Create a chat completion (OpenAI-compatible)

OpenAI-compatible chat completion endpoint. Accepts the same body shape as OpenAI's `POST /v1/chat/completions` and returns the same response shape, either a JSON `ChatCompletion` (when `stream: false`, the default) or a `text/event-stream` of `ChatCompletionChunk` objects (when `stream: true`). The bearer token is either a Privy access token (the dashboard) or an `sk-` API key (services). Threads are persisted via a custom header — see the Inference module docs for the full list of non-OpenAI headers. When the upstream provider fails, a `stream: false` call returns 502 with the provider's message, and a streamed call ends with a chunk that carries an `error` object (`{ message }`) before `data: [DONE]`. A streamed call never ends without either a `finish_reason` or that error chunk. For a client-side tool loop, send every earlier `tool_calls` entry together with its `role: "tool"` result: an unmatched tool call is a request the upstream cannot serve.

curl -X POST "https://example.com/v1/chat/completions" \  -H "Content-Type: application/json" \  -d '{    "model": "string",    "messages": [      {        "role": "system",        "content": "string"      }    ]  }'

{  "id": "string",  "object": "chat.completion",  "created": 0,  "model": "string",  "choices": [    null  ],  "usage": null}