Add files via upload

internal/attackchain/builder.go

@@ -301,7 +301,7 @@ func (b *Builder) formatProcessDetailsForAttackChain(details []database.ProcessD
 		// 目标：以主 agent（编排器）视角输出整轮迭代
 		// - 保留：编排器工具调用/结果、对子代理的 task 调度、子代理最终回复（不含推理）
 		// - 丢弃：thinking/planning/progress 等噪声、子代理的工具细节与推理过程
-		if d.EventType == "progress" || d.EventType == "thinking" || d.EventType == "planning" {
+		if d.EventType == "progress" || d.EventType == "thinking" || d.EventType == "reasoning_chain" || d.EventType == "planning" {
 			continue
 		}
 

internal/handler/agent.go

@@ -19,6 +19,7 @@ import (
 	"cyberstrike-ai/internal/agent"
 	"cyberstrike-ai/internal/config"
 	"cyberstrike-ai/internal/database"
+	"cyberstrike-ai/internal/reasoning"
 	"cyberstrike-ai/internal/mcp"
 	"cyberstrike-ai/internal/mcp/builtin"
 	"cyberstrike-ai/internal/multiagent"
@@ -201,6 +202,14 @@ type ChatAttachment struct {
 	ServerPath string `json:"serverPath,omitempty"` // 已保存在 chat_uploads 下的绝对路径（由 POST /api/chat-uploads 返回）
 }
 
+// ChatReasoningRequest 对话页「模型推理」意图（仅 Eino 路径消费；原生 agent-loop 忽略）。
+type ChatReasoningRequest struct {
+	// Mode: default（跟随系统）| off | on | auto
+	Mode string `json:"mode,omitempty"`
+	// Effort: low | medium | high | max；空表示不指定（由系统默认与各 profile 决定）。
+	Effort string `json:"effort,omitempty"`
+}
+
 // ChatRequest 聊天请求
 type ChatRequest struct {
 	Message              string           `json:"message" binding:"required"`
@@ -209,10 +218,18 @@ type ChatRequest struct {
 	Attachments          []ChatAttachment `json:"attachments,omitempty"`
 	WebShellConnectionID string           `json:"webshellConnectionId,omitempty"` // WebShell 管理 - AI 助手：当前选中的连接 ID，仅使用 webshell_* 工具
 	Hitl                 *HITLRequest     `json:"hitl,omitempty"`
+	Reasoning            *ChatReasoningRequest `json:"reasoning,omitempty"`
 	// Orchestration 仅对 /api/multi-agent、/api/multi-agent/stream：deep | plan_execute | supervisor；空则等同 deep。机器人/批量等无请求体时由服务端默认 deep。/api/eino-agent* 不使用此字段。
 	Orchestration string `json:"orchestration,omitempty"`
 }
 
+func chatReasoningToClientIntent(r *ChatReasoningRequest) *reasoning.ClientIntent {
+	if r == nil {
+		return nil
+	}
+	return &reasoning.ClientIntent{Mode: r.Mode, Effort: r.Effort}
+}
+
 type HITLRequest struct {
 	Enabled        bool     `json:"enabled"`
 	Mode           string   `json:"mode,omitempty"`
@@ -567,14 +584,7 @@ func (h *AgentHandler) AgentLoop(c *gin.Context) {
 			h.logger.Warn("获取历史消息失败", zap.Error(err))
 			agentHistoryMessages = []agent.ChatMessage{}
 		} else {
-			// 将数据库消息转换为Agent消息格式
-			agentHistoryMessages = make([]agent.ChatMessage, 0, len(historyMessages))
-			for _, msg := range historyMessages {
-				agentHistoryMessages = append(agentHistoryMessages, agent.ChatMessage{
-					Role:    msg.Role,
-					Content: msg.Content,
-				})
-			}
+			agentHistoryMessages = dbMessagesToAgentChatMessages(historyMessages)
 			h.logger.Info("从消息表加载历史消息", zap.Int("count", len(agentHistoryMessages)))
 		}
 	} else {
@@ -775,6 +785,7 @@ func (h *AgentHandler) ProcessMessageForRobot(ctx context.Context, conversationI
 			progressCallback,
 			h.agentsMarkdownDir,
 			"deep",
+			nil,
 		)
 		if errMA != nil {
 			if shouldPersistEinoAgentTraceAfterRunError(ctx) {
@@ -788,17 +799,8 @@ func (h *AgentHandler) ProcessMessageForRobot(ctx context.Context, conversationI
 			return "", conversationID, errMA
 		}
 		if assistantMessageID != "" {
-			mcpIDsJSON := ""
-			if len(resultMA.MCPExecutionIDs) > 0 {
-				jsonData, _ := json.Marshal(resultMA.MCPExecutionIDs)
-				mcpIDsJSON = string(jsonData)
-			}
-			_, err = h.db.Exec(
-				"UPDATE messages SET content = ?, mcp_execution_ids = ?, updated_at = ? WHERE id = ?",
-				resultMA.Response, mcpIDsJSON, time.Now(), assistantMessageID,
-			)
-			if err != nil {
-				h.logger.Warn("机器人：更新助手消息失败", zap.Error(err))
+			if errU := h.db.UpdateAssistantMessageFinalize(assistantMessageID, resultMA.Response, resultMA.MCPExecutionIDs, multiagent.AggregatedReasoningFromTraceJSON(resultMA.LastAgentTraceInput)); errU != nil {
+				h.logger.Warn("机器人：更新助手消息失败", zap.Error(errU))
 			}
 		} else {
 			if _, err = h.db.AddMessage(conversationID, "assistant", resultMA.Response, resultMA.MCPExecutionIDs); err != nil {
@@ -823,17 +825,8 @@ func (h *AgentHandler) ProcessMessageForRobot(ctx context.Context, conversationI
 
