AgentNexus

AgentNexus is a Java-based enterprise intelligent agent infrastructure that stores models, tools, memories, prompts, and other core components in a database and supports API-driven, hot-reloadable agent assembly. It leverages a Brain model for autonomous multi-agent orchestration, a pluggable tools system, a production-ready RAG layer with vector storage via PgVector, and API-triggered dynamic recomposition of agents without restarting the service. You can trigger updates, load new configurations, and assemble temporary agent teams on the fly through REST endpoints or via Nacos as a config source. This enables real-time, low-latency reconfiguration for complex tasks such as autonomous task planning, tool invocation, and memory-enabled conversations across sessions. The system also supports building workflows by chaining multiple agents and scheduling tasks with Cron for automated processes such as reporting or data analysis. To get started, deploy the service (for example via Docker), ensure your databases and Nacos configuration are accessible, and begin querying the API to create or modify agents, assign tools, and manage memory to support your business workflows.

Prerequisites:

• Java 17+ (or a compatible JDK)
• Maven 3.6+
• MySQL 8.0+ and PostgreSQL 15+ (with pgvector extension for vector storage)
• Nacos 2.2.0+ (config center)
• Docker (optional but recommended for quick setup)

Install steps:

1. Clone the repository (or pull the pre-built Docker image if provided):

   • git clone https://github.com/wozhenbang2004/AgentNexus.git
   • cd AgentNexus

2. Set up databases:

   • Create MySQL database (e.g., agent_nexus) and run the schema.sql from sql/ to initialize tables.
   • Create PostgreSQL database and install/enable the pgvector extension.

3. Configure Nacos:

   • Start Nacos (Docker or standalone).
   • In Nacos, create a Data ID named agent-nexus-dev.yml (or any name you prefer) under the DEFAULT_GROUP with YAML format. Paste application.yml content as a template and modify database URLs and API keys as needed.

4. Build and run (if building from source):

   • mvn clean package -DskipTests
   • java -jar target/AgentNexus.jar

5. Docker-based run (if using Docker image):

   • Ensure environment variables for DB, PgVector, and Nacos are configured (see mcp_config for placeholders).
   • Run: docker run -i your-image-name with the appropriate -e VAR=value options as needed, or use docker-compose if provided by the project.

6. Access API documentation:

   • Swagger: http://localhost:8091/swagger-ui.html (port may vary based on configuration)

Note: Adjust ports, endpoints, and environment variables according to your deployment environment.

Tips and common issues:

• Ensure the PgVector extension is installed and accessible for PostgreSQL; otherwise, the RAG features may fail.
• If you use Nacos for dynamic config, verify the Data ID and Group match what AgentNexus expects; mismatch can lead to no updates being loaded.
• When hot-reloading configurations, changes to models, tools, or memories should reflect within milliseconds, but very large configurations may take longer to assemble.
• Enable tool auto-registration by ensuring any Bean implementing java.util.function.Function is registered in the Spring context; this allows tools to be automatically discovered as Agent tools.
• Check environment variable mappings carefully when using Docker to ensure database connections and API keys are correctly provided.
• For persistent chat memory, ensure the ChatMemoryRepository is connected to the underlying database to achieve cross-session memory.