Responses API streaming - the simple guide to "events"

Addendum — echoed request configuration fields also present on the Response object
(these are returned so clients can introspect how the response was configured)

  model: string                                  # required [model used to generate the response; echo of request]
  instructions: string | array | null            # required [system/developer instructions used for this response; echo of request]
  tools: array                                   # required [tool definitions provided to the model; echo of request]
  tool_choice: string | object                   # required [tool selection mode / constraints; echo of request]
  metadata: object | null                        # required [developer metadata map; echo of request]
  temperature: number | null                     # required [sampling temperature; echo of request]
  top_p: number | null                           # required [nucleus sampling; echo of request]
  parallel_tool_calls: boolean                   # required [whether tool calls may run in parallel; echo of request]
  conversation: object | null                    # optional [conversation association; if present contains `{ id: string }` matching the request]

  previous_response_id: string | null            # optional [multi-turn linkage; echo of request]
  top_logprobs: integer | null                   # optional [how many per-token alternatives to return; echo of request]

  reasoning: object | null                       # optional [reasoning configuration; echo of request]
  background: boolean | null                     # optional [background processing requested; echo of request]
  max_output_tokens: integer | null              # optional [output token cap; echo of request]
  max_tool_calls: integer | null                 # optional [tool-call cap; echo of request]
  text: object | null                            # optional [text output formatting config; echo of request]

  prompt: object | null                          # optional [prompt template reference; echo of request]
  truncation: "auto" | "disabled" | null         # optional [truncation strategy; echo of request]

  safety_identifier: string                      # optional [stable safety identifier; echo of request]
  prompt_cache_key: string                       # optional [prompt caching key; echo of request]
  prompt_cache_retention: "in-memory" | "24h" | null  # optional [prompt cache retention policy; echo of request]
  user: string                                   # optional [deprecated stable end-user identifier; echo of request]

  service_tier: "auto" | "default" | "flex" | "scale" | "priority" | null  # optional [requested tier; response may reflect actual tier used]

Note: this is not the “Open Responses” new spec - this is the API.