Update app.py

app.py

@@ -76,17 +76,27 @@ config.read('config.txt')
 
 # API Keys
 anthropic_api_key = config.get('API', 'anthropic_api_key', fallback=None)
+logging.debug(f"Loaded Anthropic API Key: {anthropic_api_key}")
+
 cohere_api_key = config.get('API', 'cohere_api_key', fallback=None)
+logging.debug(f"Loaded cohere API Key: {cohere_api_key}")
+
 groq_api_key = config.get('API', 'groq_api_key', fallback=None)
+logging.debug(f"Loaded groq API Key: {groq_api_key}")
+
 openai_api_key = config.get('API', 'openai_api_key', fallback=None)
+logging.debug(f"Loaded openAI Face API Key: {openai_api_key}")
+
 huggingface_api_key = config.get('API', 'huggingface_api_key', fallback=None)
+logging.debug(f"Loaded HuggingFace Face API Key: {huggingface_api_key}")
+
 
 # Models
 anthropic_model = config.get('API', 'anthropic_model', fallback='claude-3-sonnet-20240229')
 cohere_model = config.get('API', 'cohere_model', fallback='command-r-plus')
 groq_model = config.get('API', 'groq_model', fallback='FIXME')
 openai_model = config.get('API', 'openai_model', fallback='gpt-4-turbo')
-huggingface_model = config.get('API', 'huggingface_model', fallback='microsoft/Phi-3-mini-128k-instruct')
+huggingface_model = config.get('API', 'huggingface_model', fallback='CohereForAI/c4ai-command-r-plus')
 
 # Local-Models
 kobold_api_IP = config.get('Local-API', 'kobold_api_IP', fallback='http://127.0.0.1:5000/api/v1/generate')
@@ -756,10 +766,10 @@ def speaker_diarize(video_file_path, segments, embedding_model = "pyannote/embed
 #
 #
 
-# Summarize with OpenAI ChatGPT
 def extract_text_from_segments(segments):
-    logging.debug(f"openai: extracting text from {segments}")
+    logging.debug(f"Main: extracting text from {segments}")
     text = ' '.join([segment['text'] for segment in segments])
+    logging.debug(f"Main: Successfully extracted text from {segments}")
     return text
 
 
@@ -1154,6 +1164,33 @@ def save_summary_to_file(summary, file_path):
 # Only to be used when configured with Gradio for HF Space
 def summarize_with_huggingface(api_key, file_path):
     logging.debug(f"huggingface: Summarization process starting...")
+
+    model = "microsoft/Phi-3-mini-128k-instruct"
+    API_URL = f"https://api-inference.huggingface.co/models/{model}"
+    headers = {"Authorization": f"Bearer {api_key}"}
+
+    with open(file_path, 'r') as file:
+        segments = json.load(file)
+    text = ''.join([segment['text'] for segment in segments])
+
+    # FIXME adjust max_length and min_length as needed
+    data = {
+        "inputs": text,
+        "parameters": {"max_length": 4096, "min_length": 100}
+    }
+
+    for attempt in range(max_retries):
+        response = requests.post(API_URL, headers=headers, json=data)
+        if response.status_code == 200:
+            summary = response.json()[0]['summary_text']
+            return summary, None
+        elif response.status_code == 503:
+            response_data = response.json()
+            wait_time = response_data.get('estimated_time', 10)
+            return None, f"Model is loading, retrying in {int(wait_time)} seconds..."
+        # Sleep before retrying....
+        time.sleep(wait_time)
+
     if api_key == "":
         api_key = os.environ.get("HF_TOKEN")
         print("HUGGINGFACE API KEY CHECK: " + api_key)
@@ -1167,15 +1204,9 @@ def summarize_with_huggingface(api_key, file_path):
 
         api_key = os.environ.get('HF_TOKEN')
         print("HUGGINGFACE API KEY CHECK #2: " + api_key)
-        headers = {
-            "Authorization": f"Bearer {api_key}"
-        }
-        model = "microsoft/Phi-3-mini-128k-instruct"
-        API_URL = f"https://api-inference.huggingface.co/models/{model}"
-        data = {
-            "inputs": text,
-            "parameters": {"max_length": 512, "min_length": 100}  # You can adjust max_length and min_length as needed
-        }
+
+
+
         
         logging.debug("huggingface: Submitting request...")
         response = requests.post(API_URL, headers=headers, json=data)
@@ -1200,6 +1231,16 @@ def summarize_with_huggingface(api_key, file_path):
 
 
 
+def process_text(api_key,text_file):
+    summary,message = summarize_with_huggingface(api_key,text_file)
+    if summary:
+        # Show summary on success
+        return "Summary:",summary 
+    else:
+        # Inform user about load/wait time
+        return "Notice:",message
+
+
 def launch_ui(demo_mode=False):
     def process_transcription(json_data):
         if json_data:
@@ -1207,24 +1248,6 @@ def launch_ui(demo_mode=False):
         else:
             return ""
 
- #    dropdown.change(None, dropdown, None, _js=js)
-#    toggle_dark.click(
-#        None,
-#        _js="""
-#        () => {
-#            document.body.classList.toggle('dark');
-#            document.querySelector('gradio-app').style.backgroundColor = 'var(--color-background-primary)'
-#        }
-#        """,
-#    )
-    
-    inputs = [
-        gr.components.Textbox(label="URL"),
-        gr.components.Number(value=2, label="Number of Speakers"),
-        gr.components.Dropdown(choices=whisper_models, value="small.en", label="Whisper Model"),
-        gr.components.Number(value=0, label="Offset")
-    ]
-
     if not demo_mode:
         inputs.extend([
             gr.components.Dropdown(choices=["huggingface", "openai", "anthropic", "cohere", "groq", "llama", "kobold", "ooba"], value="huggingface", label="API Name"),
@@ -1233,12 +1256,19 @@ def launch_ui(demo_mode=False):
             gr.components.Checkbox(value=False, label="Download Video")
         ])
 
+    inputs = [
+        gr.components.Textbox(label="URL"),
+        gr.components.Number(value=2, label="Number of Speakers"),
+        gr.components.Dropdown(choices=whisper_models, value="small.en", label="Whisper Model"),
+        gr.components.Number(value=0, label="Offset")
+    ]
+
     iface = gr.Interface(
         fn=lambda *args: process_url(*args, demo_mode=demo_mode),
         inputs=inputs,
         outputs=[
             gr.components.Textbox(label="Transcription", value=lambda: "", max_lines=10),
-            gr.components.Textbox(label="Summary"),
+            gr.components.Textbox(label="Summary or Status Message"),
             gr.components.File(label="Download Transcription as JSON"),
             gr.components.File(label="Download Summary as text", visible=lambda summary_file_path: summary_file_path is not None)
         ],
@@ -1415,7 +1445,6 @@ if __name__ == "__main__":
     parser.add_argument('input_path', type=str, help='Path or URL of the video', nargs='?')
     parser.add_argument('-v','--video',  action='store_true', help='Download the video instead of just the audio')
     parser.add_argument('-api', '--api_name', type=str, help='API name for summarization (optional)')
-    parser.add_argument('-key', '--api_key', type=str, help='API key for summarization (optional)')
     parser.add_argument('-ns', '--num_speakers', type=int, default=2, help='Number of speakers (default: 2)')
     parser.add_argument('-wm', '--whisper_model', type=str, default='small.en', help='Whisper model (default: small.en)')
     parser.add_argument('-off', '--offset', type=int, default=0, help='Offset in seconds (default: 0)')