Enhance Assistant methods with options parameter for improved context management

- Updated the Stream, BuildContent, and LLM execution methods to accept an Options parameter, allowing for more flexible context handling.
- Removed debug print statements to clean up the code and improve readability.
- Enhanced locale handling in the loadMap function to automatically inject assistant name and description into all locales, ensuring better localization support.
- Introduced output skipping functionality in context options to manage internal A2A calls more effectively.
- Improved output writer resolution logic to prioritize context settings, enhancing output management during agent calls.

agent/assistant/agent.go

@@ -51,13 +51,6 @@ func (ast *Assistant) Stream(ctx *context.Context, inputMessages []context.Messa
 	_, _, done := context.EnterStack(ctx, ast.ID, opts)
 	defer done()
 
-	fmt.Println("--- Stack debug ---")
-	if ctx.Stack != nil {
-		fmt.Println(ctx.Stack.IsRoot())
-		utils.Dump(ctx.Stack)
-	}
-	fmt.Println("------ end stack debug ------")
-
 	// Determine stream handler
 	streamHandler := ast.getStreamHandler(ctx, opts)
 
@@ -123,7 +116,7 @@ func (ast *Assistant) Stream(ctx *context.Context, inputMessages []context.Messa
 		}
 
 		// Build content - convert extended types (file, data) to standard LLM types (text, image_url, input_audio)
-		completionMessages, err = ast.BuildContent(ctx, completionMessages, completionOptions)
+		completionMessages, err = ast.BuildContent(ctx, completionMessages, completionOptions, opts)
 		if err != nil {
 			ast.traceAgentFail(agentNode, err)
 			ast.sendStreamEndOnError(ctx, streamHandler, streamStartTime, err)
@@ -131,7 +124,7 @@ func (ast *Assistant) Stream(ctx *context.Context, inputMessages []context.Messa
 		}
 
 		// Execute the LLM streaming call
-		completionResponse, err = ast.executeLLMStream(ctx, completionMessages, completionOptions, agentNode, streamHandler)
+		completionResponse, err = ast.executeLLMStream(ctx, completionMessages, completionOptions, agentNode, streamHandler, opts)
 		if err != nil {
 			ast.traceAgentFail(agentNode, err)
 			// Send error stream_end for root stack
@@ -210,7 +203,7 @@ func (ast *Assistant) Stream(ctx *context.Context, inputMessages []context.Messa
 
 			// Retry LLM call (streaming to keep user informed)
 			log.Trace("[AGENT] Retrying LLM for tool call correction (attempt %d/%d)", attempt+1, maxToolRetries-1)
-			currentResponse, err = ast.executeLLMForToolRetry(ctx, retryMessages, completionOptions, agentNode, streamHandler)
+			currentResponse, err = ast.executeLLMForToolRetry(ctx, retryMessages, completionOptions, agentNode, streamHandler, opts)
 			if err != nil {
 				log.Error("[AGENT] LLM retry failed: %v", err)
 				ast.traceAgentFail(agentNode, err)
@@ -479,6 +472,10 @@ func (ast *Assistant) initializeCapabilities(ctx *context.Context, opts *context
 		return err
 	}
 
+	fmt.Println("--- initializeCapabilities debug ---")
+	utils.Dump(capabilities)
+	fmt.Println("--- end initializeCapabilities debug ---")
+
 	// Set capabilities in context for output adapters to use
 	if capabilities != nil {
 		ctx.Capabilities = capabilities

agent/assistant/build_content.go

@@ -11,9 +11,9 @@ import (
 // (file, data) to standard LLM-compatible types (text, image_url, input_audio)
 //
 // This should be called after BuildRequest and before executing LLM call
-func (ast *Assistant) BuildContent(ctx *context.Context, messages []context.Message, options *context.CompletionOptions) ([]context.Message, error) {
+func (ast *Assistant) BuildContent(ctx *context.Context, messages []context.Message, options *context.CompletionOptions, opts *context.Options) ([]context.Message, error) {
 	// Get connector and capabilities
-	_, capabilities, err := ast.GetConnector(ctx)
+	_, capabilities, err := ast.GetConnector(ctx, opts)
 	if err != nil {
 		return nil, fmt.Errorf("failed to get connector: %w", err)
 	}

agent/assistant/llm.go

@@ -18,6 +18,7 @@ func (ast *Assistant) executeLLMStream(
 	completionOptions *context.CompletionOptions,
 	agentNode types.Node,
 	streamHandler message.StreamFunc,
+	opts *context.Options,
 ) (*context.CompletionResponse, error) {
 
 	// === Debug LLM Stream Start ===
@@ -27,7 +28,7 @@ func (ast *Assistant) executeLLMStream(
 	// === End Debug ===
 
