Update app.py

app.py

@@ -9,7 +9,7 @@ from flask import Flask, request, jsonify
 from transformers import WhisperProcessor, WhisperForConditionalGeneration
 
 ###############################################################################
-# 1) Configure environment to avoid permission issues & set up model
+# 1) Configure environment & set up model
 ###############################################################################
 os.environ["HF_HOME"] = "/tmp/hf_cache"
 os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf_cache"
@@ -18,7 +18,6 @@ os.environ["XDG_CACHE_HOME"] = "/tmp"
 
 app = Flask(__name__)
 
-# Example: your custom Hebrew model
 model_id = "ivrit-ai/whisper-large-v3-turbo"
 processor = WhisperProcessor.from_pretrained(model_id)
 model = WhisperForConditionalGeneration.from_pretrained(model_id)
@@ -35,26 +34,26 @@ WEBHOOK_URL = "https://hook.eu1.make.com/86zogci73u394k2uqpulp5yjjwgm8b9x"
 ###############################################################################
 # 2) Background transcription function
 ###############################################################################
-def transcribe_in_background(audio_url):
+def transcribe_in_background(audio_url, file_id, company, user):
     """
     Called by a background thread. Downloads & transcribes audio,
     then sends results to your Make.com webhook.
     """
     try:
-        # Download audio
+        # 1) Download the audio
         r = requests.get(audio_url)
         audio_path = "/tmp/temp_audio.wav"
         with open(audio_path, "wb") as f:
             f.write(r.content)
 
-        # Load with librosa
+        # 2) Load with librosa
         waveform, sr = librosa.load(audio_path, sr=16000)
 
-        # Optional limit ~1 hour
+        # Optional: limit ~1 hour
         max_sec = 3600
         waveform = waveform[: sr * max_sec]
 
-        # Split audio into 25-second chunks
+        # 3) Split audio into 25-second chunks
         chunk_sec = 25
         chunk_size = sr * chunk_sec
         chunks = [waveform[i : i + chunk_size] for i in range(0, len(waveform), chunk_size)]
@@ -73,39 +72,56 @@ def transcribe_in_background(audio_url):
             transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
             partial_text += transcription + "\n"
 
-        # Post final transcription back to Make.com
-        payload = {"Transcription": partial_text.strip()}
+        # 4) Post final transcription back to Make.com, including the extra fields
+        payload = {
+            "Transcription": partial_text.strip(),
+            "fileId": file_id,
+            "company": company,
+            "user": user
+        }
         requests.post(WEBHOOK_URL, json=payload)
 
     except Exception as e:
         # In case of errors, notify the webhook
-        error_payload = {"error": str(e)}
+        error_payload = {
+            "error": str(e),
+            "fileId": file_id,
+            "company": company,
+            "user": user
+        }
         requests.post(WEBHOOK_URL, json=error_payload)
 
-
 ###############################################################################
-# 3) Flask route: returns immediately, does the heavy lifting in a thread
+# 3) Flask route: returns immediately, transcribes in a separate thread
 ###############################################################################
 @app.route("/transcribe", methods=["POST"])
 def transcribe_endpoint():
+    # 1) Get JSON data from request
     data = request.get_json()
     audio_url = data.get("audio_url")
     if not audio_url:
         return jsonify({"error": "Missing 'audio_url' in request"}), 400
 
-    # Spawn a thread to handle transcription & webhook
-    thread = threading.Thread(target=transcribe_in_background, args=(audio_url,))
+    # 2) Read custom headers (fileId, company, user)
+    file_id = request.headers.get("fileId", "")
+    company = request.headers.get("company", "")
+    user = request.headers.get("user", "")
+
+    # 3) Spawn a thread to handle transcription
+    thread = threading.Thread(
+        target=transcribe_in_background,
+        args=(audio_url, file_id, company, user)
+    )
     thread.start()
 
-    # Immediately return a JSON response to Make.com
+    # 4) Immediately return a JSON response
     return jsonify({
         "status": "Received. Transcription in progress.",
         "note": "Results will be sent via webhook once done."
     }), 202
 
-
 ###############################################################################
-# 4) Run app if local, else Hugging Face will use gunicorn.
+# 4) Run app if local; on HF Spaces, gunicorn is used
 ###############################################################################
 if __name__ == "__main__":
     app.run(host="0.0.0.0", port=7860)