Q4 — chat completions reasoning field name

[weselben]

Question

What field name should the chat completions response carry the
extracted reasoning under? Candidates:

• reasoning_content — DeepSeek-style, several gateways already use
• reasoning — OpenAI's own o-series response field name in the
  Responses API and newer chat completions
• both, with reasoning as primary and reasoning_content as
  legacy alias

Check what the existing providers in this repo already emit
(internal/providers/openai et al.) to avoid creating a second
canonical name. Pick one, document the choice in the PR description.

[weselben]

idk both maybe - maybe some detection of what the client can handle ? or based on already reicived context idk tbh this needs investigation!

[weselben]

Interim state: PR #41 ships with reasoning_content (the repo-wide canonical field, already used by the Anthropic/Cohere providers and all three dialect converters). Client-capability detection stays an open research question — relabeled to wayfinder:research.