Update app.py

app.py

@@ -185,7 +185,7 @@ def convert_to_jpeg_bytes(img_bytes: bytes, base_h: int = 480) -> bytes:
     w = max(1, int(img.width * (base_h / img.height)))
     img = img.resize((w, base_h), Image.LANCZOS)
     buf = BytesIO()
-    img.save(buf, format="JPEG", quality=85)
+    img.save(buf, format="JPEG", quality=90) # Increased quality from 85 to 90
     return buf.getvalue()
 
 def b64_bytes(b: bytes, mime: str = "image/jpeg") -> str:
@@ -238,7 +238,8 @@ def _get_video_info_and_timestamps(media_path: str, sample_count: int) -> Tuple[
             timestamps = [step * (i + 1) for i in range(actual_sample_count)]
 
     if not timestamps:
-        timestamps = [0.5, 1.0, 2.0, 3.0, 4.0][:sample_count]
+        # Fallback for very short videos or if duration couldn't be determined
+        timestamps = [0.5, 1.0, 2.0, 3.0, 4.0, 5.0][:sample_count] # Ensure enough fallback timestamps
 
     return info, timestamps
 
@@ -366,10 +367,9 @@ def upload_file_to_mistral(client: Mistral, path: str, purpose: str = "batch", t
             if progress is not None:
                 progress(0.5 + 0.01 * attempt, desc=f"Uploading file to model service (attempt {attempt+1}/{max_retries})...")
 
-            with open(path, "rb") as fh:
-                # Mistral client expects a file-like object directly for the 'file' parameter.
-                # It will handle filename and mimetype internally.
-                res = client.files.upload(file=fh, purpose=purpose)
+            # CHANGE: Pass the file path (str) directly, allowing the mistralai client
+            # to handle opening the file and inferring filename/mimetype.
+            res = client.files.upload(file=path, purpose=purpose)
             fid = getattr(res, "id", None)
             if not fid:
                 raise RuntimeError(f"Mistral API upload response missing file ID: {res}")
@@ -464,6 +464,7 @@ def analyze_video_cohesive(client: Mistral, video_path: str, prompt: str, progre
         ]
         result = chat_complete(client, VIDEO_MODEL, messages, progress=progress)
 
+        # Always extract frames for gallery, even if full analysis worked
         _, gallery_frame_paths = extract_frames_for_model_and_gallery(
             video_path, sample_count=6, gallery_base_h=1080, model_base_h=1024, progress=progress
         )
@@ -524,21 +525,18 @@ def _convert_video_for_preview_if_needed(path: str) -> str:
         print(f"Error: Could not create temporary file for video conversion from {path}.")
         return path
 
+    audio_codec_args = []
+    video_info = _ffprobe_streams(path)
+    if video_info and any(s.get("codec_type") == "audio" for s in video_info.get("streams", [])):
+        audio_codec_args = ["-c:a", "aac", "-b:a", "128k"]
+
     cmd = [
         FFMPEG_BIN, "-y", "-i", path,
         "-c:v", "libx264", "-preset", "veryfast", "-crf", "28",
-        "-c:a", " " if not _ffprobe_streams(path) or not any(s.get("codec_type") == "audio" for s in _ffprobe_streams(path).get("streams", [])) else "aac",
-        "-b:a", "128k",
+        *audio_codec_args, # Unpack the list
         "-movflags", "+faststart", out_path,
         "-map_metadata", "-1"
     ]
-    # Remove audio options if no audio stream exists to prevent ffmpeg errors
-    if "-c:a" in cmd and cmd[cmd.index("-c:a") + 1] == " ":
-        cmd.pop(cmd.index("-c:a") + 1)
-        cmd.pop(cmd.index("-c:a"))
-        cmd.pop(cmd.index("-b:a") + 1)
-        cmd.pop(cmd.index("-b:a"))
-
 
     try:
         subprocess.run(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, timeout=60)
@@ -636,7 +634,8 @@ def create_demo():
             with gr.Column(scale=1):
                 preview_image = gr.Image(label="Preview Image", type="filepath", elem_classes="preview_media", visible=False)
                 preview_video = gr.Video(label="Preview Video", elem_classes="preview_media", visible=False, format="mp4")
-                screenshot_gallery = gr.Gallery(label="Extracted Screenshots", columns=5, rows=1, height="auto", object_fit="contain", visible=False)
+                # CHANGE: Set columns to 6 to display all 6 extracted frames without scrolling
+                screenshot_gallery = gr.Gallery(label="Extracted Screenshots", columns=6, rows=1, height="auto", object_fit="contain", visible=False)
                 # Initially hidden, will become visible when a preview status is set
                 preview_status_text = gr.Textbox(label="Preview Status", interactive=False, lines=1, value="", visible=False)
             with gr.Column(scale=2):
@@ -737,10 +736,13 @@ def create_demo():
             main_path_clear = ""
             screenshot_paths_clear = []
             raw_media_path_clear = ""
+            progress_markdown_update_clear = gr.update(value="Idle")
+
 
