Embeddings

The AI Gateway proxies embedding requests the same way it proxies chat: point an OpenAI-compatible client at a gateway route and call its /embeddings endpoint. Embedding models are discovered per route, listed separately from chat models in the Connect tab, and traced like any other gateway request. Usage is recorded for every embeddings call; estimated cost is tracked when pricing data is available for the model, and where the charge lands depends on the provider type: built-in provider usage draws from your prepaid gateway balance, while bring-your-own-key (BYOK) usage is billed directly by the upstream provider (see Providers).

Which providers can serve embeddings

Embeddings are available for provider types whose /embeddings endpoint the gateway proxies as an OpenAI-compatible request: OpenAI, Azure Foundry, OVHcloud, Doubleword, Ollama, and custom OpenAI-compatible providers. These carry an Embeddings badge on the Providers tab.

Anthropic and Google Vertex AI don’t expose an OpenAI-compatible embeddings API, so they’re chat-only and show no badge.

The badge means the provider can serve embeddings; which embedding models are actually available depends on what the provider advertises (see below).

Discovering a route’s embedding models

Each route reports its chat and embedding models separately. List them with GET /proxy/models.

The requests on this page need two values. Create or reveal a gateway API key on the API Keys tab and use it in place of <YOUR_GATEWAY_API_KEY>; for the route, use a provider slug from the Providers tab (or a routing group slug from Routing). The examples below use openai.

Terminal

curl "https://gateway-us.pydantic.dev/proxy/models?route=openai" \
  -H "Authorization: Bearer <YOUR_GATEWAY_API_KEY>"

The response is one entry per route, with chat models under models and embedding models split out under embedding_models:

[
  {
    "route": "openai",
    "provider": "openai",
    "models": [{ "id": "gpt-5.2", "name": "gpt-5.2", "context_window": 400000 }],
    "embedding_models": [
      { "id": "text-embedding-3-small", "name": "text-embedding-3-small", "context_window": 8192 },
      { "id": "text-embedding-3-large", "name": "text-embedding-3-large", "context_window": 8192 }
    ]
  }
]

In the Connect tab, the model picker groups these into Chat models and Embedding models, and selecting an embedding model switches the generated snippets to embeddings requests. If a provider advertises no embedding models, the picker stays a flat list with no split. (Google Vertex AI doesn’t support model listing, so its Connect tab shows a free-text model box instead.)

Sending an embeddings request

Address the request to <gateway-base-url>/<route>/embeddings with an OpenAI-compatible body: a model and an input that is either a string or a list of strings. The response is the standard OpenAI shape, a data array with one embedding (a list of floats) per input, so you can embed several inputs in one request.

The examples below use the openai route in the US region; see the gateway base URLs for other regions.

Terminal

curl https://gateway-us.pydantic.dev/proxy/openai/embeddings \
  -H "Authorization: Bearer <YOUR_GATEWAY_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "text-embedding-3-small",
    "input": "The quick brown fox jumps over the lazy dog."
  }'

from openai import OpenAI

client = OpenAI(
    api_key='<YOUR_GATEWAY_API_KEY>',
    base_url='https://gateway-us.pydantic.dev/proxy/openai',
)

response = client.embeddings.create(
    model='text-embedding-3-small',
    input='The quick brown fox jumps over the lazy dog.',
)
print(response.data[0].embedding[:8])

import OpenAI from 'openai'

const client = new OpenAI({
  apiKey: '<YOUR_GATEWAY_API_KEY>',
  baseURL: 'https://gateway-us.pydantic.dev/proxy/openai',
})

const response = await client.embeddings.create({
  model: 'text-embedding-3-small',
  input: 'The quick brown fox jumps over the lazy dog.',
})
console.log(response.data[0].embedding.slice(0, 8))

import asyncio
import os

from pydantic_ai import Embedder

os.environ['PYDANTIC_AI_GATEWAY_API_KEY'] = '<YOUR_GATEWAY_API_KEY>'
os.environ['PYDANTIC_AI_GATEWAY_BASE_URL'] = 'https://gateway-us.pydantic.dev/proxy'

embedder = Embedder('gateway/openai:text-embedding-3-small')


async def main() -> None:
    result = await embedder.embed_documents(['The quick brown fox jumps over the lazy dog.'])
    print(result.embeddings[0][:8])


asyncio.run(main())

Usage is recorded on every embeddings request like any other gateway call (and estimated cost too, when pricing data is available for the model) so embeddings show up in your Spending analytics and (when telemetry is enabled) as traces alongside the rest of your LLM traffic. As with chat, built-in provider usage draws from your prepaid gateway balance while BYOK usage is billed by the upstream provider.

Embeddings

Which providers can serve embeddings

Discovering a route’s embedding models

Sending an embeddings request

See also