Spaces:
Running
Running
Upload 10 files
Browse files- .gitignore +39 -0
- .huggingface-space +2 -2
- README.md +7 -30
- app.py +35 -157
- app_hf.py +20 -156
- example_client.py +73 -71
- requirements.txt +7 -9
- requirements_hf.txt +2 -3
- test_local.py +40 -40
.gitignore
ADDED
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Python
|
2 |
+
__pycache__/
|
3 |
+
*.py[cod]
|
4 |
+
*$py.class
|
5 |
+
*.so
|
6 |
+
.Python
|
7 |
+
build/
|
8 |
+
develop-eggs/
|
9 |
+
dist/
|
10 |
+
downloads/
|
11 |
+
eggs/
|
12 |
+
.eggs/
|
13 |
+
lib/
|
14 |
+
lib64/
|
15 |
+
parts/
|
16 |
+
sdist/
|
17 |
+
var/
|
18 |
+
wheels/
|
19 |
+
*.egg-info/
|
20 |
+
.installed.cfg
|
21 |
+
*.egg
|
22 |
+
|
23 |
+
# Virtual Environment
|
24 |
+
venv/
|
25 |
+
env/
|
26 |
+
ENV/
|
27 |
+
|
28 |
+
# IDE
|
29 |
+
.idea/
|
30 |
+
.vscode/
|
31 |
+
*.swp
|
32 |
+
*.swo
|
33 |
+
|
34 |
+
# OS
|
35 |
+
.DS_Store
|
36 |
+
Thumbs.db
|
37 |
+
|
38 |
+
# Logs
|
39 |
+
*.log
|
.huggingface-space
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
-
title: Nomic Vision Embedding
|
2 |
emoji: 🖼️
|
3 |
colorFrom: blue
|
4 |
colorTo: indigo
|
5 |
sdk: gradio
|
6 |
-
sdk_version:
|
7 |
app_file: app_hf.py
|
8 |
pinned: false
|
9 |
license: mit
|
|
|
1 |
+
title: Nomic Vision Embedding Model
|
2 |
emoji: 🖼️
|
3 |
colorFrom: blue
|
4 |
colorTo: indigo
|
5 |
sdk: gradio
|
6 |
+
sdk_version: 4.19.0
|
7 |
app_file: app_hf.py
|
8 |
pinned: false
|
9 |
license: mit
|
README.md
CHANGED
@@ -1,39 +1,16 @@
|
|
1 |
-
|
2 |
-
title: Nomic MCP Tool
|
3 |
-
emoji: 🗂️
|
4 |
-
colorFrom: indigo
|
5 |
-
colorTo: pink
|
6 |
-
sdk: gradio
|
7 |
-
sdk_version: "5.26.0"
|
8 |
-
app_file: app.py
|
9 |
-
pinned: false
|
10 |
-
---
|
11 |
-
# Nomic Vision Embedding MCP Server
|
12 |
|
13 |
-
This
|
14 |
|
15 |
## Features
|
16 |
|
17 |
- Generate embeddings for images using the nomic-ai/nomic-embed-vision-v1.5 model
|
18 |
-
-
|
19 |
-
-
|
20 |
|
21 |
## How It Works
|
22 |
|
23 |
-
|
24 |
-
|
25 |
-
1. **Web Interface**: A Gradio UI that allows users to upload images and view the generated embeddings
|
26 |
-
2. **MCP Interface**: An implementation of the Model Context Protocol that exposes the embedding functionality as a tool
|
27 |
-
|
28 |
-
## MCP Tool
|
29 |
-
|
30 |
-
The server exposes the following MCP tool:
|
31 |
-
|
32 |
-
- **embed_image**: Generate embeddings for an image
|
33 |
-
- Input:
|
34 |
-
- `image_url`: URL of the image to embed, OR
|
35 |
-
- `image_data`: Base64-encoded image data
|
36 |
-
- Output: JSON object containing the embedding vector and its dimension
|
37 |
|
38 |
## Deployment
|
39 |
|
@@ -55,12 +32,12 @@ To run this application locally:
|
|
55 |
## Requirements
|
56 |
|
57 |
- Python 3.7+
|
58 |
-
- Gradio 4.0+
|
59 |
- Transformers
|
60 |
- PyTorch
|
61 |
- Pillow
|
62 |
- NumPy
|
63 |
-
-
|
64 |
|
65 |
## License
|
66 |
|
|
|
1 |
+
# Nomic Vision Embedding Model
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
+
This project provides a Gradio interface for the [nomic-ai/nomic-embed-vision-v1.5](https://huggingface.co/nomic-ai/nomic-embed-vision-v1.5) image embedding model. It allows you to upload images and generate embeddings that can be used for various computer vision tasks.
