Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Upload folder using huggingface_hub
Browse files- README.md +1 -1
- app.py +24 -17
- requirements.txt +0 -1
README.md
CHANGED
|
@@ -9,7 +9,7 @@ app_file: app.py
|
|
| 9 |
pinned: false
|
| 10 |
license: mit
|
| 11 |
short_description: Gemini understands audio and video!
|
| 12 |
-
tags: [webrtc, websocket, gradio, secret|
|
| 13 |
---
|
| 14 |
|
| 15 |
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
|
| 9 |
pinned: false
|
| 10 |
license: mit
|
| 11 |
short_description: Gemini understands audio and video!
|
| 12 |
+
tags: [webrtc, websocket, gradio, secret|HF_TOKEN secret|GEMINI_API_KEY]
|
| 13 |
---
|
| 14 |
|
| 15 |
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
app.py
CHANGED
|
@@ -5,16 +5,18 @@ import time
|
|
| 5 |
from io import BytesIO
|
| 6 |
|
| 7 |
import gradio as gr
|
| 8 |
-
from gradio.utils import get_space
|
| 9 |
import numpy as np
|
| 10 |
-
|
| 11 |
from dotenv import load_dotenv
|
| 12 |
from fastrtc import (
|
| 13 |
AsyncAudioVideoStreamHandler,
|
| 14 |
Stream,
|
| 15 |
-
get_twilio_turn_credentials,
|
| 16 |
WebRTC,
|
|
|
|
|
|
|
| 17 |
)
|
|
|
|
|
|
|
| 18 |
from PIL import Image
|
| 19 |
|
| 20 |
load_dotenv()
|
|
@@ -44,12 +46,10 @@ class GeminiHandler(AsyncAudioVideoStreamHandler):
|
|
| 44 |
super().__init__(
|
| 45 |
"mono",
|
| 46 |
output_sample_rate=24000,
|
| 47 |
-
output_frame_size=480,
|
| 48 |
input_sample_rate=16000,
|
| 49 |
)
|
| 50 |
self.audio_queue = asyncio.Queue()
|
| 51 |
self.video_queue = asyncio.Queue()
|
| 52 |
-
self.quit = asyncio.Event()
|
| 53 |
self.session = None
|
| 54 |
self.last_frame_time = 0
|
| 55 |
self.quit = asyncio.Event()
|
|
@@ -69,10 +69,14 @@ class GeminiHandler(AsyncAudioVideoStreamHandler):
|
|
| 69 |
print("set session")
|
| 70 |
while not self.quit.is_set():
|
| 71 |
turn = self.session.receive()
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
|
|
|
| 75 |
self.audio_queue.put_nowait(audio)
|
|
|
|
|
|
|
|
|
|
| 76 |
|
| 77 |
async def video_receive(self, frame: np.ndarray):
|
| 78 |
if self.session:
|
|
@@ -87,7 +91,11 @@ class GeminiHandler(AsyncAudioVideoStreamHandler):
|
|
| 87 |
self.video_queue.put_nowait(frame)
|
| 88 |
|
| 89 |
async def video_emit(self):
|
| 90 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 91 |
|
| 92 |
async def receive(self, frame: tuple[int, np.ndarray]) -> None:
|
| 93 |
_, array = frame
|
|
@@ -97,13 +105,14 @@ class GeminiHandler(AsyncAudioVideoStreamHandler):
|
|
| 97 |
await self.session.send(input=audio_message)
|
| 98 |
|
| 99 |
async def emit(self):
|
| 100 |
-
array = await self.audio_queue
|
| 101 |
-
|
|
|
|
| 102 |
|
| 103 |
async def shutdown(self) -> None:
|
| 104 |
if self.session:
|
| 105 |
self.quit.set()
|
| 106 |
-
await self.session.
