Create huggingfaceinferenceclient.py
Browse files- huggingfaceinferenceclient.py +241 -0
huggingfaceinferenceclient.py
ADDED
@@ -0,0 +1,241 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import base64
|
3 |
+
import requests
|
4 |
+
from typing import Dict, Any, Optional
|
5 |
+
from PIL import Image
|
6 |
+
import io
|
7 |
+
|
8 |
+
class HuggingFaceInferenceClient:
|
9 |
+
"""
|
10 |
+
Comprehensive client for interacting with Hugging Face Inference API endpoints.
|
11 |
+
|
12 |
+
## Core Features
|
13 |
+
- Secure API authentication
|
14 |
+
- Flexible image encoding
|
15 |
+
- Advanced error handling
|
16 |
+
- Configurable generation parameters
|
17 |
+
|
18 |
+
## Technical Design Considerations
|
19 |
+
- Environment-based configuration
|
20 |
+
- Type-hinted method signatures
|
21 |
+
- Comprehensive logging and error management
|
22 |
+
"""
|
23 |
+
|
24 |
+
def __init__(
|
25 |
+
self,
|
26 |
+
api_url: Optional[str] = None,
|
27 |
+
api_token: Optional[str] = None
|
28 |
+
):
|
29 |
+
"""
|
30 |
+
Initialize Hugging Face Inference API client.
|
31 |
+
|
32 |
+
Args:
|
33 |
+
api_url (str, optional): Inference endpoint URL
|
34 |
+
api_token (str, optional): Authentication token
|
35 |
+
"""
|
36 |
+
self.api_url = api_url or os.getenv('HF_INFERENCE_ENDPOINT')
|
37 |
+
self.api_token = api_token or os.getenv('HF_API_TOKEN')
|
38 |
+
|
39 |
+
if not self.api_url or not self.api_token:
|
40 |
+
raise ValueError(
|
41 |
+
"Missing Hugging Face Inference endpoint or API token. "
|
42 |
+
"Please provide via parameters or environment variables."
|
43 |
+
)
|
44 |
+
|
45 |
+
def encode_image(
|
46 |
+
self,
|
47 |
+
image_path: str,
|
48 |
+
format: str = 'JPEG'
|
49 |
+
) -> str:
|
50 |
+
"""
|
51 |
+
Encode image to base64 data URI.
|
52 |
+
|
53 |
+
Args:
|
54 |
+
image_path (str): Path to input image
|
55 |
+
format (str): Output image format
|
56 |
+
|
57 |
+
Returns:
|
58 |
+
str: Base64 encoded data URI
|
59 |
+
"""
|
60 |
+
try:
|
61 |
+
with Image.open(image_path) as img:
|
62 |
+
# Ensure RGB compatibility
|
63 |
+
if img.mode != "RGB":
|
64 |
+
img = img.convert("RGB")
|
65 |
+
|
66 |
+
# Convert to byte array
|
67 |
+
img_byte_arr = io.BytesIO()
|
68 |
+
img.save(img_byte_arr, format=format)
|
69 |
+
|
70 |
+
# Encode to base64
|
71 |
+
base64_encoded = base64.b64encode(
|
72 |
+
img_byte_arr.getvalue()
|
73 |
+
).decode('utf-8')
|
74 |
+
|
75 |
+
return f"data:image/{format.lower()};base64,{base64_encoded}"
|
76 |
+
|
77 |
+
except Exception as e:
|
78 |
+
raise ValueError(f"Image encoding failed: {e}")
|
79 |
+
|
80 |
+
def generate_image(
|
81 |
+
self,
|
82 |
+
payload: Dict[str, Any]
|
83 |
+
) -> Dict[str, Any]:
|
84 |
+
"""
|
85 |
+
Execute image generation request.
|
86 |
+
|
87 |
+
Args:
|
88 |
+
payload (Dict): Generation configuration payload
|
89 |
+
|
90 |
+
Returns:
|
91 |
+
Dict: API response containing generation results
|
92 |
+
"""
|
93 |
+
headers = {
|
94 |
+
"Accept": "application/json",
|
95 |
+
"Authorization": f"Bearer {self.api_token}",
|
96 |
+
"Content-Type": "application/json"
|
97 |
+
}
|
98 |
+
|
99 |
+
try:
|
100 |
+
response = requests.post(
|
101 |
+
self.api_url,
|
102 |
+
headers=headers,
|
103 |
+
json=payload
|
104 |
+
)
|
105 |
+
response.raise_for_status()
|
106 |
+
return response.json()
|
107 |
+
|
108 |
+
except requests.RequestException as e:
|
109 |
+
return {
|
110 |
+
"error": f"API request failed: {e}",
|
111 |
+
"status_code": response.status_code if 'response' in locals() else None
|
112 |
+
}
|
113 |
+
|
114 |
+
def save_generated_media(
|
115 |
+
self,
|
116 |
+
response: Dict[str, Any],
|
117 |
+
output_filename: str
|
118 |
+
) -> Optional[str]:
|
119 |
+
"""
|
120 |
+
Save generated media from API response.