|
4 |
|
5 |
## Features
|
6 |
|
7 |
- Generate embeddings for images using the nomic-ai/nomic-embed-vision-v1.5 model
|
8 |
+
- Simple and intuitive Gradio web interface
|
9 |
+
- Support for various image formats
|
10 |
|
11 |
## How It Works
|
12 |
|
13 |
+
The application uses the Hugging Face Transformers library to load the nomic-ai/nomic-embed-vision-v1.5 model and generate embeddings for uploaded images. The embeddings are high-dimensional vector representations of the images that capture their semantic content.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
## Deployment
|
16 |
|
|
|
32 |
## Requirements
|
33 |
|
34 |
- Python 3.7+
|
35 |
+
- Gradio 4.19.0+
|
36 |
- Transformers
|
37 |
- PyTorch
|
38 |
- Pillow
|
39 |
- NumPy
|
40 |
+
- Requests
|
41 |
|
42 |
## License
|
43 |
|
app.py
CHANGED
@@ -10,16 +10,6 @@ import requests
|
|
10 |
from typing import Dict, List, Any, Optional
|
11 |
from transformers.pipelines import pipeline
|
12 |
|
13 |
-
# MCP imports
|
14 |
-
from modelcontextprotocol.server import Server
|
15 |
-
from modelcontextprotocol.server.gradio import GradioServerTransport
|
16 |
-
from modelcontextprotocol.types import (
|
17 |
-
CallToolRequestSchema,
|
18 |
-
ErrorCode,
|
19 |
-
ListToolsRequestSchema,
|
20 |
-
McpError,
|
21 |
-
)
|
22 |
-
|
23 |
# Initialize the model
|
24 |
model = pipeline("image-feature-extraction", model="nomic-ai/nomic-embed-vision-v1.5", trust_remote_code=True)
|
25 |
|
@@ -76,157 +66,45 @@ def generate_embedding(image):
|
|
76 |
except Exception as e:
|
77 |
print(f"Error generating embedding: {str(e)}")
|
78 |
return None
|
79 |
-
|
80 |
-
return {
|
81 |
-
"embedding": embedding_list,
|
82 |
-
"dimension": embedding_dim
|
83 |
-
}
|
84 |
|
85 |
-
#
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
with gr.Column():
|
92 |
-
input_image = gr.Image(type="pil", label="Input Image")
|
93 |
-
embed_btn = gr.Button("Generate Embedding")
|
94 |
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
def update_embedding(img):
|
100 |
-
result = generate_embedding(img)
|
101 |
-
if result is None:
|
102 |
-
return {
|
103 |
-
embedding_json: None,
|
104 |
-
embedding_dim: "No embedding generated"
|
105 |
-
}
|
106 |
-
return {
|
107 |
-
embedding_json: result,
|
108 |
-
embedding_dim: f"Dimension: {len(result['embedding'])}"
|
109 |
-
}
|
110 |
-
|
111 |
-
embed_btn.click(
|
112 |
-
fn=update_embedding,
|
113 |
-
inputs=[input_image],
|
114 |
-
outputs=[embedding_json, embedding_dim]
|
115 |
-
)
|
116 |
|
117 |
-
#
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
"version": "0.1.0",
|
124 |
-
},
|
125 |
-
{
|
126 |
-
"capabilities": {
|
127 |
-
"tools": {},
|
128 |
-
},
|
129 |
-
}
|
130 |
-
)
|
131 |
-
|
132 |
-
self.setup_tool_handlers()
|
133 |
-
|
134 |
-
# Error handling
|
135 |
-
self.server.onerror = lambda error: print(f"[MCP Error] {error}")
|
136 |
-
|
137 |
-
def setup_tool_handlers(self):
|
138 |
-
self.server.set_request_handler(ListToolsRequestSchema, self.handle_list_tools)
|
139 |
-
self.server.set_request_handler(CallToolRequestSchema, self.handle_call_tool)
|
140 |
-
|
141 |
-
async def handle_list_tools(self, request):
|
142 |
-
return {
|
143 |
-
"tools": [
|
144 |
-
{
|
145 |
-
"name": "embed_image",
|
146 |
-
"description": "Generate embeddings for an image using nomic-ai/nomic-embed-vision-v1.5",
|
147 |
-
"inputSchema": {
|
148 |
-
"type": "object",
|
149 |
-
"properties": {
|
150 |
-
"image_url": {
|
151 |
-
"type": "string",
|
152 |
-
"description": "URL of the image to embed",
|
153 |
-
},
|
154 |
-
"image_data": {
|
155 |
-
"type": "string",
|
156 |
-
"description": "Base64-encoded image data (alternative to image_url)",
|
157 |
-
},
|
158 |
-
},
|
159 |
-
"anyOf": [
|
160 |
-
{"required": ["image_url"]},
|
161 |
-