 	// 更新助手消息内容与 MCP 执行 ID（与 stream 一致）
 	if assistantMessageID != "" {
-		mcpIDsJSON := ""
-		if len(result.MCPExecutionIDs) > 0 {
-			jsonData, _ := json.Marshal(result.MCPExecutionIDs)
-			mcpIDsJSON = string(jsonData)
-		}
-		_, err = h.db.Exec(
-		"UPDATE messages SET content = ?, mcp_execution_ids = ?, updated_at = ? WHERE id = ?",
-		result.Response, mcpIDsJSON, time.Now(), assistantMessageID,
-		)
-		if err != nil {
-			h.logger.Warn("机器人：更新助手消息失败", zap.Error(err))
+		if errU := h.db.UpdateAssistantMessageFinalize(assistantMessageID, result.Response, result.MCPExecutionIDs, multiagent.AggregatedReasoningFromTraceJSON(result.LastAgentTraceInput)); errU != nil {
+			h.logger.Warn("机器人：更新助手消息失败", zap.Error(errU))
 		}
 	} else {
 		if _, err = h.db.AddMessage(conversationID, "assistant", result.Response, result.MCPExecutionIDs); err != nil {
@@ -891,10 +884,12 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 		return ""
 	}
 
-	// thinking_stream_*：不逐条落库，按 streamId 聚合，在后续关键事件前补一条可持久化的 thinking
+	// thinking_stream_*（ReAct 等助手正文流）与 reasoning_chain_stream_*（Eino ReasoningContent）：
+	// 不逐条落库，按 streamId 聚合，flush 时分别落 thinking / reasoning_chain。
 	type thinkingBuf struct {
-		b    strings.Builder
-		meta map[string]interface{}
+		b         strings.Builder
+		meta      map[string]interface{}
+		persistAs string // "thinking" | "reasoning_chain"
 	}
 	thinkingStreams := make(map[string]*thinkingBuf) // streamId -> buf
 	flushedThinking := make(map[string]bool)         // streamId -> flushed
@@ -948,8 +943,12 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 				}
 				data[k] = v
 			}
-			if err := h.db.AddProcessDetail(assistantMessageID, conversationID, "thinking", content, data); err != nil {
-				h.logger.Warn("保存过程详情失败", zap.Error(err), zap.String("eventType", "thinking"))
+			persist := tb.persistAs
+			if persist != "reasoning_chain" {
+				persist = "thinking"
+			}
+			if err := h.db.AddProcessDetail(assistantMessageID, conversationID, persist, content, data); err != nil {
+				h.logger.Warn("保存过程详情失败", zap.Error(err), zap.String("eventType", persist))
 			}
 			flushedThinking[sid] = true
 		}
@@ -1177,14 +1176,20 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 			return
 		}
 
