src/main.py

# -------------------------------------------------------------------
# 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 os
import torch

from fastapi import FastAPI, Path, Depends, HTTPException, UploadFile, Form, File, status, Request
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
from pydantic import BaseModel
from typing import Annotated
import json

import logging
import sys
import base64


from transformers import pipeline

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)


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)



class ResponseModel(BaseModel):
    """ Default response model for endpoints. """
    message: str
    success: bool = True


@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}


from typing import Optional



# =========================
# Translation Task
# =========================

class TranslationRequest(BaseModel):
    inputs: str
    parameters: Optional[dict] = None
    options: Optional[dict] = None

async def get_translation_request(
    request: Request
)  -> TranslationRequest: 
    content_type = request.headers.get("content-type", "")
    if content_type.startswith("application/json"):
        data = await request.json()
        return TranslationRequest(**data)
    if content_type.startswith("application/x-www-form-urlencoded"):
        raw = await request.body()
        try:
            data = json.loads(raw)
            return TranslationRequest(**data)
        except Exception:
            try:
                data = json.loads(raw.decode("utf-8"))
                return TranslationRequest(**data)
            except Exception:
                raise HTTPException(status_code=400, detail="Invalid request body")
    raise HTTPException(status_code=400, detail="Unsupported content type")



@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.
    """

    translationRequest: TranslationRequest = await get_translation_request(request)

    try:
        pipe = pipeline("translation", model=model_name)
    except Exception as e:
        logger.error(f"Failed to load model '{model_name}': {str(e)}")
        raise HTTPException(
            status_code=404,
            detail=f"Model '{model_name}' could not be loaded: {str(e)}"
        )

    try:       
        result = pipe(translationRequest.inputs, **(translationRequest.parameters or {}))
    except Exception as e:
        logger.error(f"Inference failed for model '{model_name}': {str(e)}")
        raise HTTPException(
            status_code=500,
            detail=f"Inference failed: {str(e)}"
        )

    return result


# =========================
# Zero-Shot Image Classification Task
# =========================


class ZeroShotImageClassificationRequest(BaseModel):
    inputs: str
    parameters: Optional[dict] = None

async def get_zero_shot_image_classification_request(
    request: Request
)  -> ZeroShotImageClassificationRequest: 
    content_type = request.headers.get("content-type", "")
    if content_type.startswith("application/json"):
        data = await request.json()
        return ZeroShotImageClassificationRequest(**data)
    if content_type.startswith("application/x-www-form-urlencoded"):
        raw = await request.body()
        try:
            data = json.loads(raw)
            return ZeroShotImageClassificationRequest(**data)
        except Exception:
            try:
                data = json.loads(raw.decode("utf-8"))
                return ZeroShotImageClassificationRequest(**data)
            except Exception:
                raise HTTPException(status_code=400, detail="Invalid request body")
    raise HTTPException(status_code=400, detail="Unsupported content type")



@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.
    """

    zeroShotRequest: ZeroShotImageClassificationRequest = await get_zero_shot_image_classification_request(request)

    try:
        pipe = pipeline("zero-shot-image-classification", model=model_name)
    except Exception as e:
        logger.error(f"Failed to load model '{model_name}': {str(e)}")
        raise HTTPException(
            status_code=404,
            detail=f"Model '{model_name}' could not be loaded: {str(e)}"
        )

    try:       
        candidate_labels = []
        if zeroShotRequest.parameters:
            candidate_labels = zeroShotRequest.parameters.get('candidate_labels', [])
            if isinstance(candidate_labels, str):
                candidate_labels = [label.strip() for label in candidate_labels.split(',')]
        result = pipe(zeroShotRequest.inputs, candidate_labels=candidate_labels)
    except Exception as e:
        logger.error(f"Inference failed for model '{model_name}': {str(e)}")
        raise HTTPException(
            status_code=500,
            detail=f"Inference failed: {str(e)}"
        )

    return result



# =========================
# Image to Text Task
# =========================


async def get_encoded_image(
    request: Request
)  -> str: 
    content_type = request.headers.get("content-type", "")
    if content_type.startswith("multipart/form-data"):
        form = await request.form()
        image = form.get("image")
        if image:
            image_bytes = await image.read()
            return base64.b64encode(image_bytes).decode("utf-8")
    if content_type.startswith("image/"):
        image_bytes = await request.body()
        return base64.b64encode(image_bytes).decode("utf-8")

    raise HTTPException(status_code=400, detail="Unsupported content type")



@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.
    """

    encoded_image = await get_encoded_image(request)

    try:
        pipe = pipeline("image-to-text", model=model_name, use_fast=True)
    except Exception as e:
        logger.error(f"Failed to load model '{model_name}': {str(e)}")
        raise HTTPException(
            status_code=404,
            detail=f"Model '{model_name}' could not be loaded: {str(e)}"
        )

    try:       
        result = pipe(encoded_image)
    except Exception as e:
        logger.error(f"Inference failed for model '{model_name}': {str(e)}")
        raise HTTPException(
            status_code=500,
            detail=f"Inference failed: {str(e)}"
        )

    return result