Spring AI Instrumentation

OpenInference provides instrumentation for Spring AI applications using Spring’s Micrometer Observation API. This enables you to trace LLM calls, model parameters, token usage, and more using OpenTelemetry.

Installation

Gradle

Add the following to your build.gradle:

dependencies {
    implementation 'com.arize:openinference-instrumentation-springAI:0.1.0'
    implementation 'org.springframework.ai:spring-ai-model:1.0.1'
    implementation 'io.micrometer:micrometer-observation:1.15.1'
}

Maven

Add the following to your pom.xml:

<dependencies>
    <dependency>
        <groupId>com.arize</groupId>
        <artifactId>openinference-instrumentation-springAI</artifactId>
        <version>0.1.0</version>
    </dependency>
    <dependency>
        <groupId>org.springframework.ai</groupId>
        <artifactId>spring-ai-model</artifactId>
        <version>1.0.1</version>
    </dependency>
    <dependency>
        <groupId>io.micrometer</groupId>
        <artifactId>micrometer-observation</artifactId>
        <version>1.15.1</version>
    </dependency>
</dependencies>

Requirements

Java 17 or higher
Spring AI 1.0.0 or higher
OpenTelemetry Java 1.49.0 or higher
Micrometer Observation 1.15.0 or higher

Quick Start

Basic Setup

import com.arize.instrumentation.OITracer;
import com.arize.instrumentation.TraceConfig;
import com.arize.instrumentation.springAI.SpringAIInstrumentor;
import io.micrometer.observation.ObservationRegistry;
import io.opentelemetry.sdk.trace.SdkTracerProvider;
import org.springframework.ai.chat.model.ChatResponse;
import org.springframework.ai.chat.prompt.Prompt;
import org.springframework.ai.openai.OpenAiChatModel;
import org.springframework.ai.openai.OpenAiChatOptions;
import org.springframework.ai.openai.api.OpenAiApi;

public class MySpringAIApp {
    public static void main(String[] args) {
        // Initialize OpenTelemetry
        SdkTracerProvider tracerProvider = initializeOpenTelemetry();
        
        // Create OITracer
        OITracer tracer = new OITracer(
            tracerProvider.get("com.mycompany.my-app"),
            TraceConfig.getDefault()
        );
        
        // Register SpringAI instrumentation with Micrometer
        ObservationRegistry registry = ObservationRegistry.create();
        registry.observationConfig()
            .observationHandler(new SpringAIInstrumentor(tracer));
        
        // Configure OpenAI chat model
        OpenAiApi openAiApi = OpenAiApi.builder()
            .apiKey(System.getenv("OPENAI_API_KEY"))
            .build();
        
        OpenAiChatOptions options = OpenAiChatOptions.builder()
            .model("gpt-4")
            .temperature(0.7)
            .maxTokens(200)
            .build();
        
        OpenAiChatModel chatModel = OpenAiChatModel.builder()
            .openAiApi(openAiApi)
            .defaultOptions(options)
            .observationRegistry(registry)  // Register the observation registry
            .build();
        
        // Use the model - traces are automatically captured
        ChatResponse response = chatModel.call(
            new Prompt("What is the capital of France?")
        );
        
        System.out.println(response.getResult().getOutput());
    }
}

OpenTelemetry Setup

Basic Setup with Phoenix

import io.opentelemetry.api.common.Attributes;
import io.opentelemetry.api.common.AttributeKey;
import io.opentelemetry.exporter.otlp.trace.OtlpGrpcSpanExporter;
import io.opentelemetry.sdk.OpenTelemetrySdk;
import io.opentelemetry.sdk.resources.Resource;
import io.opentelemetry.sdk.trace.SdkTracerProvider;
import io.opentelemetry.sdk.trace.export.BatchSpanProcessor;
import java.time.Duration;

public static SdkTracerProvider initializeOpenTelemetry() {
    // Create resource with service information
    Resource resource = Resource.getDefault()
        .merge(Resource.create(Attributes.of(
            AttributeKey.stringKey("service.name"), "spring-ai-app",
            AttributeKey.stringKey("service.version"), "1.0.0"
        )));
    