{"required": ["image_data"]},
|
162 |
-
],
|
163 |
-
},
|
164 |
-
}
|
165 |
-
]
|
166 |
-
}
|
167 |
-
|
168 |
-
async def handle_call_tool(self, request):
|
169 |
-
if request.params.name != "embed_image":
|
170 |
-
raise McpError(
|
171 |
-
ErrorCode.MethodNotFound,
|
172 |
-
f"Unknown tool: {request.params.name}"
|
173 |
-
)
|
174 |
-
|
175 |
-
args = request.params.arguments
|
176 |
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
from io import BytesIO
|
182 |
-
|
183 |
-
response = requests.get(args["image_url"])
|
184 |
-
image = Image.open(BytesIO(response.content))
|
185 |
-
|
186 |
-
# Handle image from base64 data
|
187 |
-
elif "image_data" in args:
|
188 |
-
import base64
|
189 |
-
from io import BytesIO
|
190 |
-
|
191 |
-
image_data = base64.b64decode(args["image_data"])
|
192 |
-
image = Image.open(BytesIO(image_data))
|
193 |
-
|
194 |
-
else:
|
195 |
-
raise McpError(
|
196 |
-
ErrorCode.InvalidParams,
|
197 |
-
"Either image_url or image_data must be provided"
|
198 |
-
)
|
199 |
-
|
200 |
-
# Generate embedding
|
201 |
-
result = generate_embedding(image)
|
202 |
-
|
203 |
-
return {
|
204 |
-
"content": [
|
205 |
-
{
|
206 |
-
"type": "text",
|
207 |
-
"text": json.dumps(result, indent=2),
|
208 |
-
}
|
209 |
-
]
|
210 |
-
}
|
211 |
-
|
212 |
-
except Exception as e:
|
213 |
-
return {
|
214 |
-
"content": [
|
215 |
-
{
|
216 |
-
"type": "text",
|
217 |
-
"text": f"Error generating embedding: {str(e)}",
|
218 |
-
}
|
219 |
-
],
|
220 |
-
"isError": True,
|
221 |
-
}
|
222 |
-
|
223 |
-
# Initialize and run the MCP server
|
224 |
-
embedding_server = NomicEmbeddingServer()
|
225 |
|
226 |
-
#
|
227 |
-
|
228 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
229 |
|
230 |
-
# Launch the
|
231 |
if __name__ == "__main__":
|
232 |
-
|
|
|
10 |
from typing import Dict, List, Any, Optional
|
11 |
from transformers.pipelines import pipeline
|
12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
# Initialize the model
|
14 |
model = pipeline("image-feature-extraction", model="nomic-ai/nomic-embed-vision-v1.5", trust_remote_code=True)
|
15 |
|
|
|
66 |
except Exception as e:
|
67 |
print(f"Error generating embedding: {str(e)}")
|
68 |
return None
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
+
# Function to generate embeddings from an image URL
|
71 |
+
def embed_image_from_url(image_url):
|
72 |
+
try:
|
73 |
+
# Download the image
|
74 |
+
response = requests.get(image_url)
|
75 |
+
image = Image.open(BytesIO(response.content))
|
|
|
|
|
|
|
76 |
|
77 |
+
# Generate embedding
|
78 |
+
return generate_embedding(image)
|
79 |
+
except Exception as e:
|
80 |
+
return {"error": str(e)}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
81 |
|
82 |
+
# Function to generate embeddings from base64 image data
|
83 |
+
def embed_image_from_base64(image_data):
|
84 |
+
try:
|
85 |
+
# Decode the base64 image
|
86 |
+
decoded_data = base64.b64decode(image_data)
|
87 |
+
image = Image.open(BytesIO(decoded_data))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
88 |
|
89 |
+
# Generate embedding
|
90 |
+
return generate_embedding(image)
|
91 |
+
except Exception as e:
|
92 |
+
return {"error": str(e)}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
|
94 |
+
# Create a Gradio app
|
95 |
+
app = gr.Interface(
|
96 |
+
fn=generate_embedding,
|
97 |
+
inputs=gr.Image(type="pil", label="Input Image"),
|
98 |
+
outputs=[
|
99 |
+
gr.JSON(label="Embedding Output"),
|
100 |
+
gr.Textbox(label="Embedding Dimension")
|
101 |
+
],
|
102 |
+
title="Nomic Vision Embedding Model (nomic-ai/nomic-embed-vision-v1.5)",
|
103 |
+
description="Upload an image to generate embeddings using the Nomic Vision model.",
|
104 |
+
examples=[["examples/example1.jpg"], ["examples/example2.jpg"]],
|
105 |
