File size: 15,227 Bytes
5c263d5
 
 
 
 
 
 
 
 
 
 
 
8f80642
5c263d5
 
 
7bac21a
 
 
8f80642
5c263d5
 
 
 
 
 
 
 
 
 
 
15a969b
 
 
 
5c263d5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fef773e
 
 
5c263d5
afab4d9
fef773e
 
 
 
 
 
 
 
 
5c263d5
 
fef773e
 
 
 
 
 
 
 
5c263d5
fef773e
5c263d5
 
 
 
 
 
 
afab4d9
7bac21a
 
fef773e
 
 
 
 
 
afab4d9
fef773e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
afab4d9
7bac21a
 
 
fef773e
7bac21a
 
 
 
afab4d9
7bac21a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fef773e
7bac21a
 
 
5c263d5
afab4d9
7bac21a
 
5c263d5
 
 
fef773e
7bac21a
fef773e
7bac21a
afab4d9
7bac21a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fef773e
7bac21a
 
 
fef773e
afab4d9
7bac21a
 
fef773e
 
7bac21a
 
 
 
afab4d9
7bac21a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
afab4d9
7bac21a
 
fef773e
 
7bac21a
 
 
 
 
 
fef773e
afab4d9
fef773e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
afab4d9
7bac21a
 
8f80642
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15a969b
8f80642
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15a969b
8f80642
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15a969b
8f80642
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15a969b
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
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
53
54
55
56
57
58
59
60
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
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
# -------------------------------------------------------------------
# This source file is available under the terms of the
# Pimcore Open Core License (POCL)
# Full copyright and license information is available in
# LICENSE.md which is distributed with this source code.
#
#  @copyright  Copyright (c) Pimcore GmbH (https://www.pimcore.com)
#  @license    Pimcore Open Core License (POCL)
# -------------------------------------------------------------------

import torch

from fastapi import FastAPI, Path, Request, File, UploadFile
import logging
import sys

from .translation_task import TranslationTaskService
from .classification import ClassificationTaskService
from .text_to_image import TextToImageTaskService
from .embeddings import ImageEmbeddingTaskService, TextEmbeddingTaskService

app = FastAPI(
    title="Pimcore Local Inference Service",
    description="This services allows HF inference provider compatible inference to models which are not available at HF inference providers.",
    version="1.0.0"
)

logging.basicConfig(format='%(asctime)s %(levelname)-8s %(message)s')
logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG)

# Create singleton instances of embedding services to enable model caching across requests
image_embedding_service = ImageEmbeddingTaskService(logger)
text_embedding_service = TextEmbeddingTaskService(logger)


class StreamToLogger(object):
    def __init__(self, logger, log_level):
        self.logger = logger
        self.log_level = log_level
        self.linebuf = ''

    def write(self, buf):
        for line in buf.rstrip().splitlines():
            self.logger.log(self.log_level, line.rstrip())

    def flush(self):
        pass

sys.stdout = StreamToLogger(logger, logging.INFO)
sys.stderr = StreamToLogger(logger, logging.ERROR)

@app.get("/gpu_check")
async def gpu_check():
    """ Check if a GPU is available """

    gpu = 'GPU not available'
    if torch.cuda.is_available():
        gpu = 'GPU is available'
        print("GPU is available")
    else:
        print("GPU is not available")

    return {'success': True, 'gpu': gpu}


# =========================
# Translation Task
# =========================
@app.post(
    "/translation/{model_name:path}", 
    openapi_extra={
        "requestBody": {
            "content": {
                "application/json": {
                    "example": {
                        "inputs": "Hello, world! foo bar",
                        "parameters": {"repetition_penalty": 1.6}
                    }
                }
            }
        }
    }        
)
async def translate(
    request: Request,
    model_name: str = Path(
        ...,
        description="The name of the translation model (e.g. Helsinki-NLP/opus-mt-en-de)",
        example="Helsinki-NLP/opus-mt-en-de"
    )
    ):
    """
    Execute translation tasks.

    Returns:
        list: The translation result(s) as returned by the pipeline.
    """

    model_name = model_name.rstrip("/")
    translationTaskService = TranslationTaskService(logger)
    return await translationTaskService.translate(request, model_name)


# =========================
# Zero-Shot Image Classification Task
# =========================
@app.post(
    "/zero-shot-image-classification/{model_name:path}",
    openapi_extra={
        "requestBody": {
            "content": {
                "application/json": {
                    "example": {
                        "inputs": "base64_encoded_image_string",
                        "parameters": {"candidate_labels": "green, yellow, blue, white, silver"}
                    }
                }
            }
        }        
    }
)
async def zero_shot_image_classification(
    request: Request,
    model_name: str = Path(
        ...,
        description="The name of the zero-shot classification model (e.g., openai/clip-vit-large-patch14-336)",
        example="openai/clip-vit-large-patch14-336"
    )
    ):
    """
    Execute zero-shot image classification tasks.

