apps/api/app/lenses/ocr.py

from __future__ import annotations

from app.providers.ocr.base import OcrProvider
from app.storage.models import VisualAssetRecord

from .base import Lens


class OcrLens(Lens):
    name = "glm_ocr"
    view_type = "ocr"

    def __init__(self, provider: OcrProvider):
        self.provider = provider

    async def execute(self, asset: VisualAssetRecord, image_bytes: bytes, parameters: dict) -> tuple[dict, float, list[dict]]:
        payload = await self.provider.extract(image_bytes, parameters.get("region"))
        confidence = min([line.get("confidence", 0.0) for line in payload.get("lines", [])] or [0.0])
        provenance = [
            {
                "asset_id": asset.id,
                "bbox": line.get("bbox"),
                "extraction_method": self.name,
                "confidence": line.get("confidence", confidence),
                "lens_version": self.version,
            }
            for line in payload.get("lines", [])
        ]
        return payload, confidence, provenance