Streaming Events

When you request a streaming response, ai& streams the model’s SSE events unchanged and then appends a single trailer event carrying cost, request ID, and final usage.

OpenAI shape

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","choices":[{"delta":{"content":"Hel"}}]}

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","choices":[{"delta":{"content":"lo"}}]}

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","choices":[{"delta":{},"finish_reason":"stop"}],"usage":{"prompt_tokens":7,"completion_tokens":2,"total_tokens":9}}

data: [DONE]

event: aiand.metadata
data: {"request_id":"req-...","cost_usd":0.000018,"model":"..."}

The aiand.metadata event is emitted after [DONE]. Robust clients should keep the connection open until the stream closes naturally.

Anthropic shape

event: message_start
data: {"type":"message_start",...}

event: content_block_delta
data: {"type":"content_block_delta",...}

event: message_delta
data: {"type":"message_delta","usage":{"output_tokens":2}}

event: message_stop
data: {"type":"message_stop"}

event: aiand.metadata
data: {"request_id":"req-...","cost_usd":0.000018,"model":"..."}

The trailer event

Field	Description
request_id	Same as X-Request-ID header.
cost_usd	Final cost in USD.
model	Model that actually served the request.
usage	(optional) Echo of the final usage block, for shapes that don’t emit it natively.

Caution

Do not parse aiand.metadata as a chat-completion chunk — it carries the event: line so clients that ignore unknown event names will skip it safely.

Why a trailer event?

ai& never modifies the response body emitted by the model. Cost, request id, and final usage are delivered as a named SSE event after the terminal message, so the byte stream above the trailer is identical to what the source API would produce — the official OpenAI and Anthropic SDKs work without modification.