### Get image embeddings with InferenceHTTPClient

Source: https://inference.roboflow.com/workflows/classifier_embeddings

Use the InferenceHTTPClient to get image embeddings from the /infer/embeddings endpoint. The example shows synchronous calls for feature vectors and logits, accessing the embeddings and info, and notes the async API has the same arguments and result format.

```python
from inference_sdk import InferenceHTTPClient

client = InferenceHTTPClient("http://localhost:9001", api_key="YOUR_API_KEY")
result = client.get_image_embeddings("product.jpg", model_id="my-project/3")
logits = client.get_image_embeddings("product.jpg", model_id="my-project/3", output_type="logits")
vector = result["embeddings"][0]
info = result["embedding_info"]

# The async API has the same arguments and result format.
# result = await client.get_image_embeddings_async("product.jpg", "my-project/3")

```

--------------------------------

### Direct model use with AutoModel.from_pretrained and embed_images

Source: https://inference.roboflow.com/workflows/classifier_embeddings

Shows how to load a model with the image_embeddings capability and generate feature vectors or logits. The output_type parameter selects which graph to preload; direct calls can override it on first use. Requires an API key and the inference-models package.

```python
from inference_models import AutoModel

model = AutoModel.from_pretrained(
    "my-project/3", api_key="YOUR_API_KEY",
    required_capabilities=["image_embeddings"],
    output_type="feature_vector",  # Or "logits"; selects which graph to preload.
)
features = model.embed_images(images, input_color_format="bgr")
# Direct calls can also override the output type, preparing that graph on first use.
logits = model.embed_images(images, output_type="logits", input_color_format="bgr")
```

--------------------------------

### InferenceHTTPClient.get_image_embeddings_async

Source: https://inference.roboflow.com/workflows/classifier_embeddings

Asynchronous SDK method to generate image embeddings. Same arguments and result format as the synchronous version.

```APIDOC
## InferenceHTTPClient.get_image_embeddings_async

### Description
Asynchronous SDK method to generate image embeddings. Same arguments and result format as the synchronous version.

### Method
SDK Call (Async)

### Endpoint
await client.get_image_embeddings_async(image, model_id, output_type=None)

### Parameters
- **image** (string) - Required - Path to the input image.
- **model_id** (string) - Required - The model identifier, e.g., "my-project/3".
- **output_type** (string) - Optional - Set to "logits" to get logits; omit for feature vectors.

### Request Example
# result = await client.get_image_embeddings_async("product.jpg", "my-project/3")

### Response
#### Success Response
- **embeddings** (array) - List of embedding vectors.
- **embedding_info** (object) - Metadata for each embedding.
- **time** (number) - Inference time.
- **metadata** (object) - Inference metadata.

#### Response Example
{
  "embeddings": [[0.1, 0.2, ...]],
  "embedding_info": {
    "model_version": "3",
    "output_type": "feature_vector",
    "feature_definition": "...",
    "vector_dimension": 768,
    "normalization": "...",
    "space_id": "..."
  },
  "time": 0.123,
  "metadata": {}
}
```

--------------------------------

### InferenceHTTPClient.get_image_embeddings

Source: https://inference.roboflow.com/workflows/classifier_embeddings

Synchronous SDK method to generate image embeddings. Accepts an image path and model ID, with optional output_type.

```APIDOC
## InferenceHTTPClient.get_image_embeddings

### Description
Synchronous SDK method to generate image embeddings. Accepts an image path and model ID, with optional output_type.

### Method
SDK Call

### Endpoint
client.get_image_embeddings(image, model_id, output_type=None)

### Parameters
- **image** (string) - Required - Path to the input image.
- **model_id** (string) - Required - The model identifier, e.g., "my-project/3".
- **output_type** (string) - Optional - Set to "logits" to get logits; omit for feature vectors.

### Request Example
from inference_sdk import InferenceHTTPClient

client = InferenceHTTPClient("http://localhost:9001", api_key="YOUR_API_KEY")
result = client.get_image_embeddings("product.jpg", model_id="my-project/3")
logits = client.get_image_embeddings("product.jpg", model_id="my-project/3", output_type="logits")

### Response
#### Success Response
- **embeddings** (array) - List of embedding vectors.
- **embedding_info** (object) - Metadata for each embedding.
- **time** (number) - Inference time.
- **metadata** (object) - Inference metadata.

#### Response Example
{
  "embeddings": [[0.1, 0.2, ...]],
  "embedding_info": {
    "model_version": "3",
    "output_type": "feature_vector",
    "feature_definition": "...",
    "vector_dimension": 768,
    "normalization": "...",
    "space_id": "..."
  },
  "time": 0.123,
  "metadata": {}
}
```

--------------------------------

### Workflow definition for embedding model

Source: https://inference.roboflow.com/workflows/classifier_embeddings

Defines a workflow step using the roboflow_core/embedding_model@v1 block with model_id resnet101 and output_type feature_vector. The outputs select the embedding and embedding_info fields.

```json
{
  "version": "1.0",
  "inputs": [{"type": "InferenceImage", "name": "image"}],
  "steps": [{
    "type": "roboflow_core/embedding_model@v1",
    "name": "embedding",
    "data": "$inputs.image",
    "model_id": "resnet101",
    "output_type": "feature_vector"
  }],
  "outputs": [
    {"type": "JsonField", "name": "embedding", "selector": "$steps.embedding.embedding"},
    {"type": "JsonField", "name": "embedding_info", "selector": "$steps.embedding.embedding_info"}
  ]
}
```

--------------------------------

### POST /infer/embeddings

Source: https://inference.roboflow.com/workflows/classifier_embeddings

Generates image embeddings from a classification model. Accepts an image or list of images and returns embeddings along with metadata.

```APIDOC
## POST /infer/embeddings

### Description
Generates image embeddings (feature vectors or logits) from a classification model. Accepts an image or list of images and returns embeddings along with metadata.

### Method
POST

### Endpoint
/infer/embeddings

### Parameters
#### Request Body
- **model_id** (string) - Required - The model identifier, e.g., "my-project/3".
- **api_key** (string) - Required - Your API key for authentication.
- **image** (image or list of images) - Required - The input image(s) in the usual Inference image format.
- **output_type** (string) - Optional - Set to "feature_vector" or "logits". Omit to default to feature vectors.
- **preprocessing overrides** (object) - Optional - Standard image preprocessing overrides; recorded in metadata.

### Request Example
{
  "model_id": "my-project/3",
  "api_key": "YOUR_API_KEY",
  "image": "product.jpg",
  "output_type": "feature_vector"
}

### Response
#### Success Response (200)
- **embeddings** (array) - List of embedding vectors.
- **embedding_info** (object) - Metadata for each embedding, including model version, output type, feature definition, vector dimension, normalization, and space_id.
- **time** (number) - Inference time.
- **metadata** (object) - Inference metadata, including preprocessing details.

#### Response Example
{
  "embeddings": [[0.1, 0.2, ...]],
  "embedding_info": {
    "model_version": "3",
    "output_type": "feature_vector",
    "feature_definition": "...",
    "vector_dimension": 768,
    "normalization": "...",
    "space_id": "..."
  },
  "time": 0.123,
  "metadata": {}
}
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.