+
allow_flagging="never"
|
106 |
+
)
|
107 |
|
108 |
+
# Launch the app
|
109 |
if __name__ == "__main__":
|
110 |
+
app.launch()
|
app_hf.py
CHANGED
@@ -10,23 +10,13 @@ import requests
|
|
10 |
from typing import Dict, List, Any, Optional
|
11 |
from transformers.pipelines import pipeline
|
12 |
|
13 |
-
# MCP imports
|
14 |
-
from modelcontextprotocol.server import Server
|
15 |
-
from modelcontextprotocol.server.gradio import GradioServerTransport
|
16 |
-
from modelcontextprotocol.types import (
|
17 |
-
CallToolRequestSchema,
|
18 |
-
ErrorCode,
|
19 |
-
ListToolsRequestSchema,
|
20 |
-
McpError,
|
21 |
-
)
|
22 |
-
|
23 |
# Initialize the model
|
24 |
model = pipeline("image-feature-extraction", model="nomic-ai/nomic-embed-vision-v1.5", trust_remote_code=True)
|
25 |
|
26 |
# Function to generate embeddings from an image
|
27 |
def generate_embedding(image):
|
28 |
if image is None:
|
29 |
-
return None
|
30 |
|
31 |
# Convert to PIL Image if needed
|
32 |
if not isinstance(image, Image.Image):
|
@@ -57,14 +47,14 @@ def generate_embedding(image):
|
|
57 |
embedding_list = list(result)
|
58 |
else:
|
59 |
print("Result is None")
|
60 |
-
return None
|
61 |
except:
|
62 |
print(f"Couldn't convert result of type {type(result)} to list")
|
63 |
-
return None
|
64 |
|
65 |
# Ensure we have a valid embedding list
|
66 |
if embedding_list is None:
|
67 |
-
return None
|
68 |
|
69 |
# Calculate embedding dimension
|
70 |
embedding_dim = len(embedding_list)
|
@@ -72,151 +62,25 @@ def generate_embedding(image):
|
|
72 |
return {
|
73 |
"embedding": embedding_list,
|
74 |
"dimension": embedding_dim
|
75 |
-
}
|
76 |
except Exception as e:
|
77 |
print(f"Error generating embedding: {str(e)}")
|
78 |
-
return None
|
79 |
|
80 |
-
# Gradio
|
81 |
-
|
82 |
-
|
83 |
-
gr.
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
embedding_dim = gr.Textbox(label="Embedding Dimension")
|
93 |
-
|
94 |
-
def update_embedding(img):
|
95 |
-
result = generate_embedding(img)
|
96 |
-
if result is None:
|
97 |
-
return {
|
98 |
-
embedding_json: None,
|
99 |
-
embedding_dim: "No embedding generated"
|
100 |
-
}
|
101 |
-
return {
|
102 |
-
embedding_json: result,
|
103 |
-
embedding_dim: f"Dimension: {len(result['embedding'])}"
|
104 |
-
}
|
105 |
-
|
106 |
-
embed_btn.click(
|
107 |
-
fn=update_embedding,
|
108 |
-
inputs=[input_image],
|
109 |
-
outputs=[embedding_json, embedding_dim]
|
110 |
-
)
|
111 |
-
|
112 |
-
# MCP Server Implementation
|
113 |
-
class NomicEmbeddingServer:
|
114 |
-
def __init__(self):
|
115 |
-
self.server = Server(
|
116 |
-
{
|
117 |
-
"name": "nomic-embedding-server",
|
118 |
-
"version": "0.1.0",
|
119 |
-
},
|
120 |
-
{
|
121 |
-
"capabilities": {
|
122 |
-
"tools": {},
|
123 |
-
},
|
124 |
-
}
|
125 |
-
)
|
126 |
-
|
127 |
-
self.setup_tool_handlers()
|
128 |
-
|
129 |
-
# Error handling
|
130 |
-
self.server.onerror = lambda error: print(f"[MCP Error] {error}")
|
131 |
-
|
132 |
-
def setup_tool_handlers(self):
|
133 |
-
self.server.set_request_handler(ListToolsRequestSchema, self.handle_list_tools)
|
134 |
-
self.server.set_request_handler(CallToolRequestSchema, self.handle_call_tool)
|
135 |
-
|
136 |
-
async def handle_list_tools(self, request):
|
137 |
-
return {
|
138 |
-
"tools": [
|
139 |
-
{
|
140 |
-
"name": "embed_image",
|
141 |
-
"description": "Generate embeddings for an image using nomic-ai/nomic-embed-vision-v1.5",
|
142 |
-
"inputSchema": {
|
143 |
-
"type": "object",
|
144 |
-
"properties": {
|
145 |
-
"image_url": {
|
146 |
-
"type": "string",
|
147 |
-
"description": "URL of the image to embed",
|
148 |
-
},
|
149 |
-
"image_data": {
|
150 |
-
"type": "string",
|
151 |
-
"description": "Base64-encoded image data (alternative to image_url)",
|
152 |
-
},
|
153 |
-
},
|
154 |
