Add files via upload

internal/multiagent/eino_adk_run_loop.go

@@ -79,6 +79,21 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 	mcpIDsMu := args.McpIDsMu
 	mcpIDs := args.McpIDs
 
+	// panic recovery：防止 Eino 框架内部 panic 导致整个 goroutine 崩溃、连接无法正常关闭。
+	defer func() {
+		if r := recover(); r != nil {
+			if logger != nil {
+				logger.Error("eino runner panic recovered", zap.Any("recover", r), zap.Stack("stack"))
+			}
+			if progress != nil {
+				progress("error", fmt.Sprintf("Internal error: %v / 内部错误: %v", r, r), map[string]interface{}{
+					"conversationId": conversationID,
+					"source":         "eino",
+				})
+			}
+		}
+	}()
+
 	var lastRunMsgs []adk.Message
 	var lastAssistant string
 	var lastPlanExecuteExecutor string
@@ -86,7 +101,8 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 
 	emptyHint := strings.TrimSpace(args.EmptyResponseMessage)
 	if emptyHint == "" {
-		emptyHint = "（Eino 会话已完成，但未捕获到助手文本输出。请查看过程详情或日志。）"
+		emptyHint = "(Eino session completed but no assistant text was captured. Check process details or logs.) " +
+			"（Eino 会话已完成，但未捕获到助手文本输出。请查看过程详情或日志。）"
 	}
 
 attemptLoop:
@@ -191,6 +207,20 @@ attemptLoop:
 		iter := runner.Run(ctx, msgs)
 
 		for {
+			// 检测 context 取消（用户关闭浏览器、请求超时等），flush pending 工具状态避免 UI 卡在 "执行中"。
+			select {
+			case <-ctx.Done():
+				flushAllPendingAsFailed(ctx.Err())
+				if progress != nil {
+					progress("error", "Request cancelled / 请求已取消", map[string]interface{}{
+						"conversationId": conversationID,
+						"source":         "eino",
+					})
+				}
+				return nil, ctx.Err()
+			default:
+			}
+
 			ev, ok := iter.Next()
 			if !ok {
 				lastRunMsgs = msgs
@@ -308,7 +338,10 @@ attemptLoop:
 							break
 						}
 						if logger != nil {
-							logger.Warn("eino stream recv", zap.Error(rerr))
+							logger.Warn("eino stream recv error, flushing incomplete stream",
+								zap.Error(rerr),
+								zap.String("agent", ev.AgentName),
+								zap.Int("toolFragments", len(toolStreamFragments)))
 						}
 						break
 					}
@@ -531,6 +564,11 @@ attemptLoop:
 	}
 	cleaned = dedupeRepeatedParagraphs(cleaned, 80)
 	cleaned = dedupeParagraphsByLineFingerprint(cleaned, 100)
+	// 防止超长响应导致 JSON 序列化慢或 OOM（多代理拼接大量工具输出时可能触发）。
+	const maxResponseRunes = 100000
+	if rs := []rune(cleaned); len(rs) > maxResponseRunes {
+		cleaned = string(rs[:maxResponseRunes]) + "\n\n... (response truncated / 响应已截断)"
+	}
 	out := &RunResult{
 		Response:         cleaned,
 		MCPExecutionIDs:  ids,

internal/multiagent/eino_orchestration.go

@@ -26,6 +26,13 @@ type PlanExecuteRootArgs struct {
 	// AppCfg / Logger 非空时为 Executor 挂载与 Deep/Supervisor 一致的 Eino summarization 中间件。
 	AppCfg *config.Config
 	Logger *zap.Logger
+	// ExecPreMiddlewares 是由 prependEinoMiddlewares 构建的前置中间件（patchtoolcalls, reduction, toolsearch, plantask），
+	// 与 Deep/Supervisor 主代理的 mainOrchestratorPre 一致。
+	ExecPreMiddlewares []adk.ChatModelAgentMiddleware
+	// SkillMiddleware 是 Eino 官方 skill 渐进式披露中间件（可选）。
+	SkillMiddleware adk.ChatModelAgentMiddleware
+	// FilesystemMiddleware 是 Eino filesystem 中间件，当 eino_skills.filesystem_tools 启用时提供本机文件读写与 Shell 能力（可选）。
+	FilesystemMiddleware adk.ChatModelAgentMiddleware
 }
 
