cj b4aa407601 fix(openai_compat): preserve reasoning_content for all DeepSeek assistant turns ... Remove the hasToolInteraction guard in filterDeepSeekReasoningTurn that was stripping reasoning_content from non-tool interaction assistant messages. DeepSeek V3/V4+ thinking mode rejects requests where any assistant message is missing reasoning_content, returning 400: "The reasoning_content in the thinking mode must be passed back to the API." This is a follow-up to the streaming fix that captured reasoning_content from SSE deltas. That fix alone wasn't sufficient because reasoning_content was still being stripped on replay for plain assistant turns.		2026-05-02 03:28:19 +00:00
..
provider.go	fix(openai_compat): preserve reasoning_content for all DeepSeek assistant turns	2026-05-02 03:28:19 +00:00
provider_test.go	fix(openai_compat): parse reasoning_content in streaming responses	2026-05-01 08:02:21 +00:00