|
121 |
+
|
122 |
+
Args:
|
123 |
+
response (Dict): API generation response
|
124 |
+
output_filename (str): Output file path
|
125 |
+
|
126 |
+
Returns:
|
127 |
+
Optional[str]: Path to saved file or None
|
128 |
+
"""
|
129 |
+
media_types = {
|
130 |
+
'image': self._save_image,
|
131 |
+
'video': self._save_video
|
132 |
+
}
|
133 |
+
|
134 |
+
try:
|
135 |
+
# Check for errors
|
136 |
+
if 'error' in response:
|
137 |
+
print(f"Generation Error: {response['error']}")
|
138 |
+
return None
|
139 |
+
|
140 |
+
# Detect media type and save
|
141 |
+
for media_type, save_func in media_types.items():
|
142 |
+
if media_type in response:
|
143 |
+
return save_func(response[media_type], output_filename)
|
144 |
+
|
145 |
+
raise ValueError("No supported media found in response")
|
146 |
+
|
147 |
+
except Exception as e:
|
148 |
+
print(f"Media saving failed: {e}")
|
149 |
+
return None
|
150 |
+
|
151 |
+
def _save_image(
|
152 |
+
self,
|
153 |
+
image_data_uri: str,
|
154 |
+
output_path: str
|
155 |
+
) -> str:
|
156 |
+
"""
|
157 |
+
Save base64 encoded image data.
|
158 |
+
|
159 |
+
Args:
|
160 |
+
image_data_uri (str): Base64 image data URI
|
161 |
+
output_path (str): Output image file path
|
162 |
+
|
163 |
+
Returns:
|
164 |
+
str: Path to saved image
|
165 |
+
"""
|
166 |
+
# Remove data URI prefix
|
167 |
+
base64_data = image_data_uri.split(",")[1]
|
168 |
+
image_data = base64.b64decode(base64_data)
|
169 |
+
|
170 |
+
with open(output_path, "wb") as f:
|
171 |
+
f.write(image_data)
|
172 |
+
|
173 |
+
return output_path
|
174 |
+
|
175 |
+
def _save_video(
|
176 |
+
self,
|
177 |
+
video_data_uri: str,
|
178 |
+
output_path: str
|
179 |
+
) -> str:
|
180 |
+
"""
|
181 |
+
Save base64 encoded video data.
|
182 |
+
|
183 |
+
Args:
|
184 |
+
video_data_uri (str): Base64 video data URI
|
185 |
+
output_path (str): Output video file path
|
186 |
+
|
187 |
+
Returns:
|
188 |
+
str: Path to saved video
|
189 |
+
"""
|
190 |
+
# Remove data URI prefix
|
191 |
+
base64_data = video_data_uri.split(",")[1]
|
192 |
+
video_data = base64.b64decode(base64_data)
|
193 |
+
|
194 |
+
with open(output_path, "wb") as f:
|
195 |
+
f.write(video_data)
|
196 |
+
|
197 |
+
return output_path
|
198 |
+
|
199 |
+
def main():
|
200 |
+
"""
|
201 |
+
Example usage demonstrating client capabilities.
|
202 |
+
"""
|
203 |
+
# Initialize client with endpoint and token
|
204 |
+
client = HuggingFaceInferenceClient(
|
205 |
+
api_url="https://your-endpoint.endpoints.huggingface.cloud",
|
206 |
+
api_token="hf_your_token_here"
|
207 |
+
)
|
208 |
+
|
209 |
+
# Prepare generation payload
|
210 |
+
image_generation_config = {
|
211 |
+
"inputs": {
|
212 |
+
"image": client.encode_image("input_image.jpg"),
|
213 |
+
"prompt": "Enhance and expand the scene creatively"
|
214 |
+
},
|
215 |
+
"parameters": {
|
216 |
+
# Configurable generation parameters
|
217 |
+
"width": 768,
|
218 |
+
"height": 480,
|
219 |
+
"num_frames": 129, # 8*16 + 1
|
220 |
+
"num_inference_steps": 50,
|
221 |
+
"guidance_scale": 4.0,
|
222 |
+
"double_num_frames": True,
|
223 |
+
"fps": 60,
|
224 |
+
"super_resolution": True,
|
225 |
+
"grain_amount": 12
|
226 |
+
}
|
227 |
+
}
|
228 |
+
|
229 |
+
# Generate media
|
230 |
+
generation_output = client.generate_image(image_generation_config)
|
231 |
+
|
232 |
+
# Save generated media
|
233 |
+
output_filename = client.save_generated_media(
|
234 |
+
generation_output,
|
235 |
+
"output_media.mp4"
|
236 |
+
)
|
237 |
+
|
238 |
+
print(f"Media saved to: {output_filename}")
|
239 |
+
|
240 |
+
if __name__ == "__main__":
|
241 |
+
main()
|