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 VectorDBs:

Managed VectorDBs

DuploCloud deploys and manages VectorDBs 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 VectorDBs if you want to keep all components within your cloud account, prefer zero setup, or don’t have an external VectorDB provider.

Third-Party VectorDBs

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

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

Integrating VectorDBs with DuploCloud

The first step for working with VectorDBs in DuploCloud is to integrate a VectorDB 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 VectorDBs (e.g., Pinecone), make sure you have your API endpoint and any necessary authentication information.

• For managed VectorDBs (e.g., Chroma, MilvusDB), ensure your Kubernetes environment is ready to deploy services.

Integrating Third-Party VectorDBs

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 VectorDB.
Vector DB Type	Select pinecone for a third-party VectorDB.
API Endpoint	Enter the endpoint URL for your Pinecone instance.
Metadata	Optionally, enter key-value pairs to organize or filter this VectorDB later.

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

Vector DBs page in the DuploCloud AI Suite Studio

Integrating Managed VectorDBs

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

Adding a Managed VectorDB

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

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 VectorDB.

1. Click