Agentic Predictive Maintenance System

Manufacturers are moving beyond traditional predictive maintenance—it's not just about forecasting failures, but about acting on issues instantly and autonomously. This project demonstrates how agentic AI, orchestrated by LangGraph.js and powered by MongoDB Atlas and AWS Bedrock, enables multi-agent systems that detect problems and coordinate rapid, intelligent responses across the shop floor.

This demo showcases:

• Autonomous action: AI agents diagnose, plan, and execute maintenance tasks in real time, minimizing human intervention and downtime.
• Operational agility: The system adapts to new data, equipment, and workflows, supporting continuous improvement.
• Unified, scalable data foundation: MongoDB makes it easy to build, operate, and evolve agentic AI solutions—handling diverse data, enabling fast search, and supporting real-time decision-making.

Architecture

How it works:

1. Detection: Agents monitor machine telemetry and logs, triggering alerts on anomalies.
2. Diagnosis: The Failure Agent uses MongoDB’s flexible data model and vector search to rapidly analyze root causes.
3. Action: The Work Order Agent drafts and routes maintenance tasks, leveraging historical data and inventory.
4. Optimization: The Planning Agent schedules work to minimize disruption, using real-time production and staff data.

Why MongoDB?

• Unified data layer: Handles structured, unstructured, and time series data for all agents and workflows.
• Real-time search & retrieval: Atlas Search and vector search enable fast, context-rich decision-making.
• Scalable, adaptable foundation: Easily extend to new agents, data sources, and operational needs.

This architecture lets manufacturers automate not just prediction, but coordinated action—unlocking the next level of operational excellence.

Getting Started

Prerequisites

• Node.js 18+
• MongoDB (local or Atlas)
• AWS Account with Bedrock access
• AWS CLI installed locally

Setup

1. Configure AWS credentials for Bedrock access: Run one of the following commands to set up your AWS credentials locally:

   aws configure

   Or with SSO (recommended):

   aws configure sso

2. Install dependencies:

   npm install

3. Set up environment variables: Copy the example environment file and update it with your credentials:

   cp .env.example .env

   Then edit your .env file and set the following variables:

   MONGODB_URI="<your-mongodb-uri>"
   DATABASE_NAME="agentic_predictive_maintenance"
   AWS_REGION="us-east-1"
   AWS_PROFILE="default"
   COMPLETION_MODEL="us.anthropic.claude-3-5-haiku-20241022-v1:0"
   EMBEDDING_MODEL="cohere.embed-english-v3"

4. Seed the demo database: To initialize the demo with all required collections, embeddings, and indexes, run:

   npm run seed

   This step ensures your database is ready for the demo application.

5. Start the application: You can now launch the app in development mode:

   npm run dev

   Or with Docker:

   docker-compose up

Open http://localhost:8080 in your browser to explore the demo.

Personalizing and Extending the Demo

This demo is designed to be flexible and extensible. Here are some ways you can tailor it to your needs:

Production Calendar Customization

• By default, the production calendar is populated with the next 6 months from today.
• You can manually edit the production_calendar collection in MongoDB to adjust tasks and schedules.
• To auto-populate the calendar, use:

  npm run generate_calendar <months>

  Replace <months> with the number of months you want to generate. Note: This script will remove the previous calendar before creating a new one.

Adding Documentation, Manuals, or Interviews

• You can add your own documentation, manual chunks, or interview transcripts directly to the relevant collections (manuals, interviews, etc.).
• After adding new documents, run:

  npm run embed

  This will embed the new content and update the vector indexes for search and retrieval.
• To customize which fields are embedded or change the embedding field name, edit the configuration in scripts/config.js:

  // Example config.js entry
  const config = [
    {
      collection: "manuals",
      textFields: ["section", "text"],
      embeddingField: "embedding",
      indexName: "default",
      similarity: "cosine",
      numDimensions: 1024, // This value depends on the embedding model selected
    },
    // Add more collections as needed
  ];
  export default config;

  • collection: The MongoDB collection name.
  • textFields: Array of fields to concatenate and embed.
  • embeddingField: Field name to store the embedding.
  • indexName, similarity, numDimensions: Vector index settings. Note: numDimensions should match the output dimension of your selected embedding model.

Creating New Agents and Tools

• To create a new agent, copy the test folder inside src/agents and give it a new name.
• Inside your new agent folder, you can:
  • Edit tools.js to define the agent's tools and capabilities.
  • Edit graph.js to set the system prompt and agent logic.
• Once your agent is ready, add it to src/agents/config.js so it appears in the demo.
• You can test your agent from the Agent Sandbox section of the application.

---

Thank you for exploring the Agentic Predictive Maintenance demo! This repository is maintained by MongoDB Industry Solutions. We encourage you to experiment, extend, and adapt the system to your own use cases. If you have questions or feedback, please reach out at industry.solutions@mongodb.com or open an issue.

Enjoy building the future of modern manufacturing!