ImageEmbeddingsClient.py

import os
import json
import requests

class ImageEmbeddingsClient:

    def __init__(self, computer_vision_endpoint: str = None, computer_vision_key: str = None,):
        self.endpoint = computer_vision_endpoint if computer_vision_endpoint else os.getenv("AZURE_COMPUTER_VISION_ENDPOINT")
        self.key = computer_vision_key if computer_vision_key else os.getenv("AZURE_COMPUTER_VISION_KEY")

    def get_embeds(self, image_file_path: str):
        with open(image_file_path, "rb") as f:
            url = f"{self.endpoint}computervision/retrieval:vectorizeImage?api-version=2023-02-01-preview&modelVersion=latest"
            headers = {"Content-Type": "application/octet-stream", "Ocp-Apim-Subscription-Key": self.key}
            resp = requests.post(url, data=f.read(), headers=headers)
            resp.raise_for_status()
            resp = json.loads(resp.text)
            if "vector" not in resp:
                return []
            return resp["vector"]

index.json

{
    "name": "",
    "fields": [
        {
            "name": "id",
            "type": "Edm.String",
            "searchable": false,
            "filterable": false,
            "sortable": false,
            "facetable": false,
            "key": true
        },
        {
            "name": "fileName",
            "type": "Edm.String",
            "searchable": true,
            "filterable": false,
            "sortable": false,
            "facetable": false
        },
        {
            "name": "contentVector",
            "type": "Collection(Edm.Single)",
            "searchable": true,
            "dimensions": 1024,
            "vectorSearchProfile": "vectorProfile"
        }
    ],
    "vectorSearch": {
        "algorithms": [
            {
                "name": "hnsw",
                "kind": "hnsw",
                "hnswParameters": {
                    "m": 10,
                    "efConstruction": 1000,
                    "efSearch": 1000,
                    "metric": "cosine"
                }
            }
        ],
        "vectorizers": [],
        "profiles": [
            {
                "name": "vectorProfile",
                "algorithm": "hnsw"
            }
        ]
    },
    "corsOptions": {
        "allowedOrigins": [
            "*"
        ],
        "maxAgeInSeconds": 300
    },
    "similarity": {
        "@odata.type": "#Microsoft.Azure.Search.BM25Similarity",
        "k1": 1.2,
        "b": 0.75
    }
}