feat(dashscope): Support stream tool call

[mengnankkkk]

Describe what this PR does / why we need it

在 DashScopeChatOptions 中新增了 enableStreamToolCalls 配置项（默认 false）
当设置为 true 时，工具调用的片段会在工具执行之前就开始流式返回

Does this pull request fix one issue?

Close #111

Describe how you did it

测试代码：

 void testStreamToolCallsWithRealApiIntegration() {
        // Get API key from environment
        String apiKey = System.getenv("AI_DASHSCOPE_API_KEY");
        if (apiKey == null || apiKey.isEmpty()) {
            throw new IllegalStateException("AI_DASHSCOPE_API_KEY environment variable is not set");
        }

        // Create real DashScope API instance
        DashScopeApi realApi = DashScopeApi.builder().apiKey(apiKey).build();

        // Create a weather request class
        record WeatherRequest(String location) {}

        // Create a real tool callback for getting weather
        ToolCallback weatherTool = FunctionToolCallback.builder("get_weather", 
                (WeatherRequest request) -> {
                    // Simulate weather API call
                    return "The weather in " + request.location() + " is sunny with 25°C";
                })
                .description("Get the current weather for a given location")
                .inputType(WeatherRequest.class)
                .build();

        // Configure options with enableStreamToolCalls=true
        DashScopeChatOptions options = DashScopeChatOptions.builder()
                .model("qwen-plus") // Use a model that supports tool calls
                .stream(true)
                .enableStreamToolCalls(true)
                .toolCallbacks(List.of(weatherTool))
                .build();

        // Create chat model with real API
        DashScopeChatModel realChatModel = DashScopeChatModel.builder()
                .dashScopeApi(realApi)
                .defaultOptions(options)
                .build();

        // Create a prompt that should trigger tool execution
        UserMessage userMessage = new UserMessage("What's the weather like in Beijing today?");
        Prompt prompt = new Prompt(List.of(userMessage), options);

        // Track states during streaming
        AtomicBoolean receivedToolCallChunks = new AtomicBoolean(false);
        AtomicBoolean receivedToolExecutionResult = new AtomicBoolean(false);
        AtomicBoolean hasToolCalls = new AtomicBoolean(false);

        // Execute streaming request
        List<ChatResponse> responses = realChatModel.stream(prompt)
                .doOnNext(response -> {
                    System.out.println("=== Received Chunk ===");
                    System.out.println("Text: " + response.getResult().getOutput().getText());
                    
                    // Check if this chunk has tool calls
                    if (!response.getResult().getOutput().getToolCalls().isEmpty()) {
                        hasToolCalls.set(true);
                        receivedToolCallChunks.set(true);
                        System.out.println("Tool Calls: " + response.getResult().getOutput().getToolCalls());
                    }

                    // Check finish reason
                    String finishReason = response.getResult().getMetadata().getFinishReason();
                    System.out.println("Finish Reason: " + finishReason);

                    // Check if tool execution result is present
                    if (response.getResult().getOutput().getText().contains("sunny") 
                            || response.getResult().getOutput().getText().contains("25°C")) {
                        receivedToolExecutionResult.set(true);
                        System.out.println("Tool execution result detected in response");
                    }
                })
                .collectList()
                .block();

        // Verify results
        assertThat(responses).isNotNull();
        assertThat(responses).isNotEmpty();

        // Verify that we received tool call chunks during streaming
        assertThat(receivedToolCallChunks.get())
                .as("Should have received tool call chunks during streaming")
                .isTrue();

        // Verify that tool execution happened and result was received
        assertThat(receivedToolExecutionResult.get())
                .as("Should have received tool execution result in final response")
                .isTrue();

        // Print summary
        System.out.println("\n=== Test Summary ===");
        System.out.println("Total chunks received: " + responses.size());
        System.out.println("Tool calls detected: " + hasToolCalls.get());
        System.out.println("Tool execution result received: " + receivedToolExecutionResult.get());

        // Verify the final response contains expected content
        String finalText = responses.stream()
                .map(r -> r.getResult().getOutput().getText())
                .reduce("", (a, b) -> a + b);
        
        System.out.println("Final combined text: " + finalText);
        
        // The final text should contain either the tool call information or the weather result
        assertThat(finalText)
                .as("Final response should contain weather-related information")
                .isNotEmpty();
    }

得到结果

Running com.alibaba.cloud.ai.dashscope.chat.DashScopeChatModelTests
OpenJDK 64-Bit Server VM warning: Sharing is only supported for boot loader classes because bootstrap classpath has been appended
21:57:13.649 [main] WARN com.alibaba.cloud.ai.tool.validator.DefaultToolCallValidator -- Filtering out toolCall with null function name: ToolCall[id=tool-call-id, type=function, function=ChatCompletionFunction[name=null, arguments={"location": "Beijing"}], index=null]
=== Received Chunk ===
Text:
Tool Calls: [ToolCall[id=call_1f37553f6d0540208c9493, type=function, name=get_weather, arguments={"location": "]]
Finish Reason:
21:57:15.047 [HttpClient-2-Worker-0] WARN com.alibaba.cloud.ai.tool.validator.DefaultToolCallValidator -- Filtering out toolCall with null function name: ToolCall[id=, type=function, function=ChatCompletionFunction[name=null, arguments=Beijing"}], index=0]
=== Received Chunk ===
Text:
Finish Reason: TOOL_CALLS
=== Received Chunk ===
Text: The
Finish Reason:
=== Received Chunk ===
Text:  weather in
Finish Reason:
=== Received Chunk ===
Text:  Beijing today is
Finish Reason:
=== Received Chunk ===
Text:  sunny with a
Finish Reason:
Tool execution result detected in response
=== Received Chunk ===
Text:  temperature of 25°C
Finish Reason:
Tool execution result detected in response
=== Received Chunk ===
Text: .
Finish Reason:
=== Received Chunk ===
Text:
Finish Reason: STOP

=== Test Summary ===
Total chunks received: 9
Tool calls detected: true
Tool execution result received: true
Final combined text: The weather in Beijing today is sunny with a temperature of 25°C.

Describe how to verify it

Special notes for reviews

[guanxuc]

DashScopeChatModelTests里麻烦加一下流式输出toolcall的单测

[guanxuc]

@mengnankkkk stream tool call好像丢失了arguments参数，这个有空帮看下

non-stream tool call:

toolCalls=[ToolCall[id=call_d5c655db312244389a49b7, type=function, name=getCurrentWeather, arguments={"location": "San Francisco, CA", "unit": "C"}]]

stream tool call:

toolCalls=[ToolCall[id=call_f469e0c8e41945aa912b02, type=function, name=getCurrentWeather, arguments=]]