-		// 聚合 thinking_stream_*（ReasoningContent），不逐条落库
-		if eventType == "thinking_stream_start" {
+		// 聚合 thinking_stream_* / reasoning_chain_stream_*，不逐条落库
+		if eventType == "thinking_stream_start" || eventType == "reasoning_chain_stream_start" {
+			persistAs := "thinking"
+			if eventType == "reasoning_chain_stream_start" {
+				persistAs = "reasoning_chain"
+			}
 			if dataMap, ok := data.(map[string]interface{}); ok {
 				if sid, ok2 := dataMap["streamId"].(string); ok2 && sid != "" {
 					tb := thinkingStreams[sid]
 					if tb == nil {
-						tb = &thinkingBuf{meta: map[string]interface{}{}}
+						tb = &thinkingBuf{meta: map[string]interface{}{}, persistAs: persistAs}
 						thinkingStreams[sid] = tb
+					} else {
+						tb.persistAs = persistAs
 					}
 					// 记录元信息（source/einoAgent/einoRole/iteration 等）
 					for k, v := range dataMap {
@@ -1194,15 +1199,21 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 			}
 			return
 		}
-		if eventType == "thinking_stream_delta" {
+		if eventType == "thinking_stream_delta" || eventType == "reasoning_chain_stream_delta" {
+			persistAs := "thinking"
+			if eventType == "reasoning_chain_stream_delta" {
+				persistAs = "reasoning_chain"
+			}
 			if dataMap, ok := data.(map[string]interface{}); ok {
 				if sid, ok2 := dataMap["streamId"].(string); ok2 && sid != "" {
 					tb := thinkingStreams[sid]
 					if tb == nil {
-						tb = &thinkingBuf{meta: map[string]interface{}{}}
+						tb = &thinkingBuf{meta: map[string]interface{}{}, persistAs: persistAs}
 						thinkingStreams[sid] = tb
+					} else if tb.persistAs == "" {
+						tb.persistAs = persistAs
 					}
-					// delta 片段直接拼接；message 本身就是 reasoning content
+					// delta 片段直接拼接
 					tb.b.WriteString(message)
 					// 有时 delta 先到 start 未到，补充元信息
 					for k, v := range dataMap {
@@ -1213,10 +1224,9 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 			return
 		}
 
-		// 当 Agent 同时发送 thinking_stream_* 和 thinking（带同一 streamId）时，
-		// thinking_stream_* 已经会在 flushThinkingStreams() 聚合落库；
-		// 这里跳过同 streamId 的 thinking，避免 processDetails 双份展示。
-		if eventType == "thinking" {
+		// 当 Agent 同时发送 *_stream_* 与同名 streamId 的 thinking/reasoning_chain 时，
+		// 流式聚合已会在 flushThinkingStreams() 落库；此处跳过逐条重复。
+		if eventType == "thinking" || eventType == "reasoning_chain" {
 			if dataMap, ok := data.(map[string]interface{}); ok {
 				if sid, ok2 := dataMap["streamId"].(string); ok2 && sid != "" {
 					if tb, exists := thinkingStreams[sid]; exists && tb != nil {
@@ -1245,7 +1255,7 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 			if eventType == "tool_result" {
 				discardPlanningIfEchoesToolResult(&respPlan, data)
 			}
-			// 在关键过程事件落库前，先把「规划中」与 thinking_stream 落库
+			// 在关键过程事件落库前，先把「规划中」与聚合中的 thinking / reasoning_chain 流落库
 			flushResponsePlan()
 			flushThinkingStreams()
 			if err := h.db.AddProcessDetail(assistantMessageID, conversationID, eventType, message, data); err != nil {
@@ -1427,14 +1437,7 @@ func (h *AgentHandler) AgentLoopStream(c *gin.Context) {
 			h.logger.Warn("获取历史消息失败", zap.Error(err))
 			agentHistoryMessages = []agent.ChatMessage{}
 		} else {
-			// 将数据库消息转换为Agent消息格式
-			agentHistoryMessages = make([]agent.ChatMessage, 0, len(historyMessages))
-			for _, msg := range historyMessages {
-				agentHistoryMessages = append(agentHistoryMessages, agent.ChatMessage{
-					Role:    msg.Role,
-					Content: msg.Content,
-				})
-			}
+			agentHistoryMessages = dbMessagesToAgentChatMessages(historyMessages)
 			h.logger.Info("从消息表加载历史消息", zap.Int("count", len(agentHistoryMessages)))
 		}
 	} else {
@@ -1727,20 +1730,8 @@ func (h *AgentHandler) AgentLoopStream(c *gin.Context) {
 
