Skip to content

Mongo DB

1. Component Intro

The MongoDB Component serves as the persistent data layer for your workflows. It allows the system to interact directly with a MongoDB database to perform CRUD (Create, Read, Update, Delete) operations. This enables your AI agents and workflows to store long-term user data, fetch specific records for context, or log complex event histories in a structured, NoSQL format.

Core JSON Structure

[[JSON]]

{
 "name": "MongoDB_Node",
 "type": "mongo_db",
 "description": "Interact with MongoDB for data retrieval and storage.",
 "output_type": "json",
 "inputs": {
  "operation": "fetch_one",
  "record_id": "{{user_profile.output.id}}",
  "limit": 50,
  "connection_string": "mongodb://localhost:27017",
  "database": "users_db",
  "collection": "users"
 }
}

2. Where to Use It

  • User Profile Retrieval: Fetching a user's preferences or history to personalize an LLM's response.

  • Audit Logging: Saving the final output of a workflow or a specific reasoning step for future review.

  • Knowledge Retrieval: Querying specific collections to provide factual context to an agent.

  • State Management: Updating a record's status (e.g., "completed", "pending") as it moves through different workflow branches.

3. How to Initialize

  1. Add Node: Drag the Mongo DB component from the Databases section of the library onto the canvas.

  2. Select Operation: Choose the desired action from the Operation dropdown (e.g., Single Fetch, Insert, Update).

  3. Configure Connection: Enter your Connection String (e.g., mongodb://localhost:27017). While this can be dynamic, it is usually a static global value.

  4. Target Database & Collection: Specify the Database name and the specific Collection you wish to query.

  5. Map Identifiers: Use the record_id field to pass dynamic values using the {{componentID.output}} syntax to find specific documents.

  6. Connect Ports: Link the input port (left) to the trigger and the output port (right) to the next processing or display node.

Mongo DB

Do's and Don'ts

✅ Do's

  • Use the Secret Vault: Always store your Connection String (which includes passwords) in the environment vault rather than hardcoding it.

  • Limit Query Results: Use the limit field when fetching multiple documents to prevent overwhelming the LLM with too much data.

  • Validate IDs: Ensure that the record_id being passed from previous nodes is in the correct format (e.g., String vs. ObjectID).

  • Index Your Collections: Ensure the fields you are querying frequently in your workflow are indexed in MongoDB to prevent timeout errors.

❌ Don'ts

  • Expose Connection Strings: Never leave raw MongoDB credentials visible in the workflow if it is being shared with unauthorized users.

  • Fetch Unnecessary Data: Avoid selecting "Fetch All" without filters; it wastes tokens and increases latency.

  • Bypass Input Connections: The node requires an input signal to execute; don't leave the input dot disconnected.

  • Ignore Error States: Don't forget to handle scenarios where a fetch_one operation returns no result using a Conditional node.

Tip: Data Pre-processing

Don't send raw MongoDB documents directly to an LLM. Use an Adapter Component after the MongoDB node to strip out system fields like _id or __v. This keeps the prompt clean and saves tokens.

Troubleshooting: Connection Timeouts

If your MongoDB node is failing to connect:

  1. IP Whitelisting: If using MongoDB Atlas, ensure the workflow engine's IP address is whitelisted in your database settings.
  2. Protocol Check: Verify if your connection string requires mongodb+srv:// for cloud instances vs mongodb:// for local ones.
  3. Collection Naming: MongoDB is case-sensitive. Verify that "Users" and "users" are not being swapped in your configuration.