# Streaming

> Stream responses as server-sent events, get token usage at the end, and learn how a disconnect is charged.

Source: https://router.africa/docs/streaming

Set `"stream": true` to receive the response as it is generated. The response is a stream of server-sent events (SSE). Each event is a line starting with `data:` that holds a JSON chunk. The stream ends with `data: [DONE]`.

**curl**

```bash
curl -N https://api.router.africa/v1/chat/completions \
  -H "Authorization: Bearer $ROUTER_AFRICA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-4o-mini",
    "stream": true,
    "stream_options": { "include_usage": true },
    "messages": [{ "role": "user", "content": "Write a two-line status message for a delayed parcel." }]
  }'
```

**Python**

```python
import os
from openai import OpenAI

client = OpenAI(base_url="https://api.router.africa/v1", api_key=os.environ["ROUTER_AFRICA_API_KEY"])

stream = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{"role": "user", "content": "Write a two-line status message for a delayed parcel."}],
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].delta.content or "", end="")
    if chunk.usage:
        print("\ntokens:", chunk.usage.total_tokens)
```

**TypeScript**

```typescript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.router.africa/v1",
  apiKey: process.env.ROUTER_AFRICA_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "openai/gpt-4o-mini",
  messages: [{ role: "user", content: "Write a two-line status message for a delayed parcel." }],
  stream: true,
  stream_options: { include_usage: true },
});

for await (const chunk of stream) {
  if (chunk.choices.length) process.stdout.write(chunk.choices[0].delta.content ?? "");
  if (chunk.usage) console.log("\ntokens:", chunk.usage.total_tokens);
}
```

## Token usage

Add `"stream_options": {"include_usage": true}` to get one last chunk before `[DONE]`. It has an empty `choices` array and a `usage` object with the tokens used.

```text
data: {"id":"gen-8f2c","choices":[],"usage":{"prompt_tokens":38,"completion_tokens":21,"total_tokens":59}}
data: [DONE]
```

## Errors during a stream

An error before the first token is a normal HTTP error. See [Errors](https://router.africa/docs/errors). Once the stream has started, the status is already `200`, so an error is sent as a final event with the same body as other errors, and then the stream ends.

```text
data: {"error":{"code":529,"message":"The model provider is overloaded. Retry shortly.","metadata":{"error_type":"provider_overloaded","reason":"provider_overloaded"}}}
```

## If you disconnect

If you close the connection before the stream ends, you are charged for the tokens used up to that point, as reported at that time, and the rest of the hold is released.

## Stream with the Anthropic SDK

`POST /v1/messages` streams in the Anthropic event format. Set the base URL to `https://api.router.africa` and pass your key as the usual `api_key`. See [Messages](https://router.africa/docs/messages) for curl, Python and TypeScript examples. Disconnecting is charged the same way as above.

## Stream with the Responses API

`POST /v1/responses` streams Responses events, each with an `event:` line and a `data:` line, and no `[DONE]` line. The `response.completed` event carries the usage. See [Responses](https://router.africa/docs/responses).