 	// 更新助手消息内容
 	if assistantMsg != nil {
-		_, err = h.db.Exec(
-			"UPDATE messages SET content = ?, mcp_execution_ids = ?, updated_at = ? WHERE id = ?",
-			result.Response,
-			func() string {
-				if len(result.MCPExecutionIDs) > 0 {
-					jsonData, _ := json.Marshal(result.MCPExecutionIDs)
-					return string(jsonData)
-				}
-				return ""
-			}(),
-			time.Now(), assistantMessageID,
-		)
-		if err != nil {
-			h.logger.Error("更新助手消息失败", zap.Error(err))
+		if errU := h.db.UpdateAssistantMessageFinalize(assistantMessageID, result.Response, result.MCPExecutionIDs, multiagent.AggregatedReasoningFromTraceJSON(result.LastAgentTraceInput)); errU != nil {
+			h.logger.Error("更新助手消息失败", zap.Error(errU))
 		}
 	} else {
 		// 如果之前创建失败，现在创建
@@ -2664,12 +2655,12 @@ func (h *AgentHandler) executeBatchQueue(queueID string) {
 			var runErr error
 			switch {
 			case useBatchMulti:
-				resultMA, runErr = multiagent.RunDeepAgent(taskCtx, h.config, &h.config.MultiAgent, h.agent, h.logger, conversationID, finalMessage, []agent.ChatMessage{}, roleTools, progressCallback, h.agentsMarkdownDir, batchOrch)
+				resultMA, runErr = multiagent.RunDeepAgent(taskCtx, h.config, &h.config.MultiAgent, h.agent, h.logger, conversationID, finalMessage, []agent.ChatMessage{}, roleTools, progressCallback, h.agentsMarkdownDir, batchOrch, nil)
 			case useEinoSingle:
 				if h.config == nil {
 					runErr = fmt.Errorf("服务器配置未加载")
 				} else {
-					resultMA, runErr = multiagent.RunEinoSingleChatModelAgent(taskCtx, h.config, &h.config.MultiAgent, h.agent, h.logger, conversationID, finalMessage, []agent.ChatMessage{}, roleTools, progressCallback)
+					resultMA, runErr = multiagent.RunEinoSingleChatModelAgent(taskCtx, h.config, &h.config.MultiAgent, h.agent, h.logger, conversationID, finalMessage, []agent.ChatMessage{}, roleTools, progressCallback, nil)
 				}
 			default:
 				result, runErr = h.agent.AgentLoopWithProgress(taskCtx, finalMessage, []agent.ChatMessage{}, conversationID, progressCallback, roleTools)
@@ -2768,17 +2759,7 @@ func (h *AgentHandler) executeBatchQueue(queueID string) {
 
 			// 更新助手消息内容
 			if assistantMessageID != "" {
-				mcpIDsJSON := ""
-				if len(mcpIDs) > 0 {
-					jsonData, _ := json.Marshal(mcpIDs)
-					mcpIDsJSON = string(jsonData)
-				}
-				if _, updateErr := h.db.Exec(
-					"UPDATE messages SET content = ?, mcp_execution_ids = ?, updated_at = ? WHERE id = ?",
-					resText,
-					mcpIDsJSON,
-					time.Now(), assistantMessageID,
-				); updateErr != nil {
+				if updateErr := h.db.UpdateAssistantMessageFinalize(assistantMessageID, resText, mcpIDs, multiagent.AggregatedReasoningFromTraceJSON(lastIn)); updateErr != nil {
 					h.logger.Warn("更新助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(updateErr))
 					// 如果更新失败，尝试创建新消息
 					_, err = h.db.AddMessage(conversationID, "assistant", resText, mcpIDs)
@@ -2870,6 +2851,10 @@ func (h *AgentHandler) loadHistoryFromAgentTrace(conversationID string) ([]agent
 		if content, ok := msgMap["content"].(string); ok {
 			msg.Content = content
 		}
+		// DeepSeek 思考模式：含工具调用的 assistant 须在后续请求中回传 reasoning_content
+		if rc, ok := msgMap["reasoning_content"].(string); ok && strings.TrimSpace(rc) != "" {
+			msg.ReasoningContent = rc
+		}
 