-
"anyOf": [
|
155 |
-
{"required": ["image_url"]},
|
156 |
-
{"required": ["image_data"]},
|
157 |
-
],
|
158 |
-
},
|
159 |
-
}
|
160 |
-
]
|
161 |
-
}
|
162 |
-
|
163 |
-
async def handle_call_tool(self, request):
|
164 |
-
if request.params.name != "embed_image":
|
165 |
-
raise McpError(
|
166 |
-
ErrorCode.MethodNotFound,
|
167 |
-
f"Unknown tool: {request.params.name}"
|
168 |
-
)
|
169 |
-
|
170 |
-
args = request.params.arguments
|
171 |
-
|
172 |
-
try:
|
173 |
-
# Handle image from URL
|
174 |
-
if "image_url" in args:
|
175 |
-
response = requests.get(args["image_url"])
|
176 |
-
image = Image.open(BytesIO(response.content))
|
177 |
-
|
178 |
-
# Handle image from base64 data
|
179 |
-
elif "image_data" in args:
|
180 |
-
image_data = base64.b64decode(args["image_data"])
|
181 |
-
image = Image.open(BytesIO(image_data))
|
182 |
-
|
183 |
-
else:
|
184 |
-
raise McpError(
|
185 |
-
ErrorCode.InvalidParams,
|
186 |
-
"Either image_url or image_data must be provided"
|
187 |
-
)
|
188 |
-
|
189 |
-
# Generate embedding
|
190 |
-
result = generate_embedding(image)
|
191 |
-
|
192 |
-
return {
|
193 |
-
"content": [
|
194 |
-
{
|
195 |
-
"type": "text",
|
196 |
-
"text": json.dumps(result, indent=2),
|
197 |
-
}
|
198 |
-
]
|
199 |
-
}
|
200 |
-
|
201 |
-
except Exception as e:
|
202 |
-
return {
|
203 |
-
"content": [
|
204 |
-
{
|
205 |
-
"type": "text",
|
206 |
-
"text": f"Error generating embedding: {str(e)}",
|
207 |
-
}
|
208 |
-
],
|
209 |
-
"isError": True,
|
210 |
-
}
|
211 |
-
|
212 |
-
# Initialize and run the MCP server
|
213 |
-
embedding_server = NomicEmbeddingServer()
|
214 |
-
|
215 |
-
# Connect the MCP server to the Gradio app
|
216 |
-
transport = GradioServerTransport(demo)
|
217 |
-
embedding_server.server.connect(transport)
|
218 |
|
219 |
-
# Launch the
|
220 |
if __name__ == "__main__":
|
221 |
# For Huggingface Spaces, we need to specify the server name and port
|
222 |
-
|
|
|
10 |
from typing import Dict, List, Any, Optional
|
11 |
from transformers.pipelines import pipeline
|
12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
# Initialize the model
|
14 |
model = pipeline("image-feature-extraction", model="nomic-ai/nomic-embed-vision-v1.5", trust_remote_code=True)
|
15 |
|
16 |
# Function to generate embeddings from an image
|
17 |
def generate_embedding(image):
|
18 |
if image is None:
|
19 |
+
return None, "No image provided"
|
20 |
|
21 |
# Convert to PIL Image if needed
|
22 |
if not isinstance(image, Image.Image):
|
|
|
47 |
embedding_list = list(result)
|
48 |
else:
|
49 |
print("Result is None")
|
50 |
+
return None, "Failed to generate embedding"
|
51 |
except:
|
52 |
print(f"Couldn't convert result of type {type(result)} to list")
|
53 |
+
return None, "Failed to process embedding"
|
54 |
|
55 |
# Ensure we have a valid embedding list
|
56 |
if embedding_list is None:
|
57 |
+
return None, "Failed to generate embedding"
|
58 |
|
59 |
# Calculate embedding dimension
|
60 |
embedding_dim = len(embedding_list)
|
|
|
62 |
return {
|
63 |
"embedding": embedding_list,
|
64 |
"dimension": embedding_dim
|
65 |
+
}, f"Dimension: {embedding_dim}"
|
66 |
except Exception as e:
|
67 |
print(f"Error generating embedding: {str(e)}")
|
68 |
+
return None, f"Error: {str(e)}"
|
69 |
|
70 |
+
# Create a Gradio app
|
71 |
+
app = gr.Interface(
|
72 |
+
fn=generate_embedding,
|
73 |
+
inputs=gr.Image(type="pil", label="Input Image"),
|
74 |
+
outputs=[
|
75 |
+
gr.JSON(label="Embedding Output"),
|
76 |
+
gr.Textbox(label="Embedding Dimension")
|
77 |
+
],
|
78 |
+
title="Nomic Vision Embedding Model (nomic-ai/nomic-embed-vision-v1.5)",
|
79 |
+
description="Upload an image to generate embeddings using the Nomic Vision model.",
|
80 |
+
allow_flagging="never"
|