    Returns:
        list: The classification result(s) as returned by the pipeline.
    """

    model_name = model_name.rstrip("/")
    zeroShotTask = ClassificationTaskService(logger, 'zero-shot-image-classification')
    return await zeroShotTask.classify(request, model_name)


# =========================
# Image Classification Task
# =========================
@app.post(
    "/image-classification/{model_name:path}",
    openapi_extra={
        "requestBody": {
            "content": {
                "application/json": {
                    "example": {
                        "inputs": "base64_encoded_image_string"
                    }
                }
            }
        }        
    }
)
async def image_classification(
    request: Request,
    model_name: str = Path(
        ...,
        description="The name of the image classification model (e.g., pimcore/car-countries-classification)",
        example="pimcore/car-countries-classification"
    )
    ):
    """
    Execute image classification tasks.

    Returns:
        list: The classification result(s) as returned by the pipeline.
    """

    model_name = model_name.rstrip("/")
    imageTask = ClassificationTaskService(logger, 'image-classification')
    return await imageTask.classify(request, model_name)



# =========================
# Zero-Shot Text Classification Task
# =========================
@app.post(
    "/zero-shot-text-classification/{model_name:path}",
    openapi_extra={
        "requestBody": {
            "content": {
                "application/json": {
                    "example": {
                        "inputs": "text to classify",
                        "parameters": {"candidate_labels": "green, yellow, blue, white, silver"}
                    }
                }
            }
        }        
    }
)
async def zero_shot_text_classification(
    request: Request,
    model_name: str = Path(
        ...,
        description="The name of the zero-shot text classification model (e.g., facebook/bart-large-mnli)",
        example="facebook/bart-large-mnli"
    )
    ):
    """
    Execute zero-shot text classification tasks.

    Returns:
        list: The classification result(s) as returned by the pipeline.
    """

    model_name = model_name.rstrip("/")
    zeroShotTask = ClassificationTaskService(logger, 'zero-shot-classification')
    return await zeroShotTask.classify(request, model_name)


# =========================
# Text Classification Task
# =========================
@app.post(
    "/text-classification/{model_name:path}",
    openapi_extra={
        "requestBody": {
            "content": {
                "application/json": {
                    "example": {
                        "inputs": "text to classify"
                    }
                }
            }
        }        
    }
)
async def text_classification(
    request: Request,
    model_name: str = Path(
        ...,
        description="The name of the text classification model (e.g., pimcore/car-class-classification)",
        example="pimcore/car-class-classification"
    )
    ):
    """
    Execute text classification tasks.

    Returns:
        list: The classification result(s) as returned by the pipeline.
    """

    model_name = model_name.rstrip("/")
    textTask = ClassificationTaskService(logger, 'text-classification')
    return await textTask.classify(request, model_name)





# =========================
# Image to Text Task
# =========================
@app.post(
    "/image-to-text/{model_name:path}",
    openapi_extra={
        "requestBody": {
            "content": {
                "multipart/form-data": {
                    "schema": {
                        "type": "object",
                        "properties": {
                            "image": {
                                "type": "string",
                                "format": "binary",
                                "description": "Image file to upload"
                            }
                        },
                        "required": ["image"]
                    }
                }
            }
        }
    }
)
async def image_to_text(
    request: Request,
    model_name: str = Path(
        ...,
        description="The name of the image-to-text (e.g., Salesforce/blip-image-captioning-base)",
        example="Salesforce/blip-image-captioning-base"
    )
    ):
    """
    Execute image-to-text tasks.

    Returns:
        list: The generated text as returned by the pipeline.
    """

    model_name = model_name.rstrip("/")
    imageToTextTask = TextToImageTaskService(logger)
    return await imageToTextTask.extract(request, model_name)


# =========================
# Image Embedding Task
# =========================
@app.post(
    "/image-embedding/{model_name:path}",
    openapi_extra={
        "requestBody": {
            "content": {
                "application/json": {
                    "example": {
                        "inputs": "base64_encoded_image_string"
                    }
                }
            }
        }        
    }
)
async def image_embedding(
    request: Request,
    model_name: str = Path(
        ...,
        description="The name of the image embedding model. Supported models include: google/siglip-so400m-patch14-384, openai/clip-vit-large-patch14, openai/clip-vit-base-patch16, laion/CLIP-ViT-bigG-14-laion2B-39B-b160k, Salesforce/blip-itm-large-flickr",
        example="google/siglip-so400m-patch14-384"
    )
    ):
    """
    Generate embedding vectors for image data.
    
