add caption hastag

app.py

@@ -33,7 +33,7 @@ hybrid_quote_generator = HybridQuoteGenerator(
     openai_client=openai_client,
 )
 
-# Initialize MCP Client (optional)
+# Initialize MCP Client (optional, not critical if missing)
 try:
     mcp_client = MCPClient("https://abidlabs-mcp-tools.hf.space")
     mcp_enabled = True
@@ -220,6 +220,62 @@ def get_trend_insights(niche: str) -> Dict[str, Any]:
     return trends.get(niche, default)
 
 
+# =============================================================================
+# CAPTION + HASHTAG GENERATION (non-MCP version)
+# =============================================================================
+
+def generate_caption_and_hashtags(niche: str, persona: str, trend_label: str) -> str:
+    """
+    Generate a posting-ready caption, hashtags, and a tiny posting tip
+    based on niche + persona + trend theme.
+    """
+    persona_instruction = get_persona_instruction(persona)
+
+    prompt = f"""
+Generate a social-media-ready caption and hashtags for a short vertical quote video.
+
+Niche: {niche}
+Persona / tone: {persona} ({persona_instruction})
+Trend theme: {trend_label}
+
+Requirements:
+- CAPTION: 1–2 sentences max
+  * Should sound natural, like a human writing for TikTok/Instagram
+  * Should NOT repeat the quote text word-for-word
+  * Can reference feelings, situation, or transformation implied by the quote
+- HASHTAGS:
+  * 8–12 hashtags total
+  * Mix of trending-style tags and niche / long-tail tags
+  * Use lowercase, no spaces (standard hashtag conventions)
+  * No banned, misleading, or spammy tags
+- POSTING TIP:
+  * 1 short sentence with a practical suggestion (sound choice, posting time, or CTA)
+
+Format the answer EXACTLY like this:
+
+CAPTION:
+<caption text>
+
+HASHTAGS:
+#tag1 #tag2 #tag3 ...
+
+POSTING TIP:
+<one short tip>
+"""
+
+    try:
+        completion = openai_client.chat.completions.create(
+            model="gpt-4o-mini",
+            messages=[{"role": "user", "content": prompt}],
+            max_tokens=220,
+            temperature=0.8,
+        )
+        text = completion.choices[0].message.content.strip()
+        return text
+    except Exception as e:
+        return f"Error generating caption/hashtags: {str(e)}"
+
+
 # =============================================================================
 # TOOLS
 # =============================================================================
@@ -269,20 +325,6 @@ def search_pexels_video_tool(style: str, niche: str, trend_label: str = "") -> d
     """
     Search and fetch a matching vertical video from Pexels based on style, niche,
     and the current trend label.
-
-    Args:
-        style: Visual style for the background (e.g. Cinematic, Nature, Urban).
-        niche: Content niche (e.g. Motivation, Business/Entrepreneurship, Fitness).
-        trend_label: Short description of the trend theme used to bias search
-            queries (e.g. "soft life vs discipline era").
-
-    Returns:
-        A dictionary with:
-            - success: Whether any suitable video was found.
-            - video_url: Direct URL to an MP4 file (if success is True).
-            - search_query: The text query used for the Pexels search.
-            - pexels_url: Link to the Pexels video page.
-            - error: Error message if success is False.
     """
     base_queries = {
         "Motivation": {
@@ -405,19 +447,6 @@ def create_quote_video_tool(
 ) -> dict:
     """
     Create the final quote video via the Modal web endpoint.
-
-    Args:
-        video_url: Direct URL to the background video file (MP4) from Pexels.
-        quote_text: The text of the quote to render on top of the video.
-        output_path: Local filesystem path where the rendered video will be saved.
-        audio_b64: Optional base64-encoded audio bytes for voice-over narration.
-        text_style: Layout style for the quote text (e.g. 'classic_center').
-
-    Returns:
-        A dictionary with:
-            - success: Whether video rendering succeeded.
-            - output_path: Path to the saved MP4 file if successful, else None.
-            - message: Human-readable status or error message.
     """
     if not MODAL_ENDPOINT_URL:
         return {
@@ -621,7 +650,11 @@ def mcp_agent_pipeline(
     text_style: str,
     voice_profile: str,
     num_variations: int = 1,
-) -> Tuple[str, List[str]]:
+) -> Tuple[str, List[str], str]:
+    """
+    Returns:
+        status_log_str, created_videos, caption_block
+    """
     status_log: List[str] = []
     status_log.append("🤖 **MCP-STYLE AGENT PIPELINE START**\n")
 
