fix: ensure reasoning_content on all assistant messages for DeepSeek/Kimi thinking mode

[highland0971]

Problem

DeepSeek and Volc Ark thinking mode requires every assistant message (including plain text replies) to carry a reasoning_content field. When this field is missing, the API returns HTTP 400:

{"error":{"message":"The reasoning_content in the thinking mode must be passed back to the API.","type":"invalid_request_error"}}

Root Cause

Three places where reasoning_content could be dropped:

1. agent_loop.py: conditionally added reasoning_content only when non-empty (if reasoning:)
2. run_agent.py: _copy_reasoning_content_for_api only covered tool-call assistant turns
3. chat_completions.py: convert_messages returned messages as-is when no Codex sanitization was needed — no field normalization

Fix (3 changes)

• environments/agent_loop.py L323 & L494: changed if reasoning: msg_dict["reasoning_content"] = reasoning → msg_dict["reasoning_content"] = reasoning or "" (unconditional)
• run_agent.py _copy_reasoning_content_for_api: expanded scope from tool-call-only to all assistant turns for DeepSeek/Kimi providers
• agent/transports/chat_completions.py convert_messages: added defense-in-depth normalization — ensures every role=="assistant" message has reasoning_content

Testing

Verified in production (Feishu bot with DeepSeek V4 Pro thinking mode). Previously failing conversations now work correctly. Three-layer defense ensures future changes won't re-introduce the issue.

[alt-glitch]

Likely duplicate of #15478 (merged) — same fix ensuring reasoning_content on all assistant messages for DeepSeek/Kimi thinking mode. Also overlaps with #15792 and #15982.

[teknium1]

Closing as redundant — the DeepSeek reasoning_content thinking-mode 400 and cross-provider leak chain of issues is now fully covered on main:

• Direct provider detection + same-provider tool-call pad: commits 93a2d6b, d58b305, ad0ac89, 5ae6081
• All-assistant-messages pad rule: commit ad0ac89
• Ordering / cross-provider isolation: commit 9daa062 + regression guard 63bf7a2
• Cross-provider leak (MiniMax reasoning → DeepSeek): PR fix(agent): block cross-provider reasoning leak to DeepSeek/Kimi (#15748) #16500 (merging shortly)

21 regression tests in tests/run_agent/test_deepseek_reasoning_content_echo.py + 2 new tests for the cross-provider scenario exercise every known path. Thanks for the submission — appreciate the digging on this area.