|
| 107 |
self.quit.clear()
|
| 108 |
|
| 109 |
|
|
@@ -111,9 +120,7 @@ stream = Stream(
|
|
| 111 |
handler=GeminiHandler(),
|
| 112 |
modality="audio-video",
|
| 113 |
mode="send-receive",
|
| 114 |
-
rtc_configuration=
|
| 115 |
-
if get_space()
|
| 116 |
-
else None,
|
| 117 |
time_limit=90 if get_space() else None,
|
| 118 |
additional_inputs=[
|
| 119 |
gr.Image(label="Image", type="numpy", sources=["upload", "clipboard"])
|
|
@@ -153,8 +160,8 @@ with gr.Blocks(css=css) as demo:
|
|
| 153 |
modality="audio-video",
|
| 154 |
mode="send-receive",
|
| 155 |
elem_id="video-source",
|
| 156 |
-
rtc_configuration=
|
| 157 |
-
if get_space()
|
| 158 |
else None,
|
| 159 |
icon="https://www.gstatic.com/lamda/images/gemini_favicon_f069958c85030456e93de685481c559f160ea06b.png",
|
| 160 |
pulse_color="rgb(255, 255, 255)",
|
|
|
|
| 5 |
from io import BytesIO
|
| 6 |
|
| 7 |
import gradio as gr
|
|
|
|
| 8 |
import numpy as np
|
| 9 |
+
import websockets
|
| 10 |
from dotenv import load_dotenv
|
| 11 |
from fastrtc import (
|
| 12 |
AsyncAudioVideoStreamHandler,
|
| 13 |
Stream,
|
|
|
|
| 14 |
WebRTC,
|
| 15 |
+
get_turn_credentials_async,
|
| 16 |
+
wait_for_item,
|
| 17 |
)
|
| 18 |
+
from google import genai
|
| 19 |
+
from gradio.utils import get_space
|
| 20 |
from PIL import Image
|
| 21 |
|
| 22 |
load_dotenv()
|
|
|
|
| 46 |
super().__init__(
|
| 47 |
"mono",
|
| 48 |
output_sample_rate=24000,
|
|
|
|
| 49 |
input_sample_rate=16000,
|
| 50 |
)
|
| 51 |
self.audio_queue = asyncio.Queue()
|
| 52 |
self.video_queue = asyncio.Queue()
|
|
|
|
| 53 |
self.session = None
|
| 54 |
self.last_frame_time = 0
|
| 55 |
self.quit = asyncio.Event()
|
|
|
|
| 69 |
print("set session")
|
| 70 |
while not self.quit.is_set():
|
| 71 |
turn = self.session.receive()
|
| 72 |
+
try:
|
| 73 |
+
async for response in turn:
|
| 74 |
+
if data := response.data:
|
| 75 |
+
audio = np.frombuffer(data, dtype=np.int16).reshape(1, -1)
|
| 76 |
self.audio_queue.put_nowait(audio)
|
| 77 |
+
except websockets.exceptions.ConnectionClosedOK:
|
| 78 |
+
print("connection closed")
|
| 79 |
+
break
|
| 80 |
|
| 81 |
async def video_receive(self, frame: np.ndarray):
|
| 82 |
if self.session:
|
|
|
|
| 91 |
self.video_queue.put_nowait(frame)
|
| 92 |
|
| 93 |
async def video_emit(self):
|
| 94 |
+
frame = await wait_for_item(self.video_queue)
|
| 95 |
+
if frame is not None:
|
| 96 |
+
return frame
|
| 97 |
+
else:
|
| 98 |
+
return np.zeros((100, 100, 3), dtype=np.uint8)
|
| 99 |
|
| 100 |
async def receive(self, frame: tuple[int, np.ndarray]) -> None:
|
| 101 |
_, array = frame
|
|
|
|
| 105 |
await self.session.send(input=audio_message)
|
| 106 |
|
| 107 |
async def emit(self):
|
| 108 |
+
array = await wait_for_item(self.audio_queue)
|
| 109 |
+
if array is not None:
|
| 110 |
+
return (self.output_sample_rate, array)
|
| 111 |
|
| 112 |
async def shutdown(self) -> None:
|
| 113 |
if self.session:
|
| 114 |
self.quit.set()
|
| 115 |
+
await self.session.close()
|
| 116 |
self.quit.clear()
|
| 117 |
|
| 118 |
|
|
|
|
| 120 |
handler=GeminiHandler(),
|
| 121 |
modality="audio-video",
|
| 122 |
mode="send-receive",
|
| 123 |
+
rtc_configuration=get_turn_credentials_async if get_space() == "spaces" else None,
|
|
|
|
|
|
|
| 124 |
time_limit=90 if get_space() else None,
|
| 125 |
additional_inputs=[
|
| 126 |
gr.Image(label="Image", type="numpy", sources=["upload", "clipboard"])
|
|
|
|
| 160 |
modality="audio-video",
|
| 161 |
mode="send-receive",
|
| 162 |
elem_id="video-source",
|
| 163 |
+
rtc_configuration=get_turn_credentials_async
|
| 164 |
+
if get_space() == "spaces"
|
| 165 |
else None,
|
| 166 |
icon="https://www.gstatic.com/lamda/images/gemini_favicon_f069958c85030456e93de685481c559f160ea06b.png",
|
| 167 |
pulse_color="rgb(255, 255, 255)",
|
requirements.txt
CHANGED
|
@@ -2,4 +2,3 @@ fastrtc
|
|
| 2 |
python-dotenv
|
| 3 |
google-genai
|
| 4 |
twilio
|
| 5 |
-
pydantic==2.10.0
|
|
|
|
| 2 |
python-dotenv
|
| 3 |
google-genai
|
| 4 |
twilio
|
|
|