 		// 解析tool_calls（如果存在）
 		if toolCallsRaw, ok := msgMap["tool_calls"]; ok && toolCallsRaw != nil {
@@ -2975,3 +2960,18 @@ func (h *AgentHandler) loadHistoryFromAgentTrace(conversationID string) ([]agent
 	)
 	return agentMessages, nil
 }
+
+// dbMessagesToAgentChatMessages maps DB rows to agent ChatMessage for history fallback
+// (includes reasoning_content for DeepSeek thinking + tool replay).
+func dbMessagesToAgentChatMessages(msgs []database.Message) []agent.ChatMessage {
+	out := make([]agent.ChatMessage, 0, len(msgs))
+	for i := range msgs {
+		m := msgs[i]
+		out = append(out, agent.ChatMessage{
+			Role:             m.Role,
+			Content:          m.Content,
+			ReasoningContent: m.ReasoningContent,
+		})
+	}
+	return out
+}

internal/handler/config.go

@@ -1312,6 +1312,19 @@ func updateOpenAIConfig(doc *yaml.Node, cfg config.OpenAIConfig) {
 	if cfg.MaxTotalTokens > 0 {
 		setIntInMap(openaiNode, "max_total_tokens", cfg.MaxTotalTokens)
 	}
+	rn := ensureMap(openaiNode, "reasoning")
+	if strings.TrimSpace(cfg.Reasoning.Mode) != "" {
+		setStringInMap(rn, "mode", cfg.Reasoning.Mode)
+	}
+	if strings.TrimSpace(cfg.Reasoning.Effort) != "" {
+		setStringInMap(rn, "effort", cfg.Reasoning.Effort)
+	}
+	if cfg.Reasoning.AllowClientReasoning != nil {
+		setBoolInMap(rn, "allow_client_reasoning", *cfg.Reasoning.AllowClientReasoning)
+	}
+	if strings.TrimSpace(cfg.Reasoning.Profile) != "" {
+		setStringInMap(rn, "profile", cfg.Reasoning.Profile)
+	}
 }
 
 func updateFOFAConfig(doc *yaml.Node, cfg config.FofaConfig) {

internal/handler/eino_single_agent.go

@@ -196,6 +196,7 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 			curHistory,
 			roleTools,
 			progressCallback,
+			chatReasoningToClientIntent(req.Reasoning),
 		)
 		timeoutCancel()
 
@@ -297,18 +298,7 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 	}
 
 	if assistantMessageID != "" {
-		mcpIDsJSON := ""
-		if len(cumulativeMCPExecutionIDs) > 0 {
-			jsonData, _ := json.Marshal(cumulativeMCPExecutionIDs)
-			mcpIDsJSON = string(jsonData)
-		}
-		_, _ = h.db.Exec(
-			"UPDATE messages SET content = ?, mcp_execution_ids = ?, updated_at = ? WHERE id = ?",
-			result.Response,
-			mcpIDsJSON,
-			time.Now(),
-			assistantMessageID,
-		)
+		_ = h.db.UpdateAssistantMessageFinalize(assistantMessageID, result.Response, cumulativeMCPExecutionIDs, multiagent.AggregatedReasoningFromTraceJSON(result.LastAgentTraceInput))
 	}
 
 	if result.LastAgentTraceInput != "" || result.LastAgentTraceOutput != "" {
@@ -376,6 +366,7 @@ func (h *AgentHandler) EinoSingleAgentLoop(c *gin.Context) {
 		prep.History,
 		prep.RoleTools,
 		progressCallback,
+		chatReasoningToClientIntent(req.Reasoning),
 	)
 	if runErr != nil {
 		if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
@@ -386,18 +377,7 @@ func (h *AgentHandler) EinoSingleAgentLoop(c *gin.Context) {
 	}
 
