Add files via upload

2026-08-15 15:40:38 +02:00 · 2026-08-15 01:36:02 +08:00
parent e691c5d9c6
commit e591fe4a9b
2 changed files with 43 additions and 66 deletions
@@ -54,9 +54,6 @@ type EnsureLocalConfigResult struct {

 const (
 	DefaultMaxCompletionTokens                        = 16384
-	DefaultMaxToolArgumentsBytes                      = 65536
-	DefaultMaxShellCommandBytes                       = 65536
-	DefaultModelOutputRepairMaxAttempts               = 1
 	DefaultSummarizationUserIntentLedgerMaxRunes      = 96000
 	DefaultSummarizationUserIntentLedgerEntryMaxRunes = 16000
 	DefaultLatestUserMessageMaxRunes                  = 48000
@@ -256,12 +253,6 @@ func (c MultiAgentEinoCallbacksConfig) EinoCallbacksMaxOutputSummaryRunes() int

 // MultiAgentEinoMiddlewareConfig optional Eino ADK middleware and Deep / supervisor tuning.
 type MultiAgentEinoMiddlewareConfig struct {
-	// MaxToolArgumentsBytes hard-rejects oversized model-generated tool arguments before execution.
-	MaxToolArgumentsBytes int `yaml:"max_tool_arguments_bytes,omitempty" json:"max_tool_arguments_bytes,omitempty"`
-	// MaxShellCommandBytes applies a stricter limit to exec/execute command strings.
-	MaxShellCommandBytes int `yaml:"max_shell_command_bytes,omitempty" json:"max_shell_command_bytes,omitempty"`
-	// ModelOutputRepairMaxAttempts limits consecutive model-output repair attempts.
-	ModelOutputRepairMaxAttempts int `yaml:"model_output_repair_max_attempts,omitempty" json:"model_output_repair_max_attempts,omitempty"`
 	// PatchToolCalls inserts placeholder tool results for dangling assistant tool_calls (nil = enabled).
 	PatchToolCalls *bool `yaml:"patch_tool_calls,omitempty" json:"patch_tool_calls,omitempty"`
 	// ToolSearch enables dynamictool/toolsearch: hide tail tools until model calls tool_search (reduces prompt tools).
@@ -297,7 +288,7 @@ type MultiAgentEinoMiddlewareConfig struct {
 	LatestUserMessageHeadRunes int `yaml:"latest_user_message_head_runes,omitempty" json:"latest_user_message_head_runes,omitempty"`
 	// LatestUserMessageTailRunes keeps the tail preview for an oversized current user turn.
 	LatestUserMessageTailRunes int `yaml:"latest_user_message_tail_runes,omitempty" json:"latest_user_message_tail_runes,omitempty"`
-	// SummarizationRetryMaxAttempts 已废弃：summarization 与 run loop 共用 run_retry_max_attempts 及 isEinoTransientRunError。
+	// SummarizationRetryMaxAttempts 已废弃：summarization 与 Eino 原生 ModelRetry 共用 model_retry_max_retries 及 isEinoTransientRunError。
 	SummarizationRetryMaxAttempts int `yaml:"summarization_retry_max_attempts,omitempty" json:"summarization_retry_max_attempts,omitempty"`
 	// PlanExecuteUserInputBudgetRatio caps planner/replanner/executor userInput prompt budget ratio (default 0.35).
 	PlanExecuteUserInputBudgetRatio float64 `yaml:"plan_execute_user_input_budget_ratio,omitempty" json:"plan_execute_user_input_budget_ratio,omitempty"`
@@ -311,11 +302,19 @@ type MultiAgentEinoMiddlewareConfig struct {
 	CheckpointDir string `yaml:"checkpoint_dir,omitempty" json:"checkpoint_dir,omitempty"`
 	// DeepOutputKey passed to deep.Config OutputKey (session final text); empty = off.
 	DeepOutputKey string `yaml:"deep_output_key,omitempty" json:"deep_output_key,omitempty"`
-	// DeepModelRetryMaxRetries 已废弃：临时错误统一由 run loop 内 isEinoTransientRunError + run_retry_max_attempts 处理。
+	// DeepModelRetryMaxRetries 已废弃：请用 model_retry_max_retries；保留字段仅为兼容旧配置。
 	DeepModelRetryMaxRetries int `yaml:"deep_model_retry_max_retries,omitempty" json:"deep_model_retry_max_retries,omitempty"`
-	// RunRetryMaxAttempts > 0：408/409/425/429/5xx/网络抖动时可退避重试次数（run loop 与 summarization 共用）；0=默认 4。
+	// ModelRetryMaxRetries configures Eino ADK native ChatModel retry attempts; 0=default 4.
+	ModelRetryMaxRetries int `yaml:"model_retry_max_retries,omitempty" json:"model_retry_max_retries,omitempty"`
+	// ModelRetryMaxBackoffSec caps native model retry backoff seconds; 0=default 30.
+	ModelRetryMaxBackoffSec int `yaml:"model_retry_max_backoff_sec,omitempty" json:"model_retry_max_backoff_sec,omitempty"`
+	// ModelFailoverChannels lists ai.channels IDs to try after native model retry is exhausted.
+	ModelFailoverChannels []string `yaml:"model_failover_channels,omitempty" json:"model_failover_channels,omitempty"`
+	// ModelFailoverMaxRetries caps distinct failover channel attempts; 0=all configured failover channels.
+	ModelFailoverMaxRetries int `yaml:"model_failover_max_retries,omitempty" json:"model_failover_max_retries,omitempty"`
+	// RunRetryMaxAttempts 已废弃：模型临时错误由 Eino 原生 ModelRetry 处理；仅保留给非模型层 run loop 兜底与 summarization 旧字段。
 	RunRetryMaxAttempts int `yaml:"run_retry_max_attempts,omitempty" json:"run_retry_max_attempts,omitempty"`
-	// RunRetryMaxBackoffSec 单次退避上限秒数；0=默认 30。
+	// RunRetryMaxBackoffSec 已废弃：请用 model_retry_max_backoff_sec；仅保留给非模型层 run loop 兜底与 summarization 旧字段。
 	RunRetryMaxBackoffSec int `yaml:"run_retry_max_backoff_sec,omitempty" json:"run_retry_max_backoff_sec,omitempty"`
 	// EmptyResponseContinueMaxAttempts Run 成功但未捕获助手正文时 Handler 层退避续跑次数；0=默认 5。
 	EmptyResponseContinueMaxAttempts int `yaml:"empty_response_continue_max_attempts,omitempty" json:"empty_response_continue_max_attempts,omitempty"`
@@ -323,27 +322,6 @@ type MultiAgentEinoMiddlewareConfig struct {
 	TaskToolDescriptionPrefix string `yaml:"task_tool_description_prefix,omitempty" json:"task_tool_description_prefix,omitempty"`
 }

