feat: getAssetOcr endpoint (#23331)

* feat: getAssetOcr endpoint

* pr feedback
This commit is contained in:
Alex
2025-10-28 15:57:03 -05:00
committed by GitHub
parent 8d25f81bec
commit 9098717c55
12 changed files with 560 additions and 0 deletions
+129
View File
@@ -2491,6 +2491,53 @@
"description": "This endpoint requires the `asset.read` permission."
}
},
"/assets/{id}/ocr": {
"get": {
"operationId": "getAssetOcr",
"parameters": [
{
"name": "id",
"required": true,
"in": "path",
"schema": {
"format": "uuid",
"type": "string"
}
}
],
"responses": {
"200": {
"content": {
"application/json": {
"schema": {
"items": {
"$ref": "#/components/schemas/AssetOcrResponseDto"
},
"type": "array"
}
}
},
"description": ""
}
},
"security": [
{
"bearer": []
},
{
"cookie": []
},
{
"api_key": []
}
],
"tags": [
"Assets"
],
"x-immich-permission": "asset.read",
"description": "This endpoint requires the `asset.read` permission."
}
},
"/assets/{id}/original": {
"get": {
"operationId": "downloadAsset",
@@ -11117,6 +11164,88 @@
],
"type": "object"
},
"AssetOcrResponseDto": {
"properties": {
"assetId": {
"format": "uuid",
"type": "string"
},
"boxScore": {
"description": "Confidence score for text detection box",
"format": "double",
"type": "number"
},
"id": {
"format": "uuid",
"type": "string"
},
"text": {
"description": "Recognized text",
"type": "string"
},
"textScore": {
"description": "Confidence score for text recognition",
"format": "double",
"type": "number"
},
"x1": {
"description": "Normalized x coordinate of box corner 1 (0-1)",
"format": "double",
"type": "number"
},
"x2": {
"description": "Normalized x coordinate of box corner 2 (0-1)",
"format": "double",
"type": "number"
},
"x3": {
"description": "Normalized x coordinate of box corner 3 (0-1)",
"format": "double",
"type": "number"
},
"x4": {
"description": "Normalized x coordinate of box corner 4 (0-1)",
"format": "double",
"type": "number"
},
"y1": {
"description": "Normalized y coordinate of box corner 1 (0-1)",
"format": "double",
"type": "number"
},
"y2": {
"description": "Normalized y coordinate of box corner 2 (0-1)",
"format": "double",
"type": "number"
},
"y3": {
"description": "Normalized y coordinate of box corner 3 (0-1)",
"format": "double",
"type": "number"
},
"y4": {
"description": "Normalized y coordinate of box corner 4 (0-1)",
"format": "double",
"type": "number"
}
},
"required": [
"assetId",
"boxScore",
"id",
"text",
"textScore",
"x1",
"x2",
"x3",
"x4",
"y1",
"y2",
"y3",
"y4"
],
"type": "object"
},
"AssetOrder": {
"enum": [
"asc",
@@ -546,6 +546,32 @@ export type AssetMetadataResponseDto = {
export type AssetMetadataUpsertDto = {
items: AssetMetadataUpsertItemDto[];
};
export type AssetOcrResponseDto = {
assetId: string;
/** Confidence score for text detection box */
boxScore: number;
id: string;
/** Recognized text */
text: string;
/** Confidence score for text recognition */
textScore: number;
/** Normalized x coordinate of box corner 1 (0-1) */
x1: number;
/** Normalized x coordinate of box corner 2 (0-1) */
x2: number;
/** Normalized x coordinate of box corner 3 (0-1) */
x3: number;
/** Normalized x coordinate of box corner 4 (0-1) */
x4: number;
/** Normalized y coordinate of box corner 1 (0-1) */
y1: number;
/** Normalized y coordinate of box corner 2 (0-1) */
y2: number;
/** Normalized y coordinate of box corner 3 (0-1) */
y3: number;
/** Normalized y coordinate of box corner 4 (0-1) */
y4: number;
};
export type AssetMediaReplaceDto = {
assetData: Blob;
deviceAssetId: string;
@@ -2390,6 +2416,19 @@ export function getAssetMetadataByKey({ id, key }: {
...opts
}));
}
/**
* This endpoint requires the `asset.read` permission.
*/
export function getAssetOcr({ id }: {
id: string;
}, opts?: Oazapfts.RequestOpts) {
return oazapfts.ok(oazapfts.fetchJson<{
status: 200;
data: AssetOcrResponseDto[];
}>(`/assets/${encodeURIComponent(id)}/ocr`, {
...opts
}));
}
/**
* This endpoint requires the `asset.download` permission.
*/