Add files via upload

internal/multiagent/runner.go

@@ -233,6 +233,9 @@ func RunDeepAgent(
 				ToolsNodeConfig: compose.ToolsNodeConfig{
 					Tools:               subTools,
 					UnknownToolsHandler: einomcp.UnknownToolReminderHandler(),
+					ToolCallMiddlewares: []compose.ToolMiddleware{
+						{Invokable: softRecoveryToolCallMiddleware()},
+					},
 				},
 				EmitInternalEvents: true,
 			},
@@ -288,6 +291,9 @@ func RunDeepAgent(
 			ToolsNodeConfig: compose.ToolsNodeConfig{
 				Tools:               mainTools,
 				UnknownToolsHandler: einomcp.UnknownToolReminderHandler(),
+				ToolCallMiddlewares: []compose.ToolMiddleware{
+					{Invokable: softRecoveryToolCallMiddleware()},
+				},
 			},
 			EmitInternalEvents: true,
 		},

internal/multiagent/tool_args_json_retry.go

@@ -10,7 +10,7 @@ import (
 // maxToolCallRecoveryAttempts 含首次运行：首次 + 自动重试次数。
 // 例如为 3 表示最多共 3 次完整 DeepAgent 运行（2 次失败后各追加一条纠错提示）。
 // 该常量同时用于 JSON 参数错误和工具执行错误（如子代理名称不存在）的恢复重试。
-const maxToolCallRecoveryAttempts = 3
+const maxToolCallRecoveryAttempts = 5
 
 // toolCallArgumentsJSONRetryHint 追加在用户消息后，提示模型输出合法 JSON 工具参数（部分云厂商会在流式阶段校验 arguments）。
 func toolCallArgumentsJSONRetryHint() *schema.Message {