    // Create OTLP exporter for Phoenix
    OtlpGrpcSpanExporter otlpExporter = OtlpGrpcSpanExporter.builder()
        .setEndpoint("http://localhost:4317")
        .setTimeout(Duration.ofSeconds(10))
        .build();
    
    // Create and configure tracer provider
    SdkTracerProvider tracerProvider = SdkTracerProvider.builder()
        .addSpanProcessor(BatchSpanProcessor.builder(otlpExporter)
            .setScheduleDelay(Duration.ofSeconds(1))
            .build())
        .setResource(resource)
        .build();
    
    // Register global OpenTelemetry instance
    OpenTelemetrySdk.builder()
        .setTracerProvider(tracerProvider)
        .buildAndRegisterGlobal();
    
    return tracerProvider;
}

Configuration

Custom Trace Configuration

Control what information is captured in traces:

import com.arize.instrumentation.TraceConfig;

// Configure what to hide in traces
TraceConfig config = TraceConfig.builder()
    .hideInputMessages(false)   // Set to true to hide input messages
    .hideOutputMessages(false)  // Set to true to hide output messages
    .build();

OITracer tracer = new OITracer(
    tracerProvider.get("my-app"),
    config
);

Chat Options

Configure model parameters:

import org.springframework.ai.openai.OpenAiChatOptions;

OpenAiChatOptions options = OpenAiChatOptions.builder()
    .model("gpt-4")
    .temperature(0.7)
    .maxTokens(500)
    .topP(0.9)
    .build();

OpenAiChatModel chatModel = OpenAiChatModel.builder()
    .openAiApi(openAiApi)
    .defaultOptions(options)
    .observationRegistry(registry)
    .build();

Spring Boot Integration

Application Configuration

import com.arize.instrumentation.OITracer;
import com.arize.instrumentation.TraceConfig;
import com.arize.instrumentation.springAI.SpringAIInstrumentor;
import io.micrometer.observation.ObservationRegistry;
import io.opentelemetry.api.GlobalOpenTelemetry;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.context.annotation.Bean;

@SpringBootApplication
public class Application {
    
    public static void main(String[] args) {
        SpringApplication.run(Application.class, args);
    }
    
    @Bean
    public ObservationRegistry observationRegistry() {
        // Create OITracer
        OITracer tracer = new OITracer(
            GlobalOpenTelemetry.getTracer("spring-ai-app"),
            TraceConfig.getDefault()
        );
        
        // Create and configure registry
        ObservationRegistry registry = ObservationRegistry.create();
        registry.observationConfig()
            .observationHandler(new SpringAIInstrumentor(tracer));
        
        return registry;
    }
}

Service Example

import org.springframework.ai.chat.model.ChatModel;
import org.springframework.ai.chat.model.ChatResponse;
import org.springframework.ai.chat.prompt.Prompt;
import org.springframework.stereotype.Service;

@Service
public class ChatService {
    
    private final ChatModel chatModel;
    
    public ChatService(ChatModel chatModel) {
        this.chatModel = chatModel;
    }
    
    public String chat(String message) {
        ChatResponse response = chatModel.call(new Prompt(message));
        return response.getResult().getOutput().getText();
    }
}

Tool Calling (Function Calling)

Spring AI supports function calling with automatic tracing:

import org.springframework.ai.tool.ToolCallback;
import org.springframework.ai.tool.function.FunctionToolCallback;
import java.util.function.Function;

public class WeatherExample {
    
    public record WeatherRequest(String location, String unit) {}
    public record WeatherResponse(double temperature, String unit) {}
    
    static class WeatherService implements Function<WeatherRequest, WeatherResponse> {
        public WeatherResponse apply(WeatherRequest request) {
            // Simulate weather API call
            return new WeatherResponse(72.5, "F");
        }
    }
    
    public static void main(String[] args) {
        // Setup OpenTelemetry and registry (as shown above)
        SdkTracerProvider tracerProvider = initializeOpenTelemetry();
        ObservationRegistry registry = createObservationRegistry(tracerProvider);
        