81 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
82 |
|
83 |
+
# Launch the app
|
84 |
if __name__ == "__main__":
|
85 |
# For Huggingface Spaces, we need to specify the server name and port
|
86 |
+
app.launch(server_name="0.0.0.0", server_port=7860)
|
example_client.py
CHANGED
@@ -6,13 +6,13 @@ import json
|
|
6 |
import matplotlib.pyplot as plt
|
7 |
import numpy as np
|
8 |
|
9 |
-
# This is an example client that demonstrates how to use the
|
10 |
# You would replace this URL with the actual URL of your deployed Huggingface Space
|
11 |
-
|
12 |
|
13 |
def embed_image_from_url(image_url):
|
14 |
"""
|
15 |
-
Generate embeddings for an image using the
|
16 |
|
17 |
Args:
|
18 |
image_url: URL of the image to embed
|
@@ -20,38 +20,43 @@ def embed_image_from_url(image_url):
|
|
20 |
Returns:
|
21 |
The embedding vector and its dimension
|
22 |
"""
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
|
|
|
|
|
|
|
|
|
|
51 |
|
52 |
def embed_image_from_file(image_path):
|
53 |
"""
|
54 |
-
Generate embeddings for an image using the
|
55 |
|
56 |
Args:
|
57 |
image_path: Path to the image file
|
@@ -59,41 +64,38 @@ def embed_image_from_file(image_path):
|
|
59 |
Returns:
|
60 |
The embedding vector and its dimension
|
61 |
"""
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
embedding_data = json.loads(content)
|
95 |
-
|
96 |
-
return embedding_data
|
97 |
|
98 |
def visualize_embedding(embedding):
|
99 |
"""
|
@@ -128,10 +130,10 @@ if __name__ == "__main__":
|
|
128 |
image_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/bert-architects.png"
|
129 |
print(f"Generating embedding for image: {image_url}")
|
130 |
|
131 |
-
embedding_data = embed_image_from_url(image_url)
|
132 |
|
133 |
if embedding_data:
|
134 |
-
print(f"Embedding dimension: {
|
135 |
print(f"First 10 values of embedding: {embedding_data['embedding'][:10]}...")
|
136 |
|
137 |
# Visualize the embedding
|
@@ -141,10 +143,10 @@ if __name__ == "__main__":
|
|
141 |
# Uncomment the following lines to use a local image file
|
142 |
# image_path = "path/to/your/image.jpg"
|
143 |
# print(f"Generating embedding for image: {image_path}")
|
144 |
-
# embedding_data = embed_image_from_file(image_path)
|
145 |
#
|
146 |
# if embedding_data:
|
147 |
-
# print(f"Embedding dimension: {
|
148 |
# print(f"First 10 values of embedding: {embedding_data['embedding'][:10]}...")
|
149 |
#
|
150 |
# # Visualize the embedding
|
|
|
6 |
import matplotlib.pyplot as plt
|
7 |
import numpy as np
|
8 |
|
9 |
+
# This is an example client that demonstrates how to use the Gradio API
|
10 |
# You would replace this URL with the actual URL of your deployed Huggingface Space
|
11 |
+
GRADIO_API_URL = "https://your-username-nomic-vision-embedding.hf.space/api/predict"
|
12 |
|
13 |
def embed_image_from_url(image_url):
|
14 |
"""
|
15 |
+
Generate embeddings for an image using the Gradio API
|
16 |
|
17 |
Args:
|
18 |
image_url: URL of the image to embed
|
|
|
20 |
Returns:
|
21 |
The embedding vector and its dimension
|
22 |
"""
|
23 |
+
try:
|
24 |
+
# Download the image
|
25 |
+
response = requests.get(image_url)
|
26 |
+
image = Image.open(BytesIO(response.content))
|
27 |
+
|
28 |
+
# Convert image to bytes
|
29 |
+
img_byte_arr = BytesIO()
|
30 |
+
image.save(img_byte_arr, format='PNG')
|
31 |
+
img_byte_arr = img_byte_arr.getvalue()
|
32 |
+
|
33 |
+
# Prepare the request
|
34 |
+
files = {
|
35 |
+
'data': ('image.png', img_byte_arr, 'image/png')
|
36 |
+
}
|
37 |
+
|
38 |
+
# Send the request to the Gradio API
|
39 |
+
response = requests.post(GRADIO_API_URL, files=files)
|
40 |
+
|
41 |
+
# Parse the response
|
42 |
+
if response.status_code == 200:
|
43 |
+
result = response.json()
|
44 |
+
embedding_data = result['data'][0]
|
45 |
+
embedding_dim = result['data'][1]