-func (c MultiAgentEinoMiddlewareConfig) MaxToolArgumentsBytesEffective() int {
-	if c.MaxToolArgumentsBytes > 0 {
-		return c.MaxToolArgumentsBytes
-	}
-	return DefaultMaxToolArgumentsBytes
-}
-
-func (c MultiAgentEinoMiddlewareConfig) MaxShellCommandBytesEffective() int {
-	if c.MaxShellCommandBytes > 0 {
-		return c.MaxShellCommandBytes
-	}
-	return DefaultMaxShellCommandBytes
-}
-
-func (c MultiAgentEinoMiddlewareConfig) ModelOutputRepairMaxAttemptsEffective() int {
-	if c.ModelOutputRepairMaxAttempts > 0 {
-		return c.ModelOutputRepairMaxAttempts
-	}
-	return DefaultModelOutputRepairMaxAttempts
-}
-
 func (c MultiAgentEinoMiddlewareConfig) SummarizationTriggerRatioEffective() float64 {
 	v := c.SummarizationTriggerRatio
 	if v <= 0 {
@@ -514,6 +492,10 @@ type MultiAgentPublic struct {
 	LatestUserMessageMaxRunes                  int      `json:"latest_user_message_max_runes"`
 	LatestUserMessageHeadRunes                 int      `json:"latest_user_message_head_runes"`
 	LatestUserMessageTailRunes                 int      `json:"latest_user_message_tail_runes"`
+	ModelRetryMaxRetries                       int      `json:"model_retry_max_retries"`
+	ModelRetryMaxBackoffSec                    int      `json:"model_retry_max_backoff_sec"`
+	ModelFailoverChannels                      []string `json:"model_failover_channels,omitempty"`
+	ModelFailoverMaxRetries                    int      `json:"model_failover_max_retries"`
 	ToolSearchAlwaysVisibleTools               []string `json:"tool_search_always_visible_tools,omitempty"`
 	ToolSearchAlwaysVisibleEffectiveTools      []string `json:"tool_search_always_visible_effective_tools,omitempty"`
 }
@@ -555,15 +537,19 @@ func NormalizeMultiAgentOrchestration(s string) string {

