fix: preserve reasoning_content for OpenAI-compatible reasoning models

Models like Moonshot kimi-k2.5 and DeepSeek-R1 return a
reasoning_content field in assistant messages. When thinking is enabled,
the API requires this field to be echoed back in subsequent requests.
PicoClaw was silently dropping it, causing 400 errors on tool-call
round-trips.

- Add ReasoningContent to Message and LLMResponse types
- Parse reasoning_content in openai_compat parseResponse()
- Carry reasoning_content through assistant tool-call messages
- Add unit test for reasoning_content parsing

Fixes #588

pkg/agent/loop.go

@@ -622,8 +622,9 @@ func (al *AgentLoop) runLLMIteration(
 
 		// Build assistant message with tool calls
 		assistantMsg := providers.Message{
-			Role:    "assistant",
+			Role:             "assistant",
-			Content: response.Content,
+			Content:          response.Content,
+			ReasoningContent: response.ReasoningContent,
 		}
 		for _, tc := range normalizedToolCalls {
 			argumentsJSON, _ := json.Marshal(tc.Arguments)

pkg/providers/openai_compat/provider.go

@@ -148,8 +148,9 @@ func parseResponse(body []byte) (*LLMResponse, error) {
 	var apiResponse struct {
 		Choices []struct {
 			Message struct {
-				Content   string `json:"content"`
+				Content          string `json:"content"`
-				ToolCalls []struct {
+				ReasoningContent string `json:"reasoning_content"`
+				ToolCalls        []struct {
 					ID       string `json:"id"`
 					Type     string `json:"type"`
 					Function *struct {
@@ -221,10 +222,11 @@ func parseResponse(body []byte) (*LLMResponse, error) {
 	}
 
 	return &LLMResponse{
-		Content:      choice.Message.Content,
+		Content:          choice.Message.Content,
-		ToolCalls:    toolCalls,
+		ReasoningContent: choice.Message.ReasoningContent,
-		FinishReason: choice.FinishReason,
+		ToolCalls:        toolCalls,
-		Usage:        apiResponse.Usage,
+		FinishReason:     choice.FinishReason,
+		Usage:            apiResponse.Usage,
 	}, nil
 }
 

pkg/providers/openai_compat/provider_test.go

@@ -101,6 +101,50 @@ func TestProviderChat_ParsesToolCalls(t *testing.T) {
 	}
 }
 
+func TestProviderChat_ParsesReasoningContent(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		resp := map[string]any{
+			"choices": []map[string]any{
+				{
+					"message": map[string]any{
+						"content":           "The answer is 2",
+						"reasoning_content": "Let me think step by step... 1+1=2",
+						"tool_calls": []map[string]any{
+							{
+								"id":   "call_1",
+								"type": "function",
+								"function": map[string]any{
+									"name":      "calculator",
+									"arguments": "{\"expr\":\"1+1\"}",
+								},
+							},
+						},
+					},
+					"finish_reason": "tool_calls",
+				},
+			},
+		}
+		w.Header().Set("Content-Type", "application/json")
+		json.NewEncoder(w).Encode(resp)
+	}))
+	defer server.Close()
+
+	p := NewProvider("key", server.URL, "")
+	out, err := p.Chat(t.Context(), []Message{{Role: "user", Content: "1+1=?"}}, nil, "kimi-k2.5", nil)
+	if err != nil {
+		t.Fatalf("Chat() error = %v", err)
+	}
+	if out.ReasoningContent != "Let me think step by step... 1+1=2" {
+		t.Fatalf("ReasoningContent = %q, want %q", out.ReasoningContent, "Let me think step by step... 1+1=2")
+	}
+	if out.Content != "The answer is 2" {
+		t.Fatalf("Content = %q, want %q", out.Content, "The answer is 2")
+	}
+	if len(out.ToolCalls) != 1 {
+		t.Fatalf("len(ToolCalls) = %d, want 1", len(out.ToolCalls))
+	}
+}
+
 func TestProviderChat_HTTPError(t *testing.T) {
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		http.Error(w, "bad request", http.StatusBadRequest)

pkg/providers/protocoltypes/types.go

@@ -25,10 +25,11 @@ type FunctionCall struct {
 }
 
 type LLMResponse struct {
-	Content      string     `json:"content"`
+	Content          string     `json:"content"`
-	ToolCalls    []ToolCall `json:"tool_calls,omitempty"`
+	ReasoningContent string     `json:"reasoning_content,omitempty"`
-	FinishReason string     `json:"finish_reason"`
+	ToolCalls        []ToolCall `json:"tool_calls,omitempty"`
-	Usage        *UsageInfo `json:"usage,omitempty"`
+	FinishReason     string     `json:"finish_reason"`
+	Usage            *UsageInfo `json:"usage,omitempty"`
 }
 
 type UsageInfo struct {
@@ -38,10 +39,11 @@ type UsageInfo struct {
 }
 
 type Message struct {
-	Role       string     `json:"role"`
+	Role             string     `json:"role"`
-	Content    string     `json:"content"`
+	Content          string     `json:"content"`
-	ToolCalls  []ToolCall `json:"tool_calls,omitempty"`
+	ReasoningContent string     `json:"reasoning_content,omitempty"`
-	ToolCallID string     `json:"tool_call_id,omitempty"`
+	ToolCalls        []ToolCall `json:"tool_calls,omitempty"`
+	ToolCallID       string     `json:"tool_call_id,omitempty"`
 }
 
 type ToolDefinition struct {