 	if prep.AssistantMessageID != "" {
-		mcpIDsJSON := ""
-		if len(result.MCPExecutionIDs) > 0 {
-			jsonData, _ := json.Marshal(result.MCPExecutionIDs)
-			mcpIDsJSON = string(jsonData)
-		}
-		_, _ = h.db.Exec(
-			"UPDATE messages SET content = ?, mcp_execution_ids = ?, updated_at = ? WHERE id = ?",
-			result.Response,
-			mcpIDsJSON,
-			time.Now(),
-			prep.AssistantMessageID,
-		)
+		_ = h.db.UpdateAssistantMessageFinalize(prep.AssistantMessageID, result.Response, result.MCPExecutionIDs, multiagent.AggregatedReasoningFromTraceJSON(result.LastAgentTraceInput))
 	}
 	if result.LastAgentTraceInput != "" || result.LastAgentTraceOutput != "" {
 		_ = h.db.SaveAgentTrace(prep.ConversationID, result.LastAgentTraceInput, result.LastAgentTraceOutput)

internal/handler/multi_agent.go

@@ -208,6 +208,7 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 			progressCallback,
 			h.agentsMarkdownDir,
 			orch,
+			chatReasoningToClientIntent(req.Reasoning),
 		)
 		timeoutCancel()
 
@@ -309,18 +310,7 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 	}
 
 	if assistantMessageID != "" {
-		mcpIDsJSON := ""
-		if len(cumulativeMCPExecutionIDs) > 0 {
-			jsonData, _ := json.Marshal(cumulativeMCPExecutionIDs)
-			mcpIDsJSON = string(jsonData)
-		}
-		_, _ = h.db.Exec(
-			"UPDATE messages SET content = ?, mcp_execution_ids = ?, updated_at = ? WHERE id = ?",
-			result.Response,
-			mcpIDsJSON,
-			time.Now(),
-			assistantMessageID,
-		)
+		_ = h.db.UpdateAssistantMessageFinalize(assistantMessageID, result.Response, cumulativeMCPExecutionIDs, multiagent.AggregatedReasoningFromTraceJSON(result.LastAgentTraceInput))
 	}
 
 	if result.LastAgentTraceInput != "" || result.LastAgentTraceOutput != "" {
@@ -390,6 +380,7 @@ func (h *AgentHandler) MultiAgentLoop(c *gin.Context) {
 		progressCallback,
 		h.agentsMarkdownDir,
 		strings.TrimSpace(req.Orchestration),
+		chatReasoningToClientIntent(req.Reasoning),
 	)
 	if runErr != nil {
 		if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
@@ -405,18 +396,7 @@ func (h *AgentHandler) MultiAgentLoop(c *gin.Context) {
 	}
 
 	if prep.AssistantMessageID != "" {
-		mcpIDsJSON := ""
-		if len(result.MCPExecutionIDs) > 0 {
-			jsonData, _ := json.Marshal(result.MCPExecutionIDs)
-			mcpIDsJSON = string(jsonData)
-		}
-		_, _ = h.db.Exec(
-			"UPDATE messages SET content = ?, mcp_execution_ids = ?, updated_at = ? WHERE id = ?",
-			result.Response,
-			mcpIDsJSON,
-			time.Now(),
-			prep.AssistantMessageID,
-		)
+		_ = h.db.UpdateAssistantMessageFinalize(prep.AssistantMessageID, result.Response, result.MCPExecutionIDs, multiagent.AggregatedReasoningFromTraceJSON(result.LastAgentTraceInput))
 	}
 
 	if result.LastAgentTraceInput != "" || result.LastAgentTraceOutput != "" {

internal/handler/multi_agent_prepare.go

@@ -55,13 +55,7 @@ func (h *AgentHandler) prepareMultiAgentSession(req *ChatRequest) (*multiAgentPr
 		if getErr != nil {
 			agentHistoryMessages = []agent.ChatMessage{}
 		} else {
-			agentHistoryMessages = make([]agent.ChatMessage, 0, len(historyMessages))
-			for _, msg := range historyMessages {
-				agentHistoryMessages = append(agentHistoryMessages, agent.ChatMessage{
-					Role:    msg.Role,
-					Content: msg.Content,
-				})
-			}
+			agentHistoryMessages = dbMessagesToAgentChatMessages(historyMessages)
 		}
 	}
 