@@ -656,7 +689,7 @@ def mcp_agent_pipeline(
     quote = generate_quote_tool(niche, style, persona)
     if quote.startswith("Error"):
         status_log.append(f"   ❌ Quote generation error: {quote}")
-        return "\n".join(status_log), []
+        return "\n".join(status_log), [], ""
 
     preview = quote if len(quote) <= 140 else quote[:140] + "..."
     status_log.append(f"   ✅ Quote: “{preview}”\n")
@@ -695,7 +728,7 @@ def mcp_agent_pipeline(
 
     if not video_results:
         status_log.append("\n❌ No background videos found. Aborting.")
-        return "\n".join(status_log), []
+        return "\n".join(status_log), [], ""
 
     status_log.append("")
 
@@ -741,7 +774,7 @@ def mcp_agent_pipeline(
 
     if not created_videos:
         status_log.append("\n❌ All video renderings failed.")
-        return "\n".join(status_log), []
+        return "\n".join(status_log), [], ""
 
     status_log.append("\n🔗 **Integrations used:**")
     status_log.append("   • Gemini – quote + variety tracking")
@@ -752,10 +785,16 @@ def mcp_agent_pipeline(
     if mcp_enabled:
         status_log.append("   • MCP server – available for extended tools")
 
+    # Step 6 – Caption + hashtags block (returned separately for the UI)
+    status_log.append(
+        "\n📝 **Step 6 – Caption + Hashtags** (see the panel next to your videos to copy-paste)"
+    )
+    caption_block = generate_caption_and_hashtags(niche, persona, trend_label)
+
     status_log.append("\n✨ **Pipeline complete!**")
     status_log.append(f"   Generated {len(created_videos)} video variation(s).")
 
-    return "\n".join(status_log), created_videos
+    return "\n".join(status_log), created_videos, caption_block
 
 
 # =============================================================================
@@ -794,7 +833,7 @@ with gr.Blocks(
     ### MCP-style agent • Gemini + OpenAI + ElevenLabs + Modal
 
     An autonomous mini-studio that generates trend-aware quote videos with voice-over,
-    cinematic stock footage, and MCP-style agent reasoning.
+    cinematic stock footage, and ready-to-post captions + hashtags.
     """
     )
 
@@ -880,11 +919,21 @@ with gr.Blocks(
                 show_label=False,
             )
 
-    gr.Markdown("### ✨ Your Quote Videos (This Run)")
+    gr.Markdown("### ✨ Your Quote Videos & Caption")
+
     with gr.Row():
-        video1 = gr.Video(label="Video 1", height=480)
-        video2 = gr.Video(label="Video 2", height=480)
-        video3 = gr.Video(label="Video 3", height=480)
+        with gr.Column(scale=3):
+            with gr.Row():
+                video1 = gr.Video(label="Video 1", height=420)
+                video2 = gr.Video(label="Video 2", height=420)
+                video3 = gr.Video(label="Video 3", height=420)
+        with gr.Column(scale=2):
+            caption_box = gr.Textbox(
+                label="📄 Caption + Hashtags + Posting Tip",
+                lines=14,
+                show_label=True,
+                interactive=False,
+            )
 
     gr.Markdown(
         """
@@ -892,8 +941,8 @@ with gr.Blocks(
     ### 🧩 Under the hood
     - Context engineering: niche + persona + trend theme
     - Mini-RAG: curated trend knowledge feeding into generation
-    - Hybrid LLM: Gemini (quotes) + OpenAI (commentary)
-    - Multimodal pipeline: text → audio → video
+    - Hybrid LLM: Gemini (quotes) + OpenAI (commentary & captions)
+    - Multimodal pipeline: text → audio → video → posting assets
     """
     )
 
@@ -905,7 +954,7 @@ with gr.Blocks(
         voice_profile_val,
         num_variations_val,
     ):
-        status, videos = mcp_agent_pipeline(
+        status, videos, caption_block = mcp_agent_pipeline(
             niche=niche_val,
             style=style_val,
             persona=persona_val,
@@ -926,7 +975,7 @@ with gr.Blocks(
         g5 = gallery_vids[4] if len(gallery_vids) > 4 else None
         g6 = gallery_vids[5] if len(gallery_vids) > 5 else None
 
-        return status, v1, v2, v3, g1, g2, g3, g4, g5, g6
+        return status, v1, v2, v3, caption_block, g1, g2, g3, g4, g5, g6
 
     generate_btn.click(
         process_and_display,
@@ -943,6 +992,7 @@ with gr.Blocks(
             video1,
             video2,
             video3,
+            caption_box,
             gallery_video1,
             gallery_video2,
             gallery_video3,
@@ -971,5 +1021,4 @@ with gr.Blocks(
     )
 
 if __name__ == "__main__":
-    demo.launch(allowed_paths=["/data/gallery_videos"])
-    
+    demo.launch(allowed_paths=["/data/gallery_videos"])