VectorDB

In the DuploCloud AI Suite, the Vector Database (VectorDB) enables you to upload documents, such as architecture diagrams, runbooks, internal wikis, or API references, that you want the AI agent to use for context during conversations. These documents are transformed into high-dimensional vector representations, which allow the system to retrieve the most relevant content when the agent processes your queries. This enhanced context allows the agent to better understand your cloud environment, use your terminology, and align with your organization’s best practices.

DuploCloud supports two types of Vector DBs:

Managed Vector DBs

DuploCloud deploys and manages Vector DBs directly within your Kubernetes environment, handling setup, environment variables, and connectivity for seamless integration. Supported engines include:

  • Chroma: Lightweight, fast, ideal for local AI workloads.

  • MilvusDB: Scalable for high-performance vector search at large scale.

Use managed Vector DBs if you want to keep all components within your cloud account, prefer zero setup, or don’t have an external vector DB provider.

Third-Party Vector DBs

These are externally hosted vector databases like Pinecone or PostgreSQL that DuploCloud connects to but does not manage or deploy.

Choose third-party Vector DBs if you already use an external provider or need to integrate with specialized vector DB services outside your Kubernetes cluster.

Integrating Vector DBs with DuploCloud

The first step for working with vector databases in DuploCloud is to integrate a Vector DB with the DuploCloud AI Suite. This allows the platform to store and retrieve vectorized content.

Prerequisites

  • You must have access to the AI Suite feature in the DuploCloud Portal.

  • For third-party Vector DBs (e.g., Pinecone), make sure you have your API endpoint and any necessary authentication information.

  • For managed Vector DBs (e.g., Chroma, Milvus), ensure your Kubernetes environment is ready to deploy services.

Integrating a Third-Party Vector DB

To integrate a third-party vector database, such as Pinecone:

  1. In the DuploCloud Platform, navigate to AI Suite → Studio → Vector DBs.

  2. Click Add. The Add Vector Database pane displays.

The Add Vector Database pane
  1. Complete the following fields:

Name

Enter a friendly name for the Vector DB.

Vector DB Type

Select pinecone for a third-party Vector DB.

API Endpoint

Enter the endpoint URL for your Pinecone instance.

Metadata

Optionally, enter key-value pairs to organize or filter this Vector DB later.

  1. Click Submit to save the Vector DB. Your third-party Vector DB is ready to use immediately.

The Vector DBs page in the DuploCloud AI Suite Studio

Integrating a Managed Vector DB

To integrate a DuploCloud-managed Vector DB (Chroma or MilvusDB), add and then deploy the database in the DuploCloud Platform.

Adding a Managed Vector DB

  1. In the DuploCloud Platform, navigate to AI Suite → Studio → Vector DBs.

  2. Click Add. The Add Vector Database pane displays.

    The Add Vector Database pane in the DuploCloud Portal
  3. Complete the following fields:

Name

Enter a friendly name for the Vector DB.

Vector DB Type

Select your Vector DB type, (e.g., chroma or milvusdb).

Deployment Environment Variables

Optionally, add custom environment variables (e.g., API keys, flags).

Metadata

Optionally, enter key-value pairs for organizing or tagging the Vector DB.

  1. Click Submit to save the Vector DB.

Note: Adding a Managed Vector DB within DuploCloud saves the configuration, but does not deploy the database. Deployment is required before you can upload or ingest files.

Deploying a Managed Vector DB

After adding a managed Vector DB, deploy it to make it active and usable.

  1. Navigate to AI Suite → Studio → Vector DBs.

  2. Select the Vector DB from the NAME column.

  3. Select the Deployment tab, and click Deploy. The Deploy pane displays.

The Deploy pane for the duplo-managed-db Vector DB
  1. Review or complete the deployment fields:

Name

Auto-filled with the Vector DB name; can be customized if desired.

Docker Image

Auto-filled for managed Vector DBs. For third-party Vector DBs, confirm or provide the correct image if applicable.

Deployment Environment Variables

Define any environment variables required for your Vector DB.

Advanced Options

Optional settings such as replicas, service name, network, volumes, and load balancer listeners.

  1. Choose either:

    • Quick Deploy to deploy with default settings immediately.

    • Advanced to customize deployment options before deploying.

  2. If using Advanced Deploy, click Next to navigate through additional configuration screens, then click Create to start deployment. For Quick Deploy, click Quick Deploy.

  3. Monitor the deployment status; it usually takes 4 to 5 minutes. Once complete, the status on the Deployment tab will show Running.

The Deployment tab for the Vector DB with status Running

Uploading Files

Upload your source documents or data files to your AWS S3 storage to make your files available for processing and ingestion into the Vector DB.

  1. In the DuploCloud portal, go to AI Suite → Studio → Vector DBs.

  2. Select the Vector DB you want to upload files to from the NAME column.

  3. Select the Uploaded Files tab.

  4. Click Browse. This will open your AWS S3 console where you can select the files you want to upload.

    The AWS S3 Console
  5. Select the files to upload (Click Upload Files → Add File, select your file(s), and click Open).

  6. Return to the DuploCloud Uploaded Files tab, and click Sync to update the Vector DB’s Uploaded Files list. The uploaded files are displayed on the Uploaded Files tab.

The Uploaded Files tab in the DuploCloud Platform

Ingesting Files

Ingesting transforms your uploaded files into vector representations.

  1. In the DuploCloud portal, go to AI Suite → Studio → Vector DBs.

  2. Select the Uploaded Files tab.

  3. Click the checkbox(s) to select one or more files you want to ingest.

  4. Click Ingest. The Trigger Build pane displays.

    The Trigger Build pane
  5. Configure the fields as needed:

    • Review the Docker Image: This field is prepopulated with the container used for ingestion. You usually do not need to change it unless you're using a custom image.

    • Timeout: Enter the maximum duration (in minutes) for the ingestion job.

    • Custom Meta Data (Optional): Use key-value pairs to customize how the ingestion job processes your data. Common options include:

      • chunk_size: Size of each text chunk in characters (e.g., 1000).

      • chunk-overlap: Number of overlapping characters between chunks (e.g., 100).

  6. Click Submit to trigger the ingestion job. Monitor the ingestion status on the Ingested Jobs tab.

    The Ingested Jobs tab in the DuploCloud Platform

Viewing Ingestion Jobs

After uploading and ingesting documents into a Vector DB, you can monitor the status and output of each job in the Ingestion Jobs tab. This tab provides access to ingestion history, logs, and detailed configuration metadata to help validate behavior and troubleshoot issues.

  • In the DuploCloud portal, go to AI Suite → Studio → Vector DBs.

  • Select the Ingestion Jobs tab.

  • Click the menu icon () next to the job you want to inspect.

    The Ingestion Jobs tab with the Logs and Details menu options highlighted
  • Choose one of the following options:

    • Logs: View output that includes source file paths, chunking progress, chunk IDs, and any success or error messages.

      The Logs for an ingested job
    • Details: Open a structured JSON summary showing Vector DB type and provider, API endpoint, file paths ingested, output directory, chunking configuration, embedding model, and other technical metadata.

      The Details for an ingested job

Using VectorDBs with DuploCloud AI Studio

To learn how to integrate the files uploaded to your VectorDBs with the DuploCloud AI agent, see the DuploCloud documentation for creating AI Agents.

Last updated

Was this helpful?