Thinking mode checks the field per round, not per call: an answer whose message
carried no `reasoning_content` came back 400 with the wording a call gets ("the
`reasoning_content` … must be passed back to the API"), while an empty string
satisfies the check. The injection now touches every assistant message instead
of only the ones carrying `tool_calls`.
DeepSeek dropped the `v4` prefix from its flash model, and the old id stays
valid only as a compatibility alias. Seed the default catalog with the current
name so a provider that omits `models` no longer offers a deprecated id, and
follow the rename in the wire fixtures and the minimal config template. An
explicit `deepseek-v4-flash` in TOML still passes through unchanged.
niquests exposes content as an async property; reading it without await
left a coroutine warning and skipped the body. Only load body on HTTP
errors for streams so SSE is not consumed early.
Previously DONE was skipped with continue, so keep-alive connections
could hang forever after the model finished. Also fail fast on HTTP
error statuses for stream and non-stream completions.
NCBM