 	// Get connector object (capabilities were already set above, before stream_start)
-	conn, capabilities, err := ast.GetConnector(ctx)
+	conn, capabilities, err := ast.GetConnector(ctx, opts)
 	if err != nil {
 		ast.traceAgentFail(agentNode, err)
 		return nil, err
@@ -92,10 +93,11 @@ func (ast *Assistant) executeLLMForToolRetry(
 	completionOptions *context.CompletionOptions,
 	agentNode types.Node,
 	streamHandler message.StreamFunc,
+	opts *context.Options,
 ) (*context.CompletionResponse, error) {
 
 	// Get connector object
-	conn, capabilities, err := ast.GetConnector(ctx)
+	conn, capabilities, err := ast.GetConnector(ctx, opts)
 	if err != nil {
 		ast.traceAgentFail(agentNode, err)
 		return nil, err

agent/assistant/load.go

@@ -531,7 +531,38 @@ func loadMap(data map[string]interface{}) (*Assistant, error) {
 	// locales
 	if locales, ok := data["locales"].(i18n.Map); ok {
 		assistant.Locales = locales
-		i18n.Locales[id] = locales.FlattenWithGlobal()
+		flattened := locales.FlattenWithGlobal()
+
+		// Auto-inject assistant name and description into all locales
+		// so that {{name}} and {{description}} templates can be resolved
+		for locale, i18nObj := range flattened {
+			if i18nObj.Messages == nil {
+				i18nObj.Messages = make(map[string]any)
+			}
+			// Add name and description if not already present
+			if _, exists := i18nObj.Messages["name"]; !exists && assistant.Name != "" {
+				i18nObj.Messages["name"] = assistant.Name
+			}
+			if _, exists := i18nObj.Messages["description"]; !exists && assistant.Description != "" {
+				i18nObj.Messages["description"] = assistant.Description
+			}
+			flattened[locale] = i18nObj
+		}
+
+		i18n.Locales[id] = flattened
+	} else {
+		// No locales defined, create default with name and description
+		if assistant.Name != "" || assistant.Description != "" {
+			defaultLocales := make(map[string]i18n.I18n)
+			defaultLocales["en"] = i18n.I18n{
+				Locale: "en",
+				Messages: map[string]any{
+					"name":        assistant.Name,
+					"description": assistant.Description,
+				},
+			}
+			i18n.Locales[id] = defaultLocales
+		}
 	}
 
 	// Search options

agent/content/tools.go

@@ -36,8 +36,8 @@ func CallAgent(ctx *agentContext.Context, agentID string, message agentContext.M
 	messages := []agentContext.Message{message}
 
 	// Note: Connector is now in Options (call-level parameter), not Context
-	// For A2A calls, we use an empty Connector to let the agent use its default
-	opts := &agentContext.Options{Skip: &agentContext.Skip{History: true}, Writer: nil} // Skip history and output to the caller
+	// For A2A calls, skip history and output (we only need the response data)
+	opts := &agentContext.Options{Skip: &agentContext.Skip{History: true, Output: true}} // Skip history and output
 	response, err := agent.Stream(ctx, messages, opts)
 	if err != nil {
 		return "", fmt.Errorf("failed to call agent %s: %w", agentID, err)

agent/context/options.go

@@ -58,6 +58,9 @@ func OptionsFromMap(m map[string]interface{}) *Options {
 		if trace, ok := skipMap["trace"].(bool); ok {
 			skip.Trace = trace
 		}
+		if output, ok := skipMap["output"].(bool); ok {
+			skip.Output = output
+		}
 		opts.Skip = skip
 	}
 	if disableGlobalPrompts, ok := m["disable_global_prompts"].(bool); ok {

agent/context/output.go

@@ -297,16 +297,33 @@ func (ctx *Context) sendRaw(msg *message.Message) error {
 	return out.Send(msg)
 }
 
+// getWriter gets the effective Writer for the current context
+// Priority: Skip.Output > Stack.Options.Writer > ctx.Writer
+func (ctx *Context) getWriter() Writer {
+	// Check if output is explicitly skipped (for internal A2A calls)
+	if ctx.Stack != nil && ctx.Stack.Options != nil && ctx.Stack.Options.Skip != nil && ctx.Stack.Options.Skip.Output {
+		return nil // Explicitly disable output
+	}
+
+	// Check if current Stack has a Writer override
+	if ctx.Stack != nil && ctx.Stack.Options != nil && ctx.Stack.Options.Writer != nil {
+		return ctx.Stack.Options.Writer
+	}
+
+	return ctx.Writer
+}
+
 // getOutput gets the output writer for the context
 func (ctx *Context) getOutput() (*output.Output, error) {
-	if ctx.output != nil {
-		return ctx.output, nil
+	// Check if current Stack has cached output
+	if ctx.Stack != nil && ctx.Stack.output != nil {
+		return ctx.Stack.output, nil
 	}
 
 	trace, _ := ctx.Trace()
 	var options message.Options = message.Options{
 		BaseURL: "/",
-		Writer:  ctx.Writer,
+		Writer:  ctx.getWriter(), // Use getWriter() to resolve Writer priority
 		Trace:   trace,
 		Locale:  ctx.Locale,
 		Accept:  string(ctx.Accept),
@@ -318,10 +335,15 @@ func (ctx *Context) getOutput() (*output.Output, error) {
 		options.Capabilities = &caps
 	}
 
