Update app.py

app.py

@@ -1,13 +1,11 @@
 #!/usr/bin/env python
 # coding: utf-8
 
-# In[7]:
-
-
 import gradio as gr
 import pandas as pd
 from sentence_transformers import SentenceTransformer, util
 import torch
+import openai # New import for Whisper API
 
 # 載入語義搜索模型
 model_checkpoint = "sickcell69/cti-semantic-search-minilm"
@@ -20,7 +18,7 @@ data = pd.read_json(data_path)
 
 # 載入嵌入文件
 embeddings_path = 'corpus_embeddings.pt'
-corpus_embeddings = torch.load(embeddings_path, map_location=torch.device('cpu')))
+corpus_embeddings = torch.load(embeddings_path)
 
 def semantic_search(query):
     query_embedding = model.encode(query, convert_to_tensor=True)
@@ -33,21 +31,28 @@ def semantic_search(query):
     
     return "\n".join(results)
 
+# New function to transcribe audio using Whisper API
+def transcribe_audio(audio_file):
+    audio_bytes = audio_file.read()
+    response = openai.Audio.transcribe("whisper-1", audio_bytes)
+    return response['text']
+
+# Modified interface to include audio input
 iface = gr.Interface(
     fn=semantic_search,
-    inputs="text",
+    inputs=["text", "file"], # Add audio file input
     outputs="text",
     title="語義搜索應用",
-    description="輸入一個查詢，然後模型將返回最相似的結果。"
+    description="輸入一個查詢或上傳一個音頻文件，然後模型將返回最相似的結果。",
+    examples=["example_audio.wav"] # Example audio file
 )
 
+# New function to handle both text and audio inputs
+def handle_input(input_text, audio_file):
+    if audio_file is not None:
+        input_text = transcribe_audio(audio_file)
+    return semantic_search(input_text)
+
 if __name__ == "__main__":
     #iface.launch()
     iface.launch(share=True) #網頁跑不出來
-
-
-# In[ ]:
-
-
-
-