    The service supports multiple model types including SigLIP, CLIP, and BLIP models.
    Returns a dense vector representation of the input image.

    Returns:
        list: The embedding vector as a list of float values.
    """

    model_name = model_name.rstrip("/")
    return await image_embedding_service.generate_embedding(request, model_name)


# =========================
# Image Embedding Upload Task (Development/Testing)
# =========================
@app.post(
    "/image-embedding-upload/{model_name:path}",
    openapi_extra={
        "requestBody": {
            "content": {
                "multipart/form-data": {
                    "schema": {
                        "type": "object",
                        "properties": {
                            "image": {
                                "type": "string",
                                "format": "binary",
                                "description": "Image file to upload for embedding generation"
                            }
                        },
                        "required": ["image"]
                    }
                }
            }
        },
        "responses": {
            "200": {
                "description": "Image embedding vector",
                "content": {
                    "application/json": {
                        "example": {
                            "embeddings": [0.1, -0.2, 0.3, "..."]
                        }
                    }
                }
            }
        }
    }
)
async def image_embedding_upload(
    image: UploadFile = File(..., description="Image file to generate embeddings for"),
    model_name: str = Path(
        ...,
        description="The name of the image embedding model. Supported models include: google/siglip-so400m-patch14-384, openai/clip-vit-large-patch14, openai/clip-vit-base-patch16, laion/CLIP-ViT-bigG-14-laion2B-39B-b160k, Salesforce/blip-itm-large-flickr",
        example="google/siglip-so400m-patch14-384"
    )
    ):
    """
    Generate embedding vectors for uploaded image data (Development/Testing endpoint).
    
    This endpoint allows you to upload an image file directly through the Swagger UI
    for development and testing purposes. The image is processed and converted to
    embedding vectors using the specified model.
    
    Supported formats: JPEG, PNG, GIF, BMP, TIFF
    
    The service supports multiple model types including SigLIP, CLIP, and BLIP models.
    Returns a dense vector representation of the uploaded image.

    Returns:
        dict: The embedding vector as a list of float values.
    """

    model_name = model_name.rstrip("/")
    return await image_embedding_service.generate_embedding_from_upload(image, model_name)


# =========================
# Text Embedding Task
# =========================
@app.post(
    "/text-embedding/{model_name:path}",
    openapi_extra={
        "requestBody": {
            "content": {
                "application/json": {
                    "example": {
                        "inputs": "text to embed"
                    }
                }
            }
        }        
    }
)
async def text_embedding(
    request: Request,
    model_name: str = Path(
        ...,
        description="The name of the text embedding model. Supported models include: google/siglip-so400m-patch14-384, openai/clip-vit-large-patch14, openai/clip-vit-base-patch16, laion/CLIP-ViT-bigG-14-laion2B-39B-b160k, Salesforce/blip-itm-large-flickr",
        example="google/siglip-so400m-patch14-384"
    )
    ):
    """
    Generate embedding vectors for text data.
    
    The service supports multiple model types including SigLIP, CLIP, and BLIP models.
    Returns a dense vector representation of the input text.

    Returns:
        list: The embedding vector as a list of float values.
    """

    model_name = model_name.rstrip("/")
    return await text_embedding_service.generate_embedding(request, model_name)


# =========================
# Embedding Vector Size
# =========================
@app.get(
    "/embedding-vector-size/{model_name:path}",
    openapi_extra={
        "responses": {
            "200": {
                "description": "Vector size information",
                "content": {
                    "application/json": {
                        "example": {
                            "model_name": "google/siglip-so400m-patch14-384",
                            "vector_size": 1152,
                            "config_attribute_used": "hidden_size"
                        }
                    }
                }
            }
        }
    } 
)
async def embedding_vector_size(
    model_name: str = Path(
        ...,
        description="The name of the embedding model. Supported models include: google/siglip-so400m-patch14-384, openai/clip-vit-large-patch14, openai/clip-vit-base-patch16, laion/CLIP-ViT-bigG-14-laion2B-39B-b160k, Salesforce/blip-itm-large-flickr",
        example="google/siglip-so400m-patch14-384"
    )
    ):
    """
    Get the vector size of embeddings for a given model.
    
    This endpoint returns the dimensionality of the embedding vectors that the model produces.
    Useful for understanding the output format before generating embeddings.

    Returns:
        dict: Information about the vector size including model name, vector size, and configuration attribute used.
    """

    model_name = model_name.rstrip("/")
    # We can use either embedding service as they inherit from the same base class
    return await image_embedding_service.get_embedding_vector_size(model_name)