|
46 |
+
|
47 |
+
return embedding_data, embedding_dim
|
48 |
+
else:
|
49 |
+
print(f"Error: HTTP {response.status_code}")
|
50 |
+
print(response.text)
|
51 |
+
return None, None
|
52 |
+
|
53 |
+
except Exception as e:
|
54 |
+
print(f"Error: {str(e)}")
|
55 |
+
return None, None
|
56 |
|
57 |
def embed_image_from_file(image_path):
|
58 |
"""
|
59 |
+
Generate embeddings for an image using the Gradio API
|
60 |
|
61 |
Args:
|
62 |
image_path: Path to the image file
|
|
|
64 |
Returns:
|
65 |
The embedding vector and its dimension
|
66 |
"""
|
67 |
+
try:
|
68 |
+
# Load the image
|
69 |
+
image = Image.open(image_path)
|
70 |
+
|
71 |
+
# Convert image to bytes
|
72 |
+
img_byte_arr = BytesIO()
|
73 |
+
image.save(img_byte_arr, format=image.format if image.format else 'PNG')
|
74 |
+
img_byte_arr = img_byte_arr.getvalue()
|
75 |
+
|
76 |
+
# Prepare the request
|
77 |
+
files = {
|
78 |
+
'data': ('image.png', img_byte_arr, 'image/png')
|
79 |
+
}
|
80 |
+
|
81 |
+
# Send the request to the Gradio API
|
82 |
+
response = requests.post(GRADIO_API_URL, files=files)
|
83 |
+
|
84 |
+
# Parse the response
|
85 |
+
if response.status_code == 200:
|
86 |
+
result = response.json()
|
87 |
+
embedding_data = result['data'][0]
|
88 |
+
embedding_dim = result['data'][1]
|
89 |
+
|
90 |
+
return embedding_data, embedding_dim
|
91 |
+
else:
|
92 |
+
print(f"Error: HTTP {response.status_code}")
|
93 |
+
print(response.text)
|
94 |
+
return None, None
|
95 |
+
|
96 |
+
except Exception as e:
|
97 |
+
print(f"Error: {str(e)}")
|
98 |
+
return None, None
|
|
|
|
|
|
|
99 |
|
100 |
def visualize_embedding(embedding):
|
101 |
"""
|
|
|
130 |
image_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/bert-architects.png"
|
131 |
print(f"Generating embedding for image: {image_url}")
|
132 |
|
133 |
+
embedding_data, embedding_dim = embed_image_from_url(image_url)
|
134 |
|
135 |
if embedding_data:
|
136 |
+
print(f"Embedding dimension: {embedding_dim}")
|
137 |
print(f"First 10 values of embedding: {embedding_data['embedding'][:10]}...")
|
138 |
|
139 |
# Visualize the embedding
|
|
|
143 |
# Uncomment the following lines to use a local image file
|
144 |
# image_path = "path/to/your/image.jpg"
|
145 |
# print(f"Generating embedding for image: {image_path}")
|
146 |
+
# embedding_data, embedding_dim = embed_image_from_file(image_path)
|
147 |
#
|
148 |
# if embedding_data:
|
149 |
+
# print(f"Embedding dimension: {embedding_dim}")
|
150 |
# print(f"First 10 values of embedding: {embedding_data['embedding'][:10]}...")
|
151 |
#
|
152 |
# # Visualize the embedding
|
requirements.txt
CHANGED
@@ -1,9 +1,7 @@
|
|
1 |
-
transformers
|
2 |
-
torch
|
3 |
-
pillow
|
4 |
-
numpy
|
5 |
-
requests
|
6 |
-
|
7 |
-
|
8 |
-
mcp
|
9 |
-
https://gradio-pypi-previews.s3.amazonaws.com/3b5cace94781b90993b596a83fb39fd1584d68ee/gradio-5.26.0-py3-none-any.whl
|
|
|
1 |
+
transformers>=4.30.0
|
2 |
+
torch>=2.0.0
|
3 |
+
pillow>=9.0.0
|
4 |
+
numpy>=1.20.0
|
5 |
+
requests>=2.25.0
|
6 |
+
gradio>=4.19.0
|
7 |
+
matplotlib>=3.5.0
|
|
|
|
requirements_hf.txt
CHANGED
@@ -1,7 +1,6 @@
|
|
1 |
-
gradio>=4.
|
2 |
transformers>=4.30.0
|
3 |
torch>=2.0.0
|
4 |
pillow>=9.0.0
|
5 |
numpy>=1.20.0
|
6 |
-
requests>=2.25.0
|
7 |
-
modelcontextprotocol>=0.1.0
|
|
|
1 |
+
gradio>=4.19.0
|
2 |
transformers>=4.30.0
|
3 |
torch>=2.0.0
|
4 |
pillow>=9.0.0
|
5 |
numpy>=1.20.0
|
6 |
+
requests>=2.25.0
|
|
test_local.py
CHANGED
@@ -1,41 +1,37 @@
|
|
1 |
import requests
|
2 |
-
import base64
|
3 |
from PIL import Image
|
4 |
import io
|
5 |
-
import json
|
6 |
import sys
|
|
|
7 |
|
8 |
def test_local_server(image_path=None):
|
9 |
"""
|
10 |
-
Test the local
|
11 |
|
12 |
Args:
|
13 |
image_path: Path to the image file. If None, a test URL will be used.