-	var err error
-	ctx.output, err = output.NewOutput(options)
+	out, err := output.NewOutput(options)
 	if err != nil {
 		return nil, err
 	}
-	return ctx.output, nil
+
+	// Cache to current Stack (each Stack has its own output with its own Writer)
+	if ctx.Stack != nil {
+		ctx.Stack.output = out
+	}
+
+	return out, nil
 }

agent/context/types.go

@@ -194,6 +194,7 @@ type AssistantInfo struct {
 type Skip struct {
 	History bool `json:"history"` // Skip saving chat history (for internal calls like title/prompt generation)
 	Trace   bool `json:"trace"`   // Skip trace logging
+	Output  bool `json:"output"`  // Skip output to client (for internal A2A calls that only need response data)
 }
 
 // MessageMetadata stores metadata for sent messages
@@ -232,7 +233,6 @@ type Context struct {
 
 	// Internal
 	trace           traceTypes.Manager    `json:"-"` // Trace manager, lazy initialized on first access
-	output          *output.Output        `json:"-"` // Output, it will be used to write response data to the client
 	messageMetadata *messageMetadataStore `json:"-"` // Thread-safe message metadata store for delta operations
 
 	// Model capabilities (set by assistant, used by output adapters)
@@ -312,6 +312,9 @@ type Stack struct {
 
 	// Metrics
 	DurationMs *int64 `json:"duration_ms,omitempty"` // Duration in milliseconds (calculated when completed)
+
+	// Runtime cache (not serialized)
+	output *output.Output `json:"-"` // Cached output instance for this stack
 }
 
 // Response the response

agent/i18n/builtin.go

@@ -21,7 +21,7 @@ func init() {
 			"assistant.agent.stream.skipping":        "Skipping output close (nested call)",
 			"assistant.agent.stream.close_error":     "Failed to close output",
 			"assistant.agent.completion.label":       "Agent Completion",
-			"assistant.agent.completion.description": "Final output from assistant",
+			"assistant.agent.completion.description": "Final output from {{name}}",
 
 			// LLM: providers/openai/openai.go Stream() function
 			"llm.openai.stream.label":        "LLM %s",
@@ -111,7 +111,7 @@ func init() {
 			"assistant.agent.stream.skipping":        "跳过输出关闭（嵌套调用）",
 			"assistant.agent.stream.close_error":     "关闭输出失败",
 			"assistant.agent.completion.label":       "智能体完成",
-			"assistant.agent.completion.description": "智能体最终输出",
+			"assistant.agent.completion.description": "{{name}} 最终输出",
 
 			// LLM: providers/openai/openai.go Stream() function
 			"llm.openai.stream.label":        "LLM %s",
@@ -173,7 +173,7 @@ func init() {
 			"assistant.agent.stream.skipping":        "跳过输出关闭（嵌套调用）",
 			"assistant.agent.stream.close_error":     "关闭输出失败",
 			"assistant.agent.completion.label":       "智能体完成",
-			"assistant.agent.completion.description": "智能体最终输出",
+			"assistant.agent.completion.description": "{{name}} 最终输出",
 
 			// LLM: providers/openai/openai.go Stream() function
 			"llm.openai.stream.label":        "LLM %s",

agent/llm/providers/openai/openai.go

@@ -211,7 +211,11 @@ func (p *Provider) Stream(ctx *context.Context, messages []context.Message, opti
 	var lastErr error
 
 	// Get Go context for cancellation support
+	// Read from Stack.Options if available (call-level override)
 	goCtx := ctx.Context
+	if ctx.Stack != nil && ctx.Stack.Options != nil && ctx.Stack.Options.Context != nil {
+		goCtx = ctx.Stack.Options.Context
+	}
 	if goCtx == nil {
 		goCtx = gocontext.Background()
 	}
@@ -808,7 +812,11 @@ func (p *Provider) Post(ctx *context.Context, messages []context.Message, option
 	var lastErr error
 
 	// Get Go context for cancellation support
+	// Read from Stack.Options if available (call-level override)
 	goCtx := ctx.Context
+	if ctx.Stack != nil && ctx.Stack.Options != nil && ctx.Stack.Options.Context != nil {
+		goCtx = ctx.Stack.Options.Context
+	}
 	if goCtx == nil {
 		goCtx = gocontext.Background()
 	}