Update app.py

app.py

@@ -1,4 +1,5 @@
 import base64
+import os
 from functools import partial
 from multiprocessing import Pool
 
@@ -14,7 +15,8 @@ title = "Whisper JAX: The Fastest Whisper API ⚡️"
 
 description = "Whisper JAX is an optimised implementation of the [Whisper model](https://huggingface.co/openai/whisper-large-v2) by OpenAI. It runs on JAX with a TPU v4-8 in the backend. Compared to PyTorch on an A100 GPU, it is over **12x** faster, making it the fastest Whisper API available."
 
-API_URL = "https://whisper-jax.ngrok.io/generate/"
+API_URL = os.getenv("API_URL")
+API_URL_FROM_FEATURES = os.getenv("API_URL_FROM_FEATURES")
 
 article = "Whisper large-v2 model by OpenAI. Backend running JAX on a TPU v4-8 through the generous support of the [TRC](https://sites.research.google/trc/about/) programme. Whisper JAX code and Gradio demo by 🤗 Hugging Face."
 
@@ -52,7 +54,7 @@ def inference(inputs, language=None, task=None, return_timestamps=False):
 
 
 def chunked_query(payload):
-    response = requests.post("https://whisper-jax.ngrok.io/generate_from_features", json=payload)
+    response = requests.post(API_URL_FROM_FEATURES, json=payload)
     return response.json()