|
14 |
"""
|
15 |
# Local server URL (default Gradio port)
|
16 |
-
server_url = "http://localhost:7860/
|
17 |
|
18 |
if image_path:
|
19 |
-
# Load the image
|
20 |
try:
|
21 |
-
|
22 |
-
|
23 |
|
24 |
-
#
|
25 |
-
|
|
|
|
|
26 |
|
27 |
-
# Prepare the
|
28 |
-
|
29 |
-
|
30 |
-
"method": "callTool",
|
31 |
-
"params": {
|
32 |
-
"name": "embed_image",
|
33 |
-
"arguments": {
|
34 |
-
"image_data": image_base64
|
35 |
-
}
|
36 |
-
},
|
37 |
-
"id": 1
|
38 |
}
|
|
|
|
|
|
|
|
|
39 |
except Exception as e:
|
40 |
print(f"Error loading image: {str(e)}")
|
41 |
return
|
@@ -44,40 +40,44 @@ def test_local_server(image_path=None):
|
|
44 |
test_image_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/bert-architects.png"
|
45 |
print(f"Using test image URL: {test_image_url}")
|
46 |
|
47 |
-
#
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
|
|
58 |
}
|
|
|
|
|
|
|
59 |
|
60 |
-
print("Sending request to local
|
61 |
|
62 |
try:
|
63 |
-
# Send the request to the MCP server
|
64 |
-
response = requests.post(server_url, json=mcp_request)
|
65 |
-
|
66 |
# Check if the request was successful
|
67 |
if response.status_code == 200:
|
68 |
# Parse the response
|
69 |
result = response.json()
|
70 |
|
71 |
if "error" in result:
|
72 |
-
print(f"Error from server: {result['error']
|
73 |
else:
|
74 |
# Extract the embedding from the response
|
75 |
-
|
76 |
-
|
77 |
|
78 |
print("✅ Test successful!")
|
79 |
-
print(f"Embedding dimension: {
|
80 |
-
|
|
|
|
|
|
|
81 |
else:
|
82 |
print(f"❌ Error: HTTP {response.status_code}")
|
83 |
print(response.text)
|
|
|
1 |
import requests
|
|
|
2 |
from PIL import Image
|
3 |
import io
|
|
|
4 |
import sys
|
5 |
+
import json
|
6 |
|
7 |
def test_local_server(image_path=None):
|
8 |
"""
|
9 |
+
Test the local Gradio server by sending a request to generate embeddings for an image
|
10 |
|
11 |
Args:
|
12 |
image_path: Path to the image file. If None, a test URL will be used.
|
13 |
"""
|
14 |
# Local server URL (default Gradio port)
|
15 |
+
server_url = "http://localhost:7860/api/predict"
|
16 |
|
17 |
if image_path:
|
|
|
18 |
try:
|
19 |
+
# Load the image
|
20 |
+
image = Image.open(image_path)
|
21 |
|
22 |
+
# Convert image to bytes
|
23 |
+
img_byte_arr = io.BytesIO()
|
24 |
+
image.save(img_byte_arr, format=image.format if image.format else 'PNG')
|
25 |
+
img_byte_arr = img_byte_arr.getvalue()
|
26 |
|
27 |
+
# Prepare the request
|
28 |
+
files = {
|
29 |
+
'data': ('image.png', img_byte_arr, 'image/png')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
}
|
31 |
+
|
32 |
+
# Send the request
|
33 |
+
response = requests.post(server_url, files=files)
|
34 |
+
|
35 |
except Exception as e:
|
36 |
print(f"Error loading image: {str(e)}")
|
37 |
return
|
|
|
40 |
test_image_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/bert-architects.png"
|
41 |
print(f"Using test image URL: {test_image_url}")
|
42 |
|
43 |
+
# Download the image
|
44 |
+
response = requests.get(test_image_url)
|
45 |
+
image = Image.open(io.BytesIO(response.content))
|
46 |
+
|
47 |
+
# Convert image to bytes
|
48 |
+
img_byte_arr = io.BytesIO()
|
49 |
+
image.save(img_byte_arr, format='PNG')
|
50 |
+
img_byte_arr = img_byte_arr.getvalue()
|
51 |
+
|
52 |
+
# Prepare the request
|
53 |
+
files = {
|
54 |
+
'data': ('image.png', img_byte_arr, 'image/png')
|
55 |
}
|
56 |
+
|
57 |
+
# Send the request
|
58 |
+
response = requests.post(server_url, files=files)
|
59 |
|
60 |
+
print("Sending request to local Gradio server...")
|
61 |
|
62 |
try:
|
|
|
|
|
|
|
63 |
# Check if the request was successful
|
64 |
if response.status_code == 200:
|
65 |
# Parse the response
|
66 |
result = response.json()
|
67 |
|
68 |
if "error" in result:
|
69 |
+
print(f"Error from server: {result['error']}")
|
70 |
else:
|
71 |
# Extract the embedding from the response
|
72 |
+
embedding_data = result['data'][0]
|
73 |
+
embedding_dim = result['data'][1]
|
74 |
|
75 |
print("✅ Test successful!")
|
76 |
+
print(f"Embedding dimension: {embedding_dim}")
|
77 |
+
if embedding_data:
|
78 |
+
print(f"First 10 values of embedding: {embedding_data['embedding'][:10]}...")
|
79 |
+
else:
|
80 |
+
print("No embedding data returned")
|
81 |
else:
|
82 |
print(f"❌ Error: HTTP {response.status_code}")
|
83 |
print(response.text)
|