internal/multiagent/tool_error_middleware.go

@@ -0,0 +1,131 @@
+package multiagent
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"strings"
+
+	"github.com/cloudwego/eino/compose"
+)
+
+// softRecoveryToolCallMiddleware returns an InvokableToolMiddleware that catches
+// specific recoverable errors from tool execution (JSON parse errors, tool-not-found,
+// etc.) and converts them into soft errors: nil error + descriptive error content
+// returned to the LLM. This allows the model to self-correct within the same
+// iteration rather than crashing the entire graph and requiring a full replay.
+//
+// Without this middleware, a JSON parse failure in any tool's InvokableRun propagates
+// as a hard error through the Eino ToolsNode → [NodeRunError] → ev.Err, which
+// either triggers the full-replay retry loop (expensive) or terminates the run
+// entirely once retries are exhausted. With it, the LLM simply sees an error message
+// in the tool result and can adjust its next tool call accordingly.
+func softRecoveryToolCallMiddleware() compose.InvokableToolMiddleware {
+	return func(next compose.InvokableToolEndpoint) compose.InvokableToolEndpoint {
+		return func(ctx context.Context, input *compose.ToolInput) (*compose.ToolOutput, error) {
+			output, err := next(ctx, input)
+			if err == nil {
+				return output, nil
+			}
+			if !isSoftRecoverableToolError(err) {
+				return output, err
+			}
+			// Convert the hard error into a soft error: the LLM will see this
+			// message as the tool's output and can self-correct.
+			msg := buildSoftRecoveryMessage(input.Name, input.Arguments, err)
+			return &compose.ToolOutput{Result: msg}, nil
+		}
+	}
+}
+
+// isSoftRecoverableToolError determines whether a tool execution error should be
+// silently converted to a tool-result message rather than crashing the graph.
+func isSoftRecoverableToolError(err error) bool {
+	if err == nil {
+		return false
+	}
+	s := strings.ToLower(err.Error())
+
+	// JSON unmarshal/parse failures — the model generated truncated or malformed arguments.
+	if isJSONRelatedError(s) {
+		return true
+	}
+
+	// Sub-agent type not found (from deep/task_tool.go)
+	if strings.Contains(s, "subagent type") && strings.Contains(s, "not found") {
+		return true
+	}
+
+	// Tool not found in ToolsNode indexes
+	if strings.Contains(s, "tool") && strings.Contains(s, "not found") {
+		return true
+	}
+
+	return false
+}
+
+// isJSONRelatedError checks whether an error string indicates a JSON parsing problem.
+func isJSONRelatedError(lower string) bool {
+	if !strings.Contains(lower, "json") {
+		return false
+	}
+	jsonIndicators := []string{
+		"unexpected end of json",
+		"unmarshal",
+		"invalid character",
+		"cannot unmarshal",
+		"invalid tool arguments",
+		"failed to unmarshal",
+		"must be in json format",
+		"unexpected eof",
+	}
+	for _, ind := range jsonIndicators {
+		if strings.Contains(lower, ind) {
+			return true
+		}
+	}
+	return false
+}
+
+// buildSoftRecoveryMessage creates a bilingual error message that the LLM can act on.
+func buildSoftRecoveryMessage(toolName, arguments string, err error) string {
+	// Truncate arguments preview to avoid flooding the context.
+	argPreview := arguments
+	if len(argPreview) > 300 {
+		argPreview = argPreview[:300] + "... (truncated)"
+	}
+
+	// Try to determine if it's specifically a JSON parse error for a friendlier message.
+	errStr := err.Error()
+	var jsonErr *json.SyntaxError
+	isJSONErr := strings.Contains(strings.ToLower(errStr), "json") ||
+		strings.Contains(strings.ToLower(errStr), "unmarshal")
+	_ = jsonErr // suppress unused
+
+	if isJSONErr {
+		return fmt.Sprintf(
+			"[Tool Error] The arguments for tool '%s' are not valid JSON and could not be parsed.\n"+
+				"Error: %s\n"+
+				"Arguments received: %s\n\n"+
+				"Please fix the JSON (ensure double-quoted keys, matched braces/brackets, no trailing commas, "+
+				"no truncation) and call the tool again.\n\n"+
+				"[工具错误] 工具 '%s' 的参数不是合法 JSON，无法解析。\n"+
+				"错误：%s\n"+
+				"收到的参数：%s\n\n"+
+				"请修正 JSON（确保双引号键名、括号配对、无尾部逗号、无截断），然后重新调用工具。",
+			toolName, errStr, argPreview,
+			toolName, errStr, argPreview,
+		)
+	}
+
+	return fmt.Sprintf(
+		"[Tool Error] Tool '%s' execution failed: %s\n"+
+			"Arguments: %s\n\n"+
+			"Please review the available tools and their expected arguments, then retry.\n\n"+
+			"[工具错误] 工具 '%s' 执行失败：%s\n"+
+			"参数：%s\n\n"+
+			"请检查可用工具及其参数要求，然后重试。",
+		toolName, errStr, argPreview,
+		toolName, errStr, argPreview,
+	)
+}