internal/reasoning/eino.go

@@ -0,0 +1,250 @@
+// Package reasoning maps user/config intent to CloudWeGo Eino OpenAI ChatModel fields
+// (ReasoningEffort, ExtraFields such as thinking / reasoning_effort / output_config).
+package reasoning
+
+import (
+	"strings"
+
+	"cyberstrike-ai/internal/config"
+
+	einoopenai "github.com/cloudwego/eino-ext/components/model/openai"
+)
+
+// ClientIntent is optional per-request override from ChatRequest.reasoning.
+type ClientIntent struct {
+	Mode   string
+	Effort string
+}
+
+type wireProfile int
+
+const (
+	wireNone wireProfile = iota
+	wireClaude
+	wireDeepseek
+	wireOpenAI
+	wireOutputConfig
+)
+
+// ApplyToEinoChatModelConfig merges reasoning-related options into cfg.
+// Precondition: cfg already has APIKey, BaseURL, Model, HTTPClient set.
+func ApplyToEinoChatModelConfig(cfg *einoopenai.ChatModelConfig, oa *config.OpenAIConfig, client *ClientIntent) {
+	if cfg == nil || oa == nil {
+		return
+	}
+	sr := &oa.Reasoning
+	allowClient := sr.AllowClientReasoningEffective()
+	mode := effectiveMode(sr, client, allowClient)
+
+	// Claude (Anthropic): merge admin extras first; optional extended thinking maps to top-level `thinking`
+	// (see internal/openai convertOpenAIToClaude). DeepSeek/OpenAI-style fields are not sent.
+	if strings.EqualFold(strings.TrimSpace(oa.Provider), "claude") ||
+		strings.EqualFold(strings.TrimSpace(oa.Provider), "anthropic") {
+		if len(sr.ExtraRequestFields) > 0 {
+			if cfg.ExtraFields == nil {
+				cfg.ExtraFields = make(map[string]any)
+			}
+			for k, v := range sr.ExtraRequestFields {
+				cfg.ExtraFields[k] = v
+			}
+		}
+		if mode == "off" {
+			return
+		}
+		applyClaudeExtendedThinking(cfg, mode, effectiveEffort(sr, client, allowClient), oa.Model)
+		return
+	}
+
+	if mode == "off" {
+		return
+	}
+	effort := effectiveEffort(sr, client, allowClient)
+	prof := resolveWireProfile(oa, sr)
+
+	// Admin-defined extra root fields (merged first; automatic keys may follow).
+	if len(sr.ExtraRequestFields) > 0 {
+		if cfg.ExtraFields == nil {
+			cfg.ExtraFields = make(map[string]any)
+		}
+		for k, v := range sr.ExtraRequestFields {
+			cfg.ExtraFields[k] = v
+		}
+	}
+
+	switch prof {
+	case wireClaude, wireNone:
+		return
+	case wireDeepseek:
+		applyDeepseek(cfg, mode, effort)
+	case wireOutputConfig:
+		applyOutputConfigEffort(cfg, mode, effort)
+	default: // wireOpenAI
+		applyOpenAICompat(cfg, mode, effort)
+	}
+}
+
+// applyClaudeExtendedThinking sets Anthropic Messages API `thinking` when absent from ExtraRequestFields.
+// Uses adaptive + summarized display by default (per Anthropic guidance for Claude 4.x); Sonnet 3.7 uses enabled+budget.
+func applyClaudeExtendedThinking(cfg *einoopenai.ChatModelConfig, mode, effort, model string) {
+	if cfg == nil || mode == "off" {
+		return
+	}
+	if cfg.ExtraFields == nil {
+		cfg.ExtraFields = make(map[string]any)
+	}
+	if _, exists := cfg.ExtraFields["thinking"]; exists {
+		return
+	}
+	m := strings.ToLower(strings.TrimSpace(model))
+	thinking := map[string]any{
+		"type":    "adaptive",
+		"display": "summarized",
+	}
+	// Sonnet 3.7: manual extended thinking is the documented path.
+	if strings.Contains(m, "claude-3-7-sonnet") || strings.Contains(m, "3-7-sonnet") || strings.Contains(m, "sonnet-3.7") {
+		thinking = map[string]any{
+			"type":          "enabled",
+			"budget_tokens": 10000,
+			"display":       "summarized",
+		}
+	}
+	// Opus 4.7+: manual enabled+budget rejected — keep adaptive only.
+	if strings.Contains(m, "opus-4-7") || strings.Contains(m, "opus-4.7") {
+		thinking = map[string]any{