 // MultiAgentAPIUpdate 设置页/API 仅更新多代理标量字段；写入 YAML 时不覆盖 sub_agents 等块。
 type MultiAgentAPIUpdate struct {
-	Enabled                                    bool   `json:"enabled"`
-	RobotDefaultAgentMode                      string `json:"robot_default_agent_mode,omitempty"`
-	BatchUseMultiAgent                         bool   `json:"batch_use_multi_agent"`
-	PlanExecuteLoopMaxIterations               *int   `json:"plan_execute_loop_max_iterations,omitempty"`
-	SummarizationUserIntentLedgerMaxRunes      *int   `json:"summarization_user_intent_ledger_max_runes,omitempty"`
-	SummarizationUserIntentLedgerEntryMaxRunes *int   `json:"summarization_user_intent_ledger_entry_max_runes,omitempty"`
-	LatestUserMessageMaxRunes                  *int   `json:"latest_user_message_max_runes,omitempty"`
-	LatestUserMessageHeadRunes                 *int   `json:"latest_user_message_head_runes,omitempty"`
-	LatestUserMessageTailRunes                 *int   `json:"latest_user_message_tail_runes,omitempty"`
+	Enabled                                    bool      `json:"enabled"`
+	RobotDefaultAgentMode                      string    `json:"robot_default_agent_mode,omitempty"`
+	BatchUseMultiAgent                         bool      `json:"batch_use_multi_agent"`
+	PlanExecuteLoopMaxIterations               *int      `json:"plan_execute_loop_max_iterations,omitempty"`
+	SummarizationUserIntentLedgerMaxRunes      *int      `json:"summarization_user_intent_ledger_max_runes,omitempty"`
+	SummarizationUserIntentLedgerEntryMaxRunes *int      `json:"summarization_user_intent_ledger_entry_max_runes,omitempty"`
+	LatestUserMessageMaxRunes                  *int      `json:"latest_user_message_max_runes,omitempty"`
+	LatestUserMessageHeadRunes                 *int      `json:"latest_user_message_head_runes,omitempty"`
+	LatestUserMessageTailRunes                 *int      `json:"latest_user_message_tail_runes,omitempty"`
+	ModelRetryMaxRetries                       *int      `json:"model_retry_max_retries,omitempty"`
+	ModelRetryMaxBackoffSec                    *int      `json:"model_retry_max_backoff_sec,omitempty"`
+	ModelFailoverChannels                      *[]string `json:"model_failover_channels,omitempty"`
+	ModelFailoverMaxRetries                    *int      `json:"model_failover_max_retries,omitempty"`
 	// 指针区分「JSON 未传该字段」与「传空数组要清空」；省略时不应覆盖 YAML 中的常驻工具白名单。
 	ToolSearchAlwaysVisibleTools *[]string `json:"tool_search_always_visible_tools,omitempty"`
 }
@@ -972,6 +958,16 @@ func (c OpenAIConfig) MaxCompletionTokensEffective() int {
 	return DefaultMaxCompletionTokens
 }