 // NewPlanExecuteRoot 返回 plan → execute → replan 预置编排根节点（与 Deep / Supervisor 并列）。
@@ -40,20 +47,39 @@ func NewPlanExecuteRoot(ctx context.Context, a *PlanExecuteRootArgs) (adk.Resuma
 	if !ok {
 		return nil, fmt.Errorf("plan_execute: 主模型需实现 ToolCallingChatModel")
 	}
-	planner, err := planexecute.NewPlanner(ctx, &planexecute.PlannerConfig{
+	plannerCfg := &planexecute.PlannerConfig{
 		ToolCallingChatModel: tcm,
-	})
+	}
+	if fn := planExecutePlannerGenInput(a.OrchInstruction); fn != nil {
+		plannerCfg.GenInputFn = fn
+	}
+	planner, err := planexecute.NewPlanner(ctx, plannerCfg)
 	if err != nil {
 		return nil, fmt.Errorf("plan_execute planner: %w", err)
 	}
 	replanner, err := planexecute.NewReplanner(ctx, &planexecute.ReplannerConfig{
 		ChatModel:  tcm,
-		GenInputFn: planExecuteReplannerGenInput,
+		GenInputFn: planExecuteReplannerGenInput(a.OrchInstruction),
 	})
 	if err != nil {
 		return nil, fmt.Errorf("plan_execute replanner: %w", err)
 	}
+
+	// 组装 executor handler 栈，顺序与 Deep/Supervisor 主代理一致（outermost first）。
 	var execHandlers []adk.ChatModelAgentMiddleware
+	// 1. patchtoolcalls, reduction, toolsearch, plantask（来自 prependEinoMiddlewares）
+	if len(a.ExecPreMiddlewares) > 0 {
+		execHandlers = append(execHandlers, a.ExecPreMiddlewares...)
+	}
+	// 2. filesystem 中间件（可选）
+	if a.FilesystemMiddleware != nil {
+		execHandlers = append(execHandlers, a.FilesystemMiddleware)
+	}
+	// 3. skill 中间件（可选）
+	if a.SkillMiddleware != nil {
+		execHandlers = append(execHandlers, a.SkillMiddleware)
+	}
+	// 4. summarization（最后，与 Deep/Supervisor 一致）
 	if a.AppCfg != nil {
 		sumMw, sumErr := newEinoSummarizationMiddleware(ctx, a.ExecModel, a.AppCfg, a.Logger)
 		if sumErr != nil {
@@ -82,6 +108,21 @@ func NewPlanExecuteRoot(ctx context.Context, a *PlanExecuteRootArgs) (adk.Resuma
 	})
 }
 
+// planExecutePlannerGenInput 将 orchestrator instruction 作为 SystemMessage 注入 planner 输入。
+// 返回 nil 时 Eino 使用内置默认 planner prompt。
+func planExecutePlannerGenInput(orchInstruction string) planexecute.GenPlannerModelInputFn {
+	oi := strings.TrimSpace(orchInstruction)
+	if oi == "" {
+		return nil
+	}
+	return func(ctx context.Context, userInput []adk.Message) ([]adk.Message, error) {
+		msgs := make([]adk.Message, 0, 1+len(userInput))
+		msgs = append(msgs, schema.SystemMessage(oi))
+		msgs = append(msgs, userInput...)
+		return msgs, nil
+	}
+}
+
 func planExecuteExecutorGenInput(orchInstruction string) planexecute.GenModelInputFn {
 	oi := strings.TrimSpace(orchInstruction)
 	return func(ctx context.Context, in *planexecute.ExecutionContext) ([]adk.Message, error) {
@@ -123,19 +164,30 @@ func planExecuteFormatExecutedSteps(results []planexecute.ExecutedStep) string {
 	return sb.String()
 }
 