+			"type":    "adaptive",
+			"display": "summarized",
+		}
+	}
+	_ = effort // reserved: map to Anthropic effort / output_config when API stabilizes in one place
+	cfg.ExtraFields["thinking"] = thinking
+}
+
+func effectiveMode(sr *config.OpenAIReasoningConfig, client *ClientIntent, allowClient bool) string {
+	server := strings.ToLower(strings.TrimSpace(sr.ModeEffective()))
+	if server == "" || server == "default" {
+		server = "auto"
+	}
+	if !allowClient || client == nil {
+		return server
+	}
+	cm := strings.ToLower(strings.TrimSpace(client.Mode))
+	if cm == "" || cm == "default" {
+		return server
+	}
+	return cm
+}
+
+func effectiveEffort(sr *config.OpenAIReasoningConfig, client *ClientIntent, allowClient bool) string {
+	se := normalizeEffort(sr.Effort)
+	if !allowClient || client == nil {
+		return se
+	}
+	ce := normalizeEffort(client.Effort)
+	if ce != "" {
+		return ce
+	}
+	return se
+}
+
+func normalizeEffort(s string) string {
+	e := strings.ToLower(strings.TrimSpace(s))
+	switch e {
+	case "low", "medium", "high", "max":
+		return e
+	default:
+		return ""
+	}
+}
+
+func resolveWireProfile(oa *config.OpenAIConfig, sr *config.OpenAIReasoningConfig) wireProfile {
+	if strings.EqualFold(strings.TrimSpace(oa.Provider), "claude") {
+		return wireClaude
+	}
+	p := strings.ToLower(strings.TrimSpace(sr.ProfileEffective()))
+	switch p {
+	case "output_config", "output_config_effort":
+		return wireOutputConfig
+	case "openai", "openai_compat":
+		return wireOpenAI
+	case "deepseek", "deepseek_compat":
+		return wireDeepseek
+	case "auto", "":
+		bu := strings.ToLower(oa.BaseURL)
+		mo := strings.ToLower(oa.Model)
+		if strings.Contains(bu, "deepseek") || strings.Contains(mo, "deepseek") {
+			return wireDeepseek
+		}
+		return wireOpenAI
+	default:
+		return wireOpenAI
+	}
+}
+
+func applyDeepseek(cfg *einoopenai.ChatModelConfig, mode, effort string) {
+	// auto: enable thinking for DeepSeek line; on: same; auto without effort still opens thinking.
+	if mode == "off" {
+		return
+	}
+	if mode == "auto" || mode == "on" {
+		if cfg.ExtraFields == nil {
+			cfg.ExtraFields = make(map[string]any)
+		}
+		cfg.ExtraFields["thinking"] = map[string]any{"type": "enabled"}
+	}
+	if effort != "" {
+		if cfg.ExtraFields == nil {
+			cfg.ExtraFields = make(map[string]any)
+		}
+		cfg.ExtraFields["reasoning_effort"] = effortStringForAPI(effort)
+	}
+}
+
+func applyOpenAICompat(cfg *einoopenai.ChatModelConfig, mode, effort string) {
+	if mode == "auto" && effort == "" {
+		return
+	}
+	e := effort
+	if mode == "on" && e == "" {
+		e = "medium"
+	}
+	if e == "" {
+		return
+	}
+	if e == "max" {
+		if cfg.ExtraFields == nil {
+			cfg.ExtraFields = make(map[string]any)
+		}
+		cfg.ExtraFields["reasoning_effort"] = "max"
+		return
+	}
+	switch e {
+	case "low":
+		cfg.ReasoningEffort = einoopenai.ReasoningEffortLevelLow
+	case "medium":
+		cfg.ReasoningEffort = einoopenai.ReasoningEffortLevelMedium
+	case "high":
+		cfg.ReasoningEffort = einoopenai.ReasoningEffortLevelHigh
+	}
+}
+
+func applyOutputConfigEffort(cfg *einoopenai.ChatModelConfig, mode, effort string) {
+	if mode == "auto" && effort == "" {
+		return
+	}
+	e := effort
+	if mode == "on" && e == "" {
+		e = "high"
+	}
+	if e == "" {
+		return
+	}
+	if cfg.ExtraFields == nil {
+		cfg.ExtraFields = make(map[string]any)
+	}
+	cfg.ExtraFields["output_config"] = map[string]any{"effort": effortStringForAPI(e)}
+}
+
+func effortStringForAPI(e string) string {
+	// Gateways expect lowercase strings; "max" kept as max.
+	return strings.ToLower(strings.TrimSpace(e))
+}