internal/multiagent/tool_error_middleware_test.go

@@ -0,0 +1,166 @@
+package multiagent
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"testing"
+
+	"github.com/cloudwego/eino/compose"
+)
+
+func TestIsSoftRecoverableToolError(t *testing.T) {
+	tests := []struct {
+		name     string
+		err      error
+		expected bool
+	}{
+		{
+			name:     "nil error",
+			err:      nil,
+			expected: false,
+		},
+		{
+			name:     "unexpected end of JSON input",
+			err:      errors.New("unexpected end of JSON input"),
+			expected: true,
+		},
+		{
+			name:     "failed to unmarshal task tool input json",
+			err:      errors.New("failed to unmarshal task tool input json: unexpected end of JSON input"),
+			expected: true,
+		},
+		{
+			name:     "invalid tool arguments JSON",
+			err:      errors.New("invalid tool arguments JSON: unexpected end of JSON input"),
+			expected: true,
+		},
+		{
+			name:     "json invalid character",
+			err:      errors.New(`invalid character '}' looking for beginning of value in JSON`),
+			expected: true,
+		},
+		{
+			name:     "subagent type not found",
+			err:      errors.New("subagent type recon_agent not found"),
+			expected: true,
+		},
+		{
+			name:     "tool not found",
+			err:      errors.New("tool nmap_scan not found in toolsNode indexes"),
+			expected: true,
+		},
+		{
+			name:     "unrelated network error",
+			err:      errors.New("connection refused"),
+			expected: false,
+		},
+		{
+			name:     "context cancelled",
+			err:      context.Canceled,
+			expected: false,
+		},
+		{
+			name: "real json unmarshal error",
+			err: func() error {
+				var v map[string]interface{}
+				return json.Unmarshal([]byte(`{"key": `), &v)
+			}(),
+			expected: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := isSoftRecoverableToolError(tt.err)
+			if got != tt.expected {
+				t.Errorf("isSoftRecoverableToolError(%v) = %v, want %v", tt.err, got, tt.expected)
+			}
+		})
+	}
+}
+
+func TestSoftRecoveryToolCallMiddleware_PassesThrough(t *testing.T) {
+	mw := softRecoveryToolCallMiddleware()
+	called := false
+	next := func(ctx context.Context, input *compose.ToolInput) (*compose.ToolOutput, error) {
+		called = true
+		return &compose.ToolOutput{Result: "success"}, nil
+	}
+	wrapped := mw(next)
+	out, err := wrapped(context.Background(), &compose.ToolInput{
+		Name:      "test_tool",
+		Arguments: `{"key": "value"}`,
+	})
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !called {
+		t.Fatal("next endpoint was not called")
+	}
+	if out.Result != "success" {
+		t.Fatalf("expected 'success', got %q", out.Result)
+	}
+}
+
+func TestSoftRecoveryToolCallMiddleware_ConvertsJSONError(t *testing.T) {
+	mw := softRecoveryToolCallMiddleware()
+	next := func(ctx context.Context, input *compose.ToolInput) (*compose.ToolOutput, error) {
+		return nil, errors.New("failed to unmarshal task tool input json: unexpected end of JSON input")
+	}
+	wrapped := mw(next)
+	out, err := wrapped(context.Background(), &compose.ToolInput{
+		Name:      "task",
+		Arguments: `{"subagent_type": "recon`,
+	})
+	if err != nil {
+		t.Fatalf("expected nil error (soft recovery), got: %v", err)
+	}
+	if out == nil || out.Result == "" {
+		t.Fatal("expected non-empty recovery message")
+	}
+	if !containsAll(out.Result, "[Tool Error]", "task", "JSON") {
+		t.Fatalf("recovery message missing expected content: %s", out.Result)
+	}
+}
+
+func TestSoftRecoveryToolCallMiddleware_PropagatesNonRecoverable(t *testing.T) {
+	mw := softRecoveryToolCallMiddleware()
+	origErr := errors.New("connection timeout to remote server")
+	next := func(ctx context.Context, input *compose.ToolInput) (*compose.ToolOutput, error) {
+		return nil, origErr
+	}
+	wrapped := mw(next)
+	_, err := wrapped(context.Background(), &compose.ToolInput{
+		Name:      "test_tool",
+		Arguments: `{}`,
+	})
+	if err == nil {
+		t.Fatal("expected error to propagate for non-recoverable errors")
+	}
+	if err != origErr {
+		t.Fatalf("expected original error, got: %v", err)
+	}
+}
+
+func containsAll(s string, subs ...string) bool {
+	for _, sub := range subs {
+		if !contains(s, sub) {
+			return false
+		}
+	}
+	return true
+}
+
+func contains(s, sub string) bool {
+	return len(s) >= len(sub) && searchString(s, sub)
+}
+
+func searchString(s, sub string) bool {
+	for i := 0; i <= len(s)-len(sub); i++ {
+		if s[i:i+len(sub)] == sub {
+			return true
+		}
+	}
+	return false
+}