        // Create tool callback
        ToolCallback weatherTool = FunctionToolCallback.builder(
            "currentWeather",
            new WeatherService()
        )
        .description("Get the current weather in a location")
        .inputType(WeatherRequest.class)
        .build();
        
        // Configure model with tools
        OpenAiChatOptions options = OpenAiChatOptions.builder()
            .model("gpt-4")
            .temperature(0.4)
            .toolCallbacks(weatherTool)
            .build();
        
        OpenAiChatModel chatModel = OpenAiChatModel.builder()
            .openAiApi(openAiApi)
            .defaultOptions(options)
            .observationRegistry(registry)
            .build();
        
        // Ask a question that requires tool calling
        ChatResponse response = chatModel.call(
            new Prompt("What's the weather like in San Francisco?")
        );
        
        System.out.println(response.getResult().getOutput());
    }
}

Captured Trace Data

The instrumentation automatically captures:

LLM Model Information: Model name, provider
Input Messages: User prompts, system messages, conversation history
Output Messages: Model responses, assistant messages
Invocation Parameters: Temperature, max tokens, top_p, etc.
Token Usage: Prompt tokens, completion tokens, total tokens
Tool Calls: Function names, arguments, and responses
Message Roles: System, user, assistant, tool
Timing Information: Request latency and duration
Error Information: Exceptions and error messages

Multi-turn Conversations

Trace complete conversations with context:

import org.springframework.ai.chat.messages.Message;
import org.springframework.ai.chat.messages.UserMessage;
import org.springframework.ai.chat.prompt.Prompt;

// First turn
Prompt prompt1 = new Prompt("What is the capital of France?");
ChatResponse response1 = chatModel.call(prompt1);

// Second turn with context
Prompt prompt2 = Prompt.builder()
    .messages(
        new UserMessage("What is the capital of France?"),
        response1.getResult().getOutput(),
        new UserMessage("What about Germany?")
    )
    .build();

ChatResponse response2 = chatModel.call(prompt2);

Viewing Traces

Using Phoenix

Start Phoenix locally:

docker run -p 6006:6006 -p 4317:4317 arizephoenix/phoenix:latest

Run your instrumented application
View traces at http://localhost:6006

Using Other Backends

OpenInference instrumentation works with any OpenTelemetry-compatible backend:

Jaeger: Change the OTLP endpoint to your Jaeger instance
Zipkin: Use the Zipkin exporter
Cloud Providers: AWS X-Ray, Google Cloud Trace, Azure Monitor

Best Practices

Singleton Registry: Create a single ObservationRegistry instance and reuse it across your application
Spring Boot Integration: Use Spring’s dependency injection for ObservationRegistry
Set Service Name: Always set a meaningful service.name in your OpenTelemetry resource
Use Batch Processing: Use BatchSpanProcessor for better performance
Handle Secrets: Never log API keys in traces
Graceful Shutdown: Flush spans before application shutdown

Troubleshooting

No traces appearing

Verify the ObservationRegistry is properly configured with SpringAIInstrumentor
Ensure the registry is passed to your ChatModel via .observationRegistry()
Check that your OTLP endpoint is accessible
Enable debug logging for Spring AI observations

Missing token counts

Token usage is only available when the model provider returns usage metadata in the response.

Tool calls not traced

Ensure you’re using Spring AI 1.0.0 or higher, which includes observation support for tool calls.

