> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getbifrost.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Wafer

> Wafer AI conversion guide - OpenAI-compatible chat, streaming, tool calling, reasoning, text completions, and file uploads referenced by ID

## Overview

Wafer is an **OpenAI-compatible provider** with a dedicated Bifrost provider implementation. Bifrost uses the shared OpenAI-compatible request and response converters, while preserving Wafer-specific extra parameters. Key characteristics:

* **OpenAI-compatible chat** - Chat Completions use `/chat/completions`
* **Streaming support** - Server-Sent Events for chat and text completions
* **Tool calling** - Function tools are passed through using the OpenAI-compatible schema
* **Reasoning support** - Reasoning models return reasoning via `reasoning_content`
* **Responses API** - Supported by converting Responses requests to Chat Completions internally
* **Text completions** - Text completions use Wafer's `/completions` endpoint
* **File uploads** - Upload media once and reference it by `file_id` in chat requests

### Supported Operations

| Operation            | Non-Streaming | Streaming | Endpoint            |
| -------------------- | ------------- | --------- | ------------------- |
| Chat Completions     | ✅             | ✅         | `/chat/completions` |
| Responses API        | ✅             | ✅         | `/chat/completions` |
| Text Completions     | ✅             | ✅         | `/completions`      |
| List Models          | ✅             | -         | `/models`           |
| File Upload          | ✅             | -         | `/files`            |
| Embeddings           | ❌             | ❌         | -                   |
| Image Generation     | ❌             | ❌         | -                   |
| Speech (TTS)         | ❌             | ❌         | -                   |
| Transcriptions (STT) | ❌             | ❌         | -                   |
| Batch                | ❌             | ❌         | -                   |

## Setup & Configuration

Configure Wafer as a provider.

<Tabs>
  <Tab title="Web UI">
    1. Navigate to **Models** > **Model Providers**. Look for **Wafer** under **Configured Providers**. If it is missing, click on **Add New Provider** and select **Wafer**.
    2. Click **Add Key** or edit an existing key.
    3. Set a name for your key.
    4. Paste your API key directly or use an environment variable (for example, `env.WAFER_API_KEY`).
    5. Set **Allowed Models** to **All Models** (default) or the specific model allowlist you want this key to serve.
    6. To upload files, enable **Use for Batch APIs** on the key (this flag also gates file operations).
    7. Save the provider configuration.
  </Tab>

  <Tab title="config.json">
    ```json theme={null}
    {
      "providers": {
        "wafer": {
          "keys": [
            {
              "name": "wafer-key-1",
              "value": "env.WAFER_API_KEY",
              "models": [
                "*"
              ],
              "weight": 1.0,
              "use_for_batch_api": true
            }
          ]
        }
      }
    }
    ```
  </Tab>

  <Tab title="API">
    Refer to the API documentation for [Provider Keys Management](https://docs.getbifrost.ai/api-reference/providers/create-a-key-for-a-provider).
  </Tab>

  <Tab title="Go SDK">
    ```go theme={null}
    case schemas.Wafer:
        return []schemas.Key{{
            Name:           "wafer-key-1",
            Value:          *schemas.NewSecretVar("env.WAFER_API_KEY"),
            Models:         []string{"*"},
            Weight:         1.0,
            UseForBatchAPI: bifrost.Ptr(true),
        }}, nil
    ```
  </Tab>
</Tabs>

***

# 1. Chat Completions

## Request Parameters

Wafer supports OpenAI-compatible chat completion parameters. For the full parameter reference and message conversion behavior, see [OpenAI Chat Completions](/providers/supported-providers/openai#1-chat-completions).

Available chat models include `glm-5.2` and `kimi-k2.6`.

```bash theme={null}
curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wafer/glm-5.2",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
```

### Reasoning Parameter

Wafer delegates through `ToOpenAIChatRequest` with provider-specific compatibility handling. Reasoning models return reasoning content via `reasoning_content`.

Assistant-message `reasoning` details are stripped before sending follow-up messages because Wafer rejects `reasoning_details` in assistant messages.

### Extra Parameters

Wafer enables passthrough extra parameters for chat and text completion requests. Provider-specific options can be sent through `extra_params` without being dropped by Bifrost.

<Note>
  Wafer **fails fast** on unsupported parameters, returning an error rather than silently ignoring them. Bifrost filters OpenAI-specific parameters before forwarding to avoid these rejections.
</Note>

***

# 2. Responses API

Bifrost converts Responses API format to Chat Completions internally, then converts the response back:

```
BifrostResponsesRequest
  → ToChatRequest()
  → ChatCompletion
  → ToBifrostResponsesResponse()
```

Same parameter support as Chat Completions with response format differences (output items instead of message content). Streaming Responses requests are also routed through Chat Completions streaming.

***

# 3. Text Completions

Wafer supports text completions through `/completions`:

| Parameter      | Mapping                 |
| -------------- | ----------------------- |
| `prompt`       | Sent as-is              |
| `max_tokens`   | max\_tokens             |
| `temperature`  | temperature             |
| `top_p`        | top\_p                  |
| `stop`         | stop sequences          |
| `extra_params` | Passed through to Wafer |

Response returns `choices[].text` with completion text. Streaming text completions use Wafer's OpenAI-compatible SSE format on `/completions`.

***

# 4. List Models

Lists available models from Wafer through `/models`.

***

# 5. File Upload

Upload a file once and reference it by `file_id` in later chat requests, avoiding per-request body size limits. Wafer accepts **raw file bytes** with metadata in HTTP headers (not a multipart form):

* `Content-Type` - the file's MIME type (required; validated against the purpose)
* `X-Wafer-Purpose` - `vision`, `document`, or `video` (required)
* `X-Wafer-Filename` - optional label

Uploads go to `/files` and return a `file_id`. Files expire after 30 days.

<Note>
  File operations reuse the batch key filter: only keys with **Use for Batch APIs** enabled are eligible. Enable it on at least one Wafer key, or uploads fail with `no config found for batch apis`.
</Note>

## Referencing a file in a chat completion

Reference an uploaded image by nesting its `file_id` inside `image_url` (in place of `url`):

```json theme={null}
{
  "type": "image_url",
  "image_url": { "file_id": "file_8a3c4d5e6f7a8b9c0d1e2f3a4b5c6d7e" }
}
```