-// planExecuteReplannerGenInput 与 Eino 默认 Replanner 输入一致，但 executed_steps 经 cap 后再写入 prompt。
+// planExecuteReplannerGenInput 与 Eino 默认 Replanner 输入一致，但 executed_steps 经 cap 后再写入 prompt，
-func planExecuteReplannerGenInput(ctx context.Context, in *planexecute.ExecutionContext) ([]adk.Message, error) {
+// 且在 orchInstruction 非空时 prepend SystemMessage 使 replanner 也能接收全局指令。
-	planContent, err := in.Plan.MarshalJSON()
+func planExecuteReplannerGenInput(orchInstruction string) planexecute.GenModelInputFn {
-	if err != nil {
+	oi := strings.TrimSpace(orchInstruction)
-		return nil, err
+	return func(ctx context.Context, in *planexecute.ExecutionContext) ([]adk.Message, error) {
+		planContent, err := in.Plan.MarshalJSON()
+		if err != nil {
+			return nil, err
+		}
+		msgs, err := planexecute.ReplannerPrompt.Format(ctx, map[string]any{
+			"plan":           string(planContent),
+			"input":          planExecuteFormatInput(in.UserInput),
+			"executed_steps": planExecuteFormatExecutedSteps(in.ExecutedSteps),
+			"plan_tool":      planexecute.PlanToolInfo.Name,
+			"respond_tool":   planexecute.RespondToolInfo.Name,
+		})
+		if err != nil {
+			return nil, err
+		}
+		if oi != "" {
+			msgs = append([]adk.Message{schema.SystemMessage(oi)}, msgs...)
+		}
+		return msgs, nil
 	}
-	return planexecute.ReplannerPrompt.Format(ctx, map[string]any{
-		"plan":           string(planContent),
-		"input":          planExecuteFormatInput(in.UserInput),
-		"executed_steps": planExecuteFormatExecutedSteps(in.ExecutedSteps),
-		"plan_tool":      planexecute.PlanToolInfo.Name,
-		"respond_tool":   planexecute.RespondToolInfo.Name,
-	})
 }
 
 // planExecuteStreamsMainAssistant 将规划/执行/重规划各阶段助手流式输出映射到主对话区。

internal/multiagent/eino_single_runner.go

@@ -212,6 +212,7 @@ func RunEinoSingleChatModelAgent(
 		McpIDsMu:             &mcpIDsMu,
 		McpIDs:               &mcpIDs,
 		DA:                   chatAgent,
-		EmptyResponseMessage: "（Eino ADK 单代理会话已完成，但未捕获到助手文本输出。请查看过程详情或日志。）",
+		EmptyResponseMessage: "(Eino ADK single-agent session completed but no assistant text was captured. Check process details or logs.) " +
+			"（Eino ADK 单代理会话已完成，但未捕获到助手文本输出。请查看过程详情或日志。）",
 	}, baseMsgs)
 }

internal/multiagent/plan_execute_executor.go

@@ -23,13 +23,13 @@ func newPlanExecuteExecutor(ctx context.Context, cfg *planexecute.ExecutorConfig
 	genInput := func(ctx context.Context, instruction string, _ *adk.AgentInput) ([]adk.Message, error) {
 		plan, ok := adk.GetSessionValue(ctx, planexecute.PlanSessionKey)
 		if !ok {
-			panic("impossible: plan not found")
+			return nil, fmt.Errorf("plan_execute executor: session value %q missing (possible session corruption)", planexecute.PlanSessionKey)
 		}
 		plan_ := plan.(planexecute.Plan)
 
 		userInput, ok := adk.GetSessionValue(ctx, planexecute.UserInputSessionKey)
 		if !ok {
-			panic("impossible: user input not found")
+			return nil, fmt.Errorf("plan_execute executor: session value %q missing (possible session corruption)", planexecute.UserInputSessionKey)
 		}
 		userInput_ := userInput.([]adk.Message)
 

internal/multiagent/runner.go

@@ -335,6 +335,9 @@ func RunDeepAgent(
 		}
 		sb.WriteString("你是监督协调者：可将任务通过 transfer 工具委派给下列专家子代理（使用其在系统中的 Agent 名称）。专家列表：")
 		for _, sa := range subAgents {
+			if sa == nil {
+				continue
+			}
 			sb.WriteString("\n- ")
 			sb.WriteString(sa.Name(ctx))
 		}
@@ -349,14 +352,15 @@ func RunDeepAgent(
 		deepShell = einoLoc
 	}
 