+// IsDeepSeekEndpointOrModel reports whether the channel targets DeepSeek's
+// official-compatible API or a DeepSeek model family. This is separate from the
+// reasoning profile: profile controls field mapping, while DeepSeek has provider
+// constraints such as default thinking mode and no tool_choice in thinking mode.
+func (c OpenAIConfig) IsDeepSeekEndpointOrModel() bool {
+	baseURL := strings.ToLower(strings.TrimSpace(c.BaseURL))
+	model := strings.ToLower(strings.TrimSpace(c.Model))
+	return strings.Contains(baseURL, "deepseek") || strings.Contains(model, "deepseek")
+}
+
 // OpenAIReasoningConfig 全局默认与网关 profile（对话页可通过 ChatRequest.reasoning 覆盖，受 AllowClientReasoning 约束）。
 type OpenAIReasoningConfig struct {
 	// Mode: auto（默认）| on | off | default（与 auto 相同）。
@@ -1377,7 +1373,7 @@ func Load(path string) (*Config, error) {
 		cfg.Audit.MaxDetailBytes = 8192
 	}
 	cfg.ApplyDefaultAIChannel()
-	if err := validateModelOutputLimits(cfg.OpenAI, cfg.MultiAgent.EinoMiddleware); err != nil {
+	if err := validateOpenAIOutputLimits(cfg.OpenAI); err != nil {
 		return nil, err
 	}
 	// 如果配置了工具目录，从目录加载工具配置
@@ -1442,22 +1438,10 @@ func Load(path string) (*Config, error) {
 	return &cfg, nil
 }

-func validateModelOutputLimits(openAI OpenAIConfig, mw MultiAgentEinoMiddlewareConfig) error {
+func validateOpenAIOutputLimits(openAI OpenAIConfig) error {
 	if openAI.MaxCompletionTokens < 0 {
 		return fmt.Errorf("openai.max_completion_tokens 必须为正数")
 	}
-	if mw.MaxToolArgumentsBytes < 0 {
-		return fmt.Errorf("multi_agent.eino_middleware.max_tool_arguments_bytes 必须为正数")
-	}
-	if mw.MaxShellCommandBytes < 0 {
-		return fmt.Errorf("multi_agent.eino_middleware.max_shell_command_bytes 必须为正数")
-	}
-	if mw.ModelOutputRepairMaxAttempts < 0 {
-		return fmt.Errorf("multi_agent.eino_middleware.model_output_repair_max_attempts 必须为正数")
-	}
-	if mw.MaxShellCommandBytesEffective() > mw.MaxToolArgumentsBytesEffective() {
-		return fmt.Errorf("multi_agent.eino_middleware.max_shell_command_bytes 不能大于 max_tool_arguments_bytes")
-	}
 	return nil
 }

@@ -169,18 +169,11 @@ func TestSummarizationOutputReserveTokensEffective(t *testing.T) {
 	}
 }

-func TestModelOutputLimitDefaultsAndValidation(t *testing.T) {
+func TestOpenAIOutputLimitValidation(t *testing.T) {
 	if got := (OpenAIConfig{}).MaxCompletionTokensEffective(); got != DefaultMaxCompletionTokens {
 		t.Fatalf("max completion default=%d", got)
 	}
-	mw := MultiAgentEinoMiddlewareConfig{}
-	if mw.MaxToolArgumentsBytesEffective() != 65536 || mw.MaxShellCommandBytesEffective() != 65536 || mw.ModelOutputRepairMaxAttemptsEffective() != 1 {
-		t.Fatalf("unexpected guard defaults: %+v", mw)
-	}
-	if err := validateModelOutputLimits(OpenAIConfig{}, MultiAgentEinoMiddlewareConfig{MaxShellCommandBytes: 100, MaxToolArgumentsBytes: 99}); err == nil {
-		t.Fatal("shell limit greater than generic limit must fail")
-	}
-	if err := validateModelOutputLimits(OpenAIConfig{MaxCompletionTokens: -1}, MultiAgentEinoMiddlewareConfig{}); err == nil {
+	if err := validateOpenAIOutputLimits(OpenAIConfig{MaxCompletionTokens: -1}); err == nil {
 		t.Fatal("negative completion limit must fail")
 	}
 }