Anvia Docs

DeepInfra exposes an OpenAI-compatible chat completions surface at https://api.deepinfra.com/v1/openai. In Anvia, configure OpenAIClient with that baseUrl, then pass DeepInfra model ids to completionModel(...). DeepInfra also hosts dedicated APIs for embeddings, image generation, speech, and reranking.

Create the Client

import { AgentBuilder } from "@anvia/core";
import { OpenAIClient } from "@anvia/openai";

const client = new OpenAIClient({
  baseUrl: "https://api.deepinfra.com/v1/openai",
  apiKey: process.env.DEEPINFRA_TOKEN,
});

const model = client.completionModel("deepseek-ai/DeepSeek-V3");

const agent = new AgentBuilder("support", model)
  .instructions("Answer support questions clearly.")
  .build();

const response = await agent.prompt("Hello!").send();

console.log(response.output);

baseUrl makes Anvia use the OpenAI-compatible chat completion adapter. DeepInfra model ids are namespaced as <org>/<model> and are not interchangeable with the upstream Hugging Face ids.

Get the Model List

DeepInfra exposes a /v1/models endpoint that returns the model ids available to your token. Because the client was created with baseUrl, listModels() calls DeepInfra's /models endpoint.

const models = await client.listModels();

console.table(
  models.data.map((model) => ({
    id: model.id,
    name: model.name,
    contextLength: model.contextLength,
  })),
);

Use the id field directly with completionModel(...).

Notes

DeepInfra API tokens are passed as bearer tokens in the Authorization header.
Supported parameters include model, messages, max_tokens, stream, temperature, top_p, stop, n, presence_penalty, frequency_penalty, response_format, tools, tool_choice, service_tier, and reasoning_effort. Coverage can vary slightly per model.
service_tier: "priority" requests priority inference with faster time-to-first-token and higher throughput during peak demand. Priority requests incur a 20% surcharge on top of the model's standard per-token price.
The hard cap for output tokens is 16384 for most models. To extend a response beyond the cap, send a follow-up chat completion with the previous assistant content included; DeepInfra returns 400 when the total context size is exceeded.
DeepInfra hosts a wide open-model catalog, including DeepSeek, Llama, Qwen, Mistral, Mixtral, and many others. Browse the full list at deepinfra.com/models.
DeepInfra also exposes dedicated APIs for embeddings, image generation, speech, and reranking that are not part of the OpenAI surface. Use DeepInfra's native endpoints or the OpenAI-compatible chat completions endpoint described here.

For current DeepInfra API details, see the DeepInfra OpenAI chat completions docs and the DeepInfra model catalog.