Chat Completions

Creates a model response for a chat conversation. This endpoint accepts OpenAI-compatible chat payloads and adds JoyToken routing, policy, wallet, and usage metadata around the provider call. When JoyToken routes a request through orchestration (multi-task planning) mode, the gateway first produces a task `plan`, executes each task with a selected sub-model, and then aggregates a final answer. In this mode: the non-streaming `choices[].message.content` is a JSON-encoded array of per-task outputs (each with `content` and `title`), the top-level `plan` lists the tasks, and `metadata` is an array with one entry per task (including the reserved `__planner__` and `__final__` tasks). When streaming, a leading planning event arrives first (a chunk whose `orchestration.phase` is `planning`, carrying the ordered `plan`) followed by the `__planner__` `metadata` event; then each task's full content arrives as a chunk that carries an `orchestration` field (`task_id`, `task_seq`, `task_status`, `title`), each followed by one standalone `metadata` event (a single object, not an array) before `[DONE]`.