added generation audio

db/audio_db/is3/is3.py

@@ -23,6 +23,7 @@ class UploadedObject(BaseModel):
     obj_id: ObjectId
     deletehash: str
     cached_obj: Any = None
+    link: str
 
     def __getstate__(self):
         d = super().__getstate__()
@@ -63,13 +64,14 @@ class StagedObject(BaseModel):
         during the same runtime does not need to download the object.
         """
         async with Imgur() as imgur:
-            oid, delete = await imgur.upload_image(self.image())
+            oid, delete, link = await imgur.upload_image(self.image())
         
         return UploadedObject(
             name=self.name,
             obj_id=oid,
             deletehash=delete,
-            cached_obj=self.obj
+            cached_obj=self.obj,
+            link=link
         )
 
 
@@ -122,6 +124,8 @@ class Bucket:
                 '\n'.join(o.name for o in self.pending.values())
             )
             raise Warning(msg)
+
+        return uploaded
     
     async def get_obj(self, name: str) -> Any:
         return await self.uploaded[name].download()

db/audio_db/is3/wrapper.py

@@ -11,7 +11,6 @@ from typing import Optional, Union, Tuple
 from .utils import image_to_b64_string, bytes_to_image
 
 dotenv.load_dotenv()
-AUTH_HEADER = {'Authorization': f"Client-ID {os.getenv('IS3_CLIENT_ID')}"}
 API_ENDPOINTS = {
     'upload': 'https://api.imgur.com/3/upload/',
     'download': 'http://i.imgur.com/',
@@ -26,7 +25,6 @@ async def get_tokens():
     r = await session.request(
             method='post',
             url=API_ENDPOINTS['auth'],
-            headers=AUTH_HEADER,
             data={
                 'refresh_token': os.getenv("IS3_REFRESH_TOKEN"),
                 'client_id': os.getenv("IS3_CLIENT_ID"),
@@ -38,6 +36,8 @@ async def get_tokens():
     return r['access_token'], r['refresh_token']
 
 ACCESS_TOKEN, REFRESH_TOKEN = asyncio.run(get_tokens())
+# AUTH_HEADER = {'Authorization': f"Client-ID {os.getenv('IS3_CLIENT_ID')}"}
+AUTH_HEADER = {'Authorization': f"Bearer {ACCESS_TOKEN}"}
 
 
 
@@ -73,7 +73,7 @@ class ImgurClient:
             headers=AUTH_HEADER, 
             data={'image': data, 'type': 'base64'}
         )
-        return r['id'], r['deletehash']
+        return r['id'], r['deletehash'], r['link']
     
     async def download_image(self, image_id: str) -> Image.Image:
         """Download the image and return the data as bytes."""
@@ -86,12 +86,3 @@ class ImgurClient:
         """Delete an image using a deletehash string"""
         url = API_ENDPOINTS['delete'] + deletehash
         await self._request('delete', url, headers=AUTH_HEADER)
-
-
-async def get_token():
-    im = ImgurClient()
-    await im.get_access_token()
-
-import asyncio
-loop = asyncio.get_event_loop()
-loop.run_until_complete(get_token())

db/generate_audio.py

@@ -1,13 +1,17 @@
-from sentence_transformers import SentenceTransformer
+import io
+
 from pymilvus import Collection
 import asyncio
-
+from bark import SAMPLE_RATE, generate_audio, preload_models
+from scipy.io.wavfile import write
 from db_connect import connect
 
 # for audio storage
-from audio_db.is3 import is3
+from audio_db.is3.is3 import StagedObject
 
 connect()
+preload_models()
+
 
 async def generate_audio():
     response = Collection("Response")
@@ -15,7 +19,8 @@ async def generate_audio():
 
     data = []
 
-    response_iterator = response.query_iterator(limit=1, output_fields=['text', 'embeddings'])
+    response_iterator = response.query_iterator(batch_size=1, output_fields=['text', 'embeddings'])
+    ids_to_delete = []
     while True:
         res = response_iterator.next()
         if len(res) == 0:
@@ -24,22 +29,27 @@ async def generate_audio():
             response_iterator.close()
             break
 
-        bucket = is3.Bucket(str(res[0]['id']))
-
         # generate audio
-        audio_bytes = open('445766006129375465.wav', 'rb').read()
+        audio_array = generate_audio(res[0]['text'], history_prompt="en_speaker_3")
+        bytes_io = io.BytesIO()
+        write(bytes_io, SAMPLE_RATE, audio_array)
+        audio_bytes = bytes_io.read()
 
         # store the audio
-        bucket.stage_obj(audio_bytes, 'audio')
-        await bucket.commit()
+        obj = StagedObject(obj=audio_bytes, name='audio')
+        uploaded_object = await obj.upload()
 
         # save the audio record to AudioResponse
-        data.append([res[0]['text'], str(res[0]['id']), res[0]['embeddings']])
+        data.append([res[0]['text'], uploaded_object.obj_id, res[0]['embeddings']])
+        ids_to_delete.append(res[0]['id'])
 
     audio_response.insert(list(zip(*data)))
     audio_response.flush()
 
+    # delete text to generate audio
+    response.delete(expr=f"id in {str(ids_to_delete)}")
+
 
 if __name__ == '__main__':
-    loop = asyncio.get_event_loop()
-    loop.run_until_complete(generate_audio())
+    loop = asyncio.new_event_loop()
+    loop.run_until_complete(generate_audio())