Capture OpenAI-compatible reasoning and replay it in assistant history

Why: Interleaved-thinking models (MiniMax, DeepSeek R1) require the
assistant's thinking to be passed back verbatim in message history; dropping
it degrades later turns. The product's P2-I1 replayed thinking only into the
request-side rebuild; with the session transcript becoming the authoritative
context (quality P2-I2), capture must live in the kernel so thinking persists
in the transcript and replays on every request.

Evidence: SELF - TestReasoningCaptureAndReplay covers capture (SSE
reasoning_content into Message.Thinking, content unpolluted) and replay
(<think> prefix on thinking assistants, untouched non-thinking messages);
go vet + go test ./... pass.

Impact: sse.go accumulates reasoning deltas into the terminal message's
Thinking; client.go toProviderMessages renders captured thinking as a
<think> block prefix for assistant messages. Anthropic path unchanged;
messages without thinking serialize exactly as before.

runtime/llm/sse.go

@@ -52,6 +52,7 @@ func ParseSSE(ctx context.Context, r io.Reader) <-chan Event {
 
 func emitParsed(ctx context.Context, r io.Reader, out chan<- Event) {
 	var content strings.Builder
+	var reasoning strings.Builder
 	calls := map[int]*ToolCall{}
 	order := []int{}
 	var usage Usage
@@ -108,8 +109,12 @@ func emitParsed(ctx context.Context, r io.Reader, out chan<- Event) {
 		}
 		// reasoning_content（国产 OpenAI 兼容模型的思考链）与正文分流：
 		// 只作为 reasoning_delta 事件外发，不计入 Message.Content（与 runner
-		// stripThinkBlocks"正文无 think"语义对齐）。
+		// stripThinkBlocks"正文无 think"语义对齐）。同时累计进 reasoning——
+		// 交错思考模型（MiniMax/DeepSeek R1 等）要求历史 assistant 原样带回
+		// 思考内容，丢弃会降低后续轮质量；捕获进 Message.Thinking 随会话
+		// 转录持久化，请求序列化侧按需回放。
 		if choice.Delta.Reasoning != "" {
+			reasoning.WriteString(choice.Delta.Reasoning)
 			select {
 			case <-ctx.Done():
 				emitAbort(out, content.String(), ctx.Err())
@@ -166,6 +171,9 @@ func emitParsed(ctx context.Context, r io.Reader, out chan<- Event) {
 		order = nil
 	}
 	msg := Message{Role: RoleAssistant, Content: content.String(), StopReason: mapFinish(finished, len(order) > 0)}
+	if reasoning.Len() > 0 {
+		msg.Thinking = []ThinkingBlock{{Thinking: reasoning.String()}}
+	}
 	for _, idx := range order {
 		msg.ToolCalls = append(msg.ToolCalls, *calls[idx])
 	}