This repository is now part of the alibaba/spring-ai-alibaba project. The source code has been fully migrated, please submit and follow updates there.

Spring AI Alibaba Repo: https://github.com/alibaba/spring-ai-alibaba

Spring AI Alibaba Website: https://java2ai.com

Spring AI Alibaba Website Repo: https://github.com/springaialibaba/spring-ai-alibaba-website

English | 中文

Project Background

Agent Studio is an AI Agent development and evaluation platform based on Spring AI Alibaba, designed to provide developers and enterprises with a complete AI Agent lifecycle management solution. The platform supports a complete workflow from Prompt engineering, dataset management, evaluator configuration to experiment execution and result analysis, helping users quickly build, test, and optimize AI Agent applications.

Core Features

🤖 Prompt Management

• Prompt Template Management: Create, update, and delete Prompt templates
• Version Control: Support for Prompt version management and history tracking
• Real-time Debugging: Provide online Prompt debugging and streaming responses
• Session Management: Support for multi-turn conversation session management

📊 Dataset Management

• Dataset Creation: Support for importing and creating datasets in multiple formats
• Version Management: Dataset version control and history management
• Data Item Management: Fine-grained data item CRUD operations
• Create from Trace: Support for creating datasets from OpenTelemetry trace data

⚖️ Evaluator Management

• Evaluator Configuration: Support for creating and configuring various evaluators
• Template System: Provide evaluator templates and custom evaluation logic
• Debugging Features: Support for online evaluator debugging and testing
• Version Management: Evaluator version control and release management

🧪 Experiment Management

• Experiment Execution: Automated execution of evaluation experiments
• Result Analysis: Detailed experiment result analysis and statistics
• Experiment Control: Support for starting, stopping, restarting, and deleting experiments
• Batch Processing: Support for batch experiment execution and result comparison

📈 Observability

• Trace Tracking: Integrated OpenTelemetry providing complete trace tracking
• Service Monitoring: Support for service list and overview statistics
• Trace Analysis: Provide detailed Trace details and Span analysis

🔧 Model Configuration

• Multi-model Support: Support for mainstream AI models including OpenAI, DashScope, DeepSeek
• Configuration Management: Unified configuration and management of model parameters
• Dynamic Switching: Support for dynamic updates of model configuration at runtime

System Architecture

Overall Architecture

🚀 Quick Start

Prerequisites

• 🐳 Docker (for containerized deployment) + Docker Compose: 2.0+
• ☕ Java 17+ (for source code execution) + Maven: 3.8+
• 🌐 AI Model Provider API Keys, supporting OpenAI, DashScope, DeepSeek

Running from Source Code

1. Clone the Project

git clone https://github.com/spring-ai-alibaba/spring-ai-alibaba-admin.git
cd spring-ai-alibaba-admin

2. Configure Your API Keys

Modify the model configuration in spring-ai-alibaba-admin-server-start/model-config.yaml according to your model provider.

• If you use DashScope, please refer to the model-config-dashscope.yaml template for configuration
• If you use DeepSeek, please refer to the model-config-deepseek.yaml template for configuration
• If you use OpenAI, please refer to the model-config-openai.yaml template for configuration

💡 Get Your DashScope API Key: Visit Alibaba Cloud Bailian Console to get a free API key.

3. Nacos Configuration (Optional)

If you need to modify the Nacos address, please update the configuration in the spring-ai-alibaba-admin-server-start/src/main/resources/application.yml file

nacos:
  server-addr: ${nacos-address}

4. Start SAA Admin

Execute the startup script in the root directory. This script will help you start the database-related services

sh start.sh

Start the application in the spring-ai-alibaba-admin-server-start directory

mvn spring-boot:run

5. Access the Application

Open your browser and visit http://localhost:8080 to use the SAA Admin platform.

At this point, you can already manage, debug, evaluate, and observe prompts on the platform. If you expect your Spring AI Alibaba Agent application to integrate with Nacos for prompt loading and dynamic updates, and observe the online running status, you can refer to step 6 to configure your AI Agent application.

6. Connect Your AI Agent Application

In your Spring AI Alibaba Agent application, add the following dependencies

<dependencies>
    <!-- Introduce spring ai alibaba agent nacos proxy module -->
    <dependency>
        <groupId>com.alibaba.cloud.ai</groupId>
        <artifactId>spring-ai-alibaba-agent-nacos</artifactId>
        <version>{spring.ai.alibaba.version}</version>
    </dependency>

    <!-- Introduce observability module -->

    <dependency>
        <groupId>com.alibaba.cloud.ai</groupId>
        <artifactId>spring-ai-alibaba-autoconfigure-arms-observation</artifactId>
        <version>{spring.ai.alibaba.version}</version>
    </dependency>
    
    
    <!-- For implementing various OTel related components, such as automatic loading of Tracer, Exporter -->
    <dependency>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-actuator</artifactId>
    </dependency>
    
    <!-- For connecting micrometer generated metrics to otlp format -->
    <dependency>
        <groupId>io.micrometer</groupId>
        <artifactId>micrometer-registry-otlp</artifactId>
    </dependency>
    
    <!-- For replacing micrometer underlying trace tracer with OTel tracer -->
    <dependency>
        <groupId>io.micrometer</groupId>
        <artifactId>micrometer-tracing-bridge-otel</artifactId>
    </dependency>
    
    <!-- For reporting spans generated by OTel tracer according to otlp protocol -->
    <dependency>
        <groupId>io.opentelemetry</groupId>
        <artifactId>opentelemetry-exporter-otlp</artifactId>
    </dependency>

    <dependency>
        <groupId>org.springframework.ai</groupId>
        <artifactId>spring-ai-autoconfigure-model-tool</artifactId>
        <version>1.0.0</version>
    </dependency>
</dependencies>

Specify Nacos address and promptKey

    spring.ai.alibaba.agent.proxy.nacos.serverAddr={replace nacos address, example: 127.0.0.1:8848}
    spring.ai.alibaba.agent.proxy.nacos.username={replace nacos username, example: nacos}
    spring.ai.alibaba.agent.proxy.nacos.password={replace nacos password, example: nacos}
    spring.ai.alibaba.agent.proxy.nacos.promptKey={replace with promptKey, example: mse-nacos-helper} 

Set observability parameters

    management.otlp.tracing.export.enabled=true
    management.tracing.sampling.probability=1.0
    management.otlp.tracing.endpoint=http://{admin address}:4318/v1/traces
    management.otlp.metrics.export.enabled=false
    management.otlp.logging.export.enabled=false
    management.opentelemetry.resource-attributes.service.name=agent-nacos-prompt-test
    management.opentelemetry.resource-attributes.service.version=1.0
    spring.ai.chat.client.observations.log-prompt=true
    spring.ai.chat.observations.log-prompt=true
    spring.ai.chat.observations.log-completion=true
    spring.ai.image.observations.log-prompt=true
    spring.ai.vectorstore.observations.log-query-response=true
    spring.ai.alibaba.arms.enabled=true
    spring.ai.alibaba.arms.tool.enabled=true
    spring.ai.alibaba.arms.model.capture-input=true
    spring.ai.alibaba.arms.model.capture-output=true

License

This project is open source under the Apache License 2.0 license.

Contributing

We welcome submitting Issues and Pull Requests to help improve the project.