-	deepHandlers := []adk.ChatModelAgentMiddleware{}
+	// noNestedTaskMiddleware 必须在最外层（最先拦截），防止 skill 或其他中间件内部触发 task 调用绕过检测。
+	deepHandlers := []adk.ChatModelAgentMiddleware{newNoNestedTaskMiddleware()}
 	if len(mainOrchestratorPre) > 0 {
 		deepHandlers = append(deepHandlers, mainOrchestratorPre...)
 	}
 	if einoSkillMW != nil {
 		deepHandlers = append(deepHandlers, einoSkillMW)
 	}
-	deepHandlers = append(deepHandlers, newNoNestedTaskMiddleware(), mainSumMw)
+	deepHandlers = append(deepHandlers, mainSumMw)
 
 	supHandlers := []adk.ChatModelAgentMiddleware{}
 	if len(mainOrchestratorPre) > 0 {
@@ -387,6 +391,14 @@ func RunDeepAgent(
 		if perr != nil {
 			return nil, fmt.Errorf("plan_execute 执行器模型: %w", perr)
 		}
+		// 构建 filesystem 中间件（与 Deep sub-agent 一致）
+		var peFsMw adk.ChatModelAgentMiddleware
+		if einoSkillMW != nil && einoFSTools && einoLoc != nil {
+			peFsMw, err = subAgentFilesystemMiddleware(ctx, einoLoc)
+			if err != nil {
+				return nil, fmt.Errorf("plan_execute filesystem 中间件: %w", err)
+			}
+		}
 		peRoot, perr := NewPlanExecuteRoot(ctx, &PlanExecuteRootArgs{
 			MainToolCallingModel: mainModel,
 			ExecModel:            execModel,
@@ -396,6 +408,9 @@ func RunDeepAgent(
 			LoopMaxIter:          ma.PlanExecuteLoopMaxIterations,
 			AppCfg:               appCfg,
 			Logger:               logger,
+			ExecPreMiddlewares:   mainOrchestratorPre,
+			SkillMiddleware:      einoSkillMW,
+			FilesystemMiddleware: peFsMw,
 		})
 		if perr != nil {
 			return nil, perr
@@ -493,7 +508,8 @@ func RunDeepAgent(
 		McpIDsMu:             &mcpIDsMu,
 		McpIDs:               &mcpIDs,
 		DA:                   da,
-		EmptyResponseMessage: "（Eino 多代理编排已完成，但未捕获到助手文本输出。请查看过程详情或日志。）",
+		EmptyResponseMessage: "(Eino multi-agent orchestration completed but no assistant text was captured. Check process details or logs.) " +
+			"（Eino 多代理编排已完成，但未捕获到助手文本输出。请查看过程详情或日志。）",
 	}, baseMsgs)
 }
 

internal/multiagent/tool_error_middleware.go

@@ -3,6 +3,7 @@ package multiagent
 import (
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"strings"
 
@@ -44,6 +45,17 @@ func isSoftRecoverableToolError(err error) bool {
 	if err == nil {
 		return false
 	}
+
+	// 用户取消 — 不应重试，让 hard error 传播以终止编排。
+	if errors.Is(err, context.Canceled) {
+		return false
+	}
+
+	// 工具执行超时 — 转为 soft error 让 LLM 知晓并选择替代方案，而非全局重试。
+	if errors.Is(err, context.DeadlineExceeded) {
+		return true
+	}
+
 	s := strings.ToLower(err.Error())
 
 	// JSON unmarshal/parse failures — the model generated truncated or malformed arguments.