Complete Example

Here’s a complete example with tool calling and conversation context:

import com.arize.instrumentation.OITracer;
import com.arize.instrumentation.TraceConfig;
import com.arize.instrumentation.springAI.SpringAIInstrumentor;
import io.micrometer.observation.ObservationRegistry;
import io.opentelemetry.api.common.Attributes;
import io.opentelemetry.sdk.OpenTelemetrySdk;
import io.opentelemetry.sdk.resources.Resource;
import io.opentelemetry.sdk.trace.SdkTracerProvider;
import org.springframework.ai.chat.messages.UserMessage;
import org.springframework.ai.chat.model.ChatResponse;
import org.springframework.ai.chat.prompt.Prompt;
import org.springframework.ai.openai.OpenAiChatModel;
import org.springframework.ai.openai.OpenAiChatOptions;
import org.springframework.ai.openai.api.OpenAiApi;

public class CompleteExample {
    public static void main(String[] args) {
        // 1. Initialize OpenTelemetry
        Resource resource = Resource.getDefault().merge(
            Resource.create(Attributes.builder()
                .put("service.name", "spring-ai-example")
                .build())
        );
        
        SdkTracerProvider tracerProvider = SdkTracerProvider.builder()
            .setResource(resource)
            .build();
        
        OpenTelemetrySdk.builder()
            .setTracerProvider(tracerProvider)
            .buildAndRegisterGlobal();
        
        // 2. Create OITracer and register with Micrometer
        OITracer tracer = new OITracer(
            tracerProvider.get("spring-ai"),
            TraceConfig.getDefault()
        );
        
        ObservationRegistry registry = ObservationRegistry.create();
        registry.observationConfig()
            .observationHandler(new SpringAIInstrumentor(tracer));
        
        // 3. Configure and create chat model
        OpenAiApi api = OpenAiApi.builder()
            .apiKey(System.getenv("OPENAI_API_KEY"))
            .build();
        
        OpenAiChatOptions options = OpenAiChatOptions.builder()
            .model("gpt-4")
            .temperature(0.7)
            .build();
        
        OpenAiChatModel chatModel = OpenAiChatModel.builder()
            .openAiApi(api)
            .defaultOptions(options)
            .observationRegistry(registry)
            .build();
        
        // 4. Use the model
        ChatResponse response = chatModel.call(
            new Prompt("Explain quantum computing in one sentence.")
        );
        
        System.out.println(response.getResult().getOutput());
        
        // 5. Shutdown and flush traces
        tracerProvider.forceFlush();
        tracerProvider.shutdown();
    }
}

Python Instrumentations

JavaScript Instrumentations

Java Instrumentations

Spring AI Instrumentation

Installation

Gradle

Maven

Requirements

Quick Start

Basic Setup

OpenTelemetry Setup

Basic Setup with Phoenix

Configuration

Custom Trace Configuration

Chat Options

Spring Boot Integration

Application Configuration

Service Example

Tool Calling (Function Calling)

Captured Trace Data

Multi-turn Conversations

Viewing Traces

Using Phoenix

Using Other Backends

Best Practices

Troubleshooting

No traces appearing

Missing token counts

Tool calls not traced

Complete Example

Resources

Build docs developers (and LLMs) love

Python Instrumentations

JavaScript Instrumentations

Java Instrumentations

​Installation

​Gradle

​Maven

​Requirements

​Quick Start

​Basic Setup

​OpenTelemetry Setup

​Basic Setup with Phoenix

​Configuration

​Custom Trace Configuration

​Chat Options

​Spring Boot Integration

​Application Configuration

​Service Example

​Tool Calling (Function Calling)

​Captured Trace Data

​Multi-turn Conversations

​Viewing Traces

​Using Phoenix

​Using Other Backends

​Best Practices

​Troubleshooting

​No traces appearing

​Missing token counts

​Tool calls not traced

​Complete Example

​Resources

Build docs developers (and LLMs) love

Installation

Gradle

Maven

Requirements

Quick Start

Basic Setup

OpenTelemetry Setup

Basic Setup with Phoenix

Configuration

Custom Trace Configuration

Chat Options

Spring Boot Integration

Application Configuration

Service Example

Tool Calling (Function Calling)

Captured Trace Data

Multi-turn Conversations

Viewing Traces

Using Phoenix

Using Other Backends

Best Practices

Troubleshooting

No traces appearing

Missing token counts

Tool calls not traced

Complete Example

Resources