             if not url:
                 return img_update_clear, video_update_clear, gallery_update_clear, \
-                       preview_status_clear, main_path_clear, raw_media_path_clear, screenshot_paths_clear
+                       preview_status_clear, main_path_clear, raw_media_path_clear, \
+                       screenshot_paths_clear, progress_markdown_update_clear
 
             temp_raw_path_for_analysis = ""
             try:
@@ -749,13 +751,15 @@ def create_demo():
                 if not raw_bytes_for_analysis:
                     return img_update_clear, video_update_clear, gallery_update_clear, \
                            gr.update(value="Preview load failed: No media bytes fetched.", visible=True), \
-                           main_path_clear, raw_media_path_clear, screenshot_paths_clear
+                           main_path_clear, raw_media_path_clear, screenshot_paths_clear, \
+                           gr.update(value="Preview load failed (Error)")
 
                 temp_raw_path_for_analysis = _temp_file(raw_bytes_for_analysis, suffix=ext_from_src(url) or ".tmp")
                 if not temp_raw_path_for_analysis:
                     return img_update_clear, video_update_clear, gallery_update_clear, \
                            gr.update(value="Preview load failed: Could not save raw media to temp file.", visible=True), \
-                           main_path_clear, raw_media_path_clear, screenshot_paths_clear
+                           main_path_clear, raw_media_path_clear, screenshot_paths_clear, \
+                           gr.update(value="Preview load failed (Error)")
 
                 progress(0.25, desc="Generating playable preview...")
                 is_img_initial, is_vid_initial = determine_media_type(url)
@@ -768,7 +772,8 @@ def create_demo():
 
                     return img_update_clear, video_update_clear, gallery_update_clear, \
                            gr.update(value="Preview load failed: could not make content playable.", visible=True), \
-                           main_path_clear, raw_media_path_clear, screenshot_paths_clear
+                           main_path_clear, raw_media_path_clear, screenshot_paths_clear, \
+                           gr.update(value="Preview load failed (Error)")
 
                 ext = ext_from_src(local_playable_path)
                 is_img_preview = ext in IMAGE_EXTENSIONS
@@ -777,11 +782,13 @@ def create_demo():
                 if is_img_preview:
                     return gr.update(value=local_playable_path, visible=True), gr.update(value=None, visible=False), \
                            gallery_update_clear, gr.update(value="Image preview loaded.", visible=True), \
-                           local_playable_path, temp_raw_path_for_analysis, screenshot_paths_clear
+                           local_playable_path, temp_raw_path_for_analysis, screenshot_paths_clear, \
+                           gr.update(value="Preview ready")
                 elif is_vid_preview:
                     return gr.update(value=None, visible=False), gr.update(value=local_playable_path, visible=True), \
                            gallery_update_clear, gr.update(value="Video preview loaded.", visible=True), \
-                           local_playable_path, temp_raw_path_for_analysis, screenshot_paths_clear
+                           local_playable_path, temp_raw_path_for_analysis, screenshot_paths_clear, \
+                           gr.update(value="Preview ready")
                 else:
                     _temp_files_to_delete.discard(local_playable_path)
                     try: os.remove(local_playable_path)
@@ -792,7 +799,8 @@ def create_demo():
 
                     return img_update_clear, video_update_clear, gallery_update_clear, \
                            gr.update(value="Preview load failed: unknown playable format.", visible=True), \
-                           main_path_clear, raw_media_path_clear, screenshot_paths_clear
+                           main_path_clear, raw_media_path_clear, screenshot_paths_clear, \
+                           gr.update(value="Preview load failed (Error)")
 
             except Exception as e:
                 if os.path.exists(temp_raw_path_for_analysis):
@@ -802,12 +810,13 @@ def create_demo():
 
                 return img_update_clear, video_update_clear, gallery_update_clear, \
                        gr.update(value=f"Preview load failed: {type(e).__name__}: {e}", visible=True), \
-                       main_path_clear, raw_media_path_clear, screenshot_paths_clear
+                       main_path_clear, raw_media_path_clear, screenshot_paths_clear, \
+                       gr.update(value="Preview load failed (Error)")
 
         url_input.change(
             fn=load_main_preview_and_setup_for_analysis,
             inputs=[url_input, main_preview_path_state, raw_media_path_state, screenshot_paths_state],
-            outputs=[preview_image, preview_video, screenshot_gallery, preview_status_text, main_preview_path_state, raw_media_path_state, screenshot_paths_state]
+            outputs=[preview_image, preview_video, screenshot_gallery, preview_status_text, main_preview_path_state, raw_media_path_state, screenshot_paths_state, progress_markdown] # Added progress_markdown to outputs
         )
 
         def worker(url: str, prompt: str, key: str, raw_media_path: str, progress=gr.Progress()):