server.py

import json
from fastapi import FastAPI, File, UploadFile, HTTPException, status
from fastapi.middleware.cors import CORSMiddleware
from paddleocr import PaddleOCR
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain
from passporteye import read_mrz
from pydantic.v1 import BaseModel as v1BaseModel
from pydantic.v1 import Field
from pydantic import BaseModel
from typing import Any, Optional, Dict, List
from huggingface_hub import InferenceClient
from langchain.llms.base import LLM
import os


HF_token = os.getenv("apiToken")

model_name = "mistralai/Mixtral-8x7B-Instruct-v0.1"
hf_token = HF_token
kwargs = {"max_new_tokens":500, "temperature":0.1, "top_p":0.95, "repetition_penalty":1.0, "do_sample":True}

class KwArgsModel(v1BaseModel):
    kwargs: Dict[str, Any] = Field(default_factory=dict)

class CustomInferenceClient(LLM, KwArgsModel):
    model_name: str
    inference_client: InferenceClient

    def __init__(self, model_name: str, hf_token: str, kwargs: Optional[Dict[str, Any]] = None):
        inference_client = InferenceClient(model=model_name, token=hf_token)
        super().__init__(
            model_name=model_name,
            hf_token=hf_token,
            kwargs=kwargs,
            inference_client=inference_client
        )

    def _call(
        self,
        prompt: str,
        stop: Optional[List[str]] = None
    ) -> str:
        if stop is not None:
            raise ValueError("stop kwargs are not permitted.")
        response_gen = self.inference_client.text_generation(prompt, **self.kwargs, stream=True, return_full_text=False)
        response = ''.join(response_gen)  
        return response

    @property
    def _llm_type(self) -> str:
        return "custom"

    @property
    def _identifying_params(self) -> dict:
        return {"model_name": self.model_name}

app = FastAPI(title="Passport Recognition API")

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

ocr = PaddleOCR(use_angle_cls=True, lang='en')
template = """below is poorly read ocr result of a passport.
OCR Result:
{ocr_result}

Fill the below catergories using the OCR Results. you can correct spellings and make other adujustments. Dates should be in 01-JAN-2000 format.

  "countryName": "",
  "dateOfBirth": "",
  "dateOfExpiry": "",
  "dateOfIssue": "",
  "documentNumber": "",
  "givenNames": "",
  "name": "",
  "surname": "",
  "mrz": ""

json output:
"""
prompt = PromptTemplate(template=template, input_variables=["ocr_result"])

class MRZData(BaseModel):
    date_of_birth: str
    expiration_date: str
    type: str
    number: str
    names: str
    country: str
    check_number: str
    check_date_of_birth: str
    check_expiration_date: str
    check_composite: str
    check_personal_number: str
    valid_number: bool
    valid_date_of_birth: bool
    valid_expiration_date: bool
    valid_composite: bool
    valid_personal_number: bool
    method: str

class OCRData(BaseModel):
    countryName: str
    dateOfBirth: str
    dateOfExpiry: str
    dateOfIssue: str
    documentNumber: str
    givenNames: str
    name: str
    surname: str
    mrz: str

class ResponseData(BaseModel):
    documentName: str
    errorCode: int
    mrz: MRZData
    ocr: OCRData
    status: str


def create_response_data(mrz, ocr_data):
    return ResponseData(
        documentName="Passport",
        errorCode=0,
        mrz=MRZData(**mrz),
        ocr=OCRData(**ocr_data),
        status="ok"
    )


@app.post("/recognize_passport", response_model=ResponseData, status_code=status.HTTP_201_CREATED)
async def recognize_passport(image: UploadFile = File(...)):
    """Passport information extraction from a provided image file."""
    try:
        image_bytes = await image.read()
        mrz = read_mrz(image_bytes)

        img_path = 'image.jpg'
        with open(img_path, 'wb') as f:
            f.write(image_bytes)

        result = ocr.ocr(img_path, cls=True)
        json_result = []
        for idx in range(len(result)):
            res = result[idx]
            for line in res:
                coordinates, text_with_confidence = line
                text, confidence = text_with_confidence
                json_result.append({
                    'coordinates': coordinates,
                    'text': text,
                    'confidence': confidence
                })

        llm = CustomInferenceClient(model_name=model_name, hf_token=hf_token, kwargs=kwargs)
        llm_chain = LLMChain(prompt=prompt, llm=llm)
        response_str = llm_chain.run(ocr_result=json_result)
        response_str = response_str.rstrip("</s>")
        #print(response_str)

        ocr_data = json.loads(response_str)

        return create_response_data(mrz.to_dict(), ocr_data)

    except HTTPException as e:
        raise e

    except Exception as e:
        raise HTTPException(
            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
            detail=f"Internal server error: {str(e)}"
        ) from e