y=[50,30,20])
-
-    pdf_path = create_pdf_report(result)
+        return None, "❌ AI Analysis failed.", None, None, None, None
 
-    return result, f"**✅ Success!** Analyzed {len(comments)} comments", fig_poll, fig_sent, result.get('summary', 'Done'), pdf_path
+    main = result.get('main_poll', {})
+    poll_values = [
+        main.get('yes_count',0) + main.get('agree_count',0) + main.get('support_count',0),
+        main.get('no_count',0) + main.get('disagree_count',0) + main.get('oppose_count',0),
+        main.get('neutral_count',0)
+    ]
 
-with gr.Blocks(title="CommentSurvey AI", theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# 📊 CommentSurvey AI\n**YouTube Comments → Smart Survey**")
-
-    url_input = gr.Textbox(
-        label="🌐 Paste YouTube Video Link",
-        placeholder="https://youtu.be/dQw4w9WgXcQ",
-        lines=1
-    )
-    analyze_btn = gr.Button("🚀 Analyze Comments", variant="primary", size="large")
+    fig_poll = px.pie(names=['Yes/Agree/Support', 'No/Disagree/Oppose', 'Neutral'], 
+                      values=poll_values, title="Poll Distribution", hole=0.4)
+    
+    sent = result.get('sentiment', {})
+    fig_sent = px.bar(x=['Positive', 'Negative', 'Neutral'], 
+                      y=[sent.get('positive',0), sent.get('negative',0), sent.get('neutral',0)], 
+                      title="Sentiment Score", color=['Positive', 'Negative', 'Neutral'])
 
-    status = gr.Markdown("**Status:** Ready")
+    df = pd.DataFrame(result.get('labeled_comments', []))
+    pdf_path = create_pdf_report(result, main.get('question', 'Analysis'))
+    
+    summary_md = f"### 📝 {main.get('question', 'Analysis')}\n{result.get('summary', '')}"
+    
+    return df, "✅ Analysis Complete", fig_poll, fig_sent, summary_md, pdf_path
 
+# ====================== UI ======================
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 📊 CommentSurvey AI")
+    
+    with gr.Row():
+        url_input = gr.Textbox(label="YouTube URL", placeholder="https://www.youtube.com/watch?v=...")
+        btn = gr.Button("Analyze", variant="primary")
+    
+    status = gr.Markdown("Status: Ready")
+    
     with gr.Tabs():
-        with gr.Tab("📊 Poll Results"):
-            poll_plot = gr.Plot()
-        with gr.Tab("Sentiment"):
-            sentiment_plot = gr.Plot()
-        with gr.Tab("Insights"):
-            insights = gr.Markdown()
-        with gr.Tab("Raw Comments"):
-            raw = gr.Dataframe()
-
-    download = gr.File(label="Download PDF")
-
-    analyze_btn.click(
-        fn=analyze,
-        inputs=url_input,
-        outputs=[gr.State(), status, poll_plot, sentiment_plot, insights, download]
-    )
+        with gr.Tab("Summary"):
+            sum_md = gr.Markdown()
+            with gr.Row():
+                p1 = gr.Plot()
+                p2 = gr.Plot()
+        with gr.Tab("Data"):
+            table = gr.Dataframe()
+            
+    report_file = gr.File(label="Download PDF Report")
+
+    btn.click(analyze, inputs=[url_input], outputs=[table, status, p1, p2, sum_md, report_file])
 
 if __name__ == "__main__":
     demo.launch()