[Feature Request] Expose context-window and output-token limits in GET /models

[yanmxa]

Current behavior

GET /models currently returns only id, object, and owned_by for each model.

Requested change

Please add optional per-model token-limit metadata to the response, for example:

{
  "id": "deepseek-chat",
  "object": "model",
  "owned_by": "deepseek",
  "context_window": 128000,
  "max_output_tokens": 8000
}

Field names may follow DeepSeek's preferred convention; the important part is that clients can discover both limits programmatically.

Semantics to document

Please clarify whether context_window is the total request context capacity, or whether callers must reserve max_output_tokens from that value.

Why this helps

Session-aware clients use this metadata to:

• display context usage accurately;
• validate requests before they exceed a model limit;
• compact conversation history safely and automatically;
• adapt when supported models or their limits change.

This lets integrations avoid hard-coded, potentially stale token-limit tables.