Merge branch 'gradio-modal-integration' into 'main'

Refactor: Switch to Modal framework for audio processing and storage

See merge request sonne-technology/bsod-tv/waveform-matching-gradio-front-end!1

app.py

@@ -1,8 +1,17 @@
-import gradio as gr
-import requests
+## ENVIRONMENT VARIABLES
+# MODAL_VOLUME
+# MODAL_TOKEN_ID
+# MODAL_ENVIRONMENT
+# MODAL_TOKEN_SECRET
+
+
+import os
+import time
 import wave
+import modal
 import contextlib
-import os
+import gradio as gr
+
 
 def process_audio(original_audio_path, dubbed_audio_path, email, company_name, tolerance):
     """
@@ -10,6 +19,11 @@ def process_audio(original_audio_path, dubbed_audio_path, email, company_name, t
     file upload to presigned URLs, and triggering the processing.
     """
     # 1. Check the duration of both audio files.
+    modal_token_id = os.environ['MODAL_TOKEN_ID']
+    modal_token_secret = os.environ['MODAL_TOKEN_SECRET']
+    modal_environment = os.environ['MODAL_ENVIRONMENT']
+    modal_volume = os.environ['MODAL_VOLUME']
+    processing_id = str(int(time.time()))
     try:
         with contextlib.closing(wave.open(original_audio_path, 'r')) as f:
             frames = f.getnframes()
@@ -26,71 +40,29 @@ def process_audio(original_audio_path, dubbed_audio_path, email, company_name, t
 
     except Exception as e:
         return f"Error reading audio files: {e}"
-
-    # --- ACTION REQUIRED ---
-    # Please replace the following placeholder URLs with your actual API endpoints.
-    presigned_url_endpoint = "https://your-api.com/get-presigned-urls" # TODO: Change URL
-    processing_endpoint = "https://your-api.com/trigger-processing" # TODO: Change URL
-    # --------------------------
-
-    # 2.1. Get presigned URLs from your endpoint. # TODO: Change Payload
-    payload = {
-        "files": [
-            {"name": os.path.basename(original_audio_path), "type": "audio/wav"},
-            {"name": os.path.basename(dubbed_audio_path), "type": "audio/wav"}
-        ]
-    }
-    try:
-        print(f"Requesting presigned URLs from: {presigned_url_endpoint}")
-        response = requests.post(presigned_url_endpoint, json=payload)
-        response.raise_for_status()  # Raise an exception for bad status codes
-        presigned_data = response.json()
-
-        # IMPORTANT: Adjust the following lines based on the actual JSON response
-        # structure of your presigned URL endpoint.
-        # This example assumes a response like:
-        # {"original_url": "...", "dubbed_url": "..."}
-        original_upload_url = presigned_data['original_url']
-        dubbed_upload_url = presigned_data['dubbed_url']
-
-    except requests.exceptions.RequestException as e:
-        return f"Error getting presigned URLs: {e}"
-    except KeyError:
-        return "Error: Could not parse the presigned URL response. Please check the JSON structure."
-
-    # 2.2. Upload the audio files to the presigned URLs. # TODO: Check for PUT accuracy
+    # 2. Upload Audio Files to Modal Storage
     try:
-        print(f"Uploading original file to: {original_upload_url}")
-        with open(original_audio_path, 'rb') as f:
-            upload_response = requests.put(original_upload_url, data=f)
-            upload_response.raise_for_status()
-
-        print(f"Uploading dubbed file to: {dubbed_upload_url}")
-        with open(dubbed_audio_path, 'rb') as f:
-            upload_response = requests.put(dubbed_upload_url, data=f)
-            upload_response.raise_for_status()
-
-    except requests.exceptions.RequestException as e:
-        return f"Error uploading files: {e}"
-
-    # 3. Call the endpoint to trigger the processing.
-    processing_payload = { # TODO: Change Payload
-        "email": email,
-        "company_name": company_name,
-        "tolerance": tolerance,
-        # The keys here ('original_file', 'dubbed_file') should match what your
-        # processing API expects.
-        "original_file": original_upload_url,
-        "dubbed_file": dubbed_upload_url
-    }
+        bsodtv_storage = modal.Volume.from_name(modal_volume)
+        with bsodtv_storage.batch_upload() as batch:
+            batch.put_file(original_audio_path, "/{}/original_audio.wav".format(processing_id))
+            batch.put_file(dubbed_audio_path, "/{}/original_audio.wav".format(processing_id))
+        bsodtv_storage.commit()
+    except:
+        return "Error uploading audio files to Cloud Storage."
+    # 3. Call modal to trigger processing
     try:
-        print(f"Triggering processing at: {processing_endpoint}")
-        processing_response = requests.post(processing_endpoint, json=processing_payload)
-        processing_response.raise_for_status()
-        # 4. Show the response as output.
-        return f"Processing triggered successfully. Server response: {processing_response.text}"
-    except requests.exceptions.RequestException as e:
-        return f"Error triggering processing: {e}"
+        waveform_matching_function = modal.Function.from_name("Waveform-Matching", "reception_handler")
+        waveform_matching_function.spawn(
+            processing_id=processing_id,
+            original_file="/{}/original_audio.wav".format(processing_id),
+            dubbed_file="/{}/original_audio.wav".format(processing_id),
+            email=email,
+            company_name=company_name,
+            tolerance_percentage=tolerance
+        )
+    except:
+        return "Error calling Outpost to trigger processing."
+    return "Processing started. Results will be emailed to you shortly."
 
 
 # Create the Gradio interface for the application.
@@ -112,4 +84,4 @@ demo = gr.Interface(
 if __name__ == "__main__":
     # To run this file locally, you'll need to install gradio and requests:
     # pip install gradio requests
-    demo.launch()
+    demo.launch()