# `HuggingfaceClient.Inference.Task.ChatCompletion`
[🔗](https://github.com/huggingface/huggingface_client/blob/v0.1.0/lib/huggingface_client/inference/tasks/text_tasks.ex#L1)

Chat completion task — OpenAI-compatible `/v1/chat/completions`.

## Non-streaming

    {:ok, response} = HuggingfaceClient.chat_completion(client, %{
      model: "meta-llama/Llama-3.1-8B-Instruct",
      messages: [%{role: "user", content: "Hello!"}],
      max_tokens: 512
    })

## Streaming

    {:ok, stream} = HuggingfaceClient.chat_completion_stream(client, %{
      model: "meta-llama/Llama-3.1-8B-Instruct",
      messages: [%{role: "user", content: "Hello!"}]
    })

    for chunk <- stream do
      IO.write(get_in(chunk, ["choices", Access.at(0), "delta", "content"]) || "")
    end

## Multi-provider

    {:ok, response} = HuggingfaceClient.chat_completion(client, %{
      model: "meta-llama/Llama-3.1-8B-Instruct",
      provider: "groq",
      messages: [%{role: "user", content: "Hi!"}]
    })

# `run`

```elixir
@spec run(HuggingfaceClient.Client.t(), map()) ::
  {:ok, map()} | {:error, Exception.t()}
```

Runs a non-streaming chat completion.

Returns `{:ok, response_map}` or `{:error, exception}`.

# `stream`

```elixir
@spec stream(HuggingfaceClient.Client.t(), map()) ::
  {:ok, Enumerable.t()} | {:error, Exception.t()}
```

Runs a streaming chat completion.

Returns `{:ok, stream}` where each element is a decoded chunk map, or `{:error, exception}`.

---

*Consult [api-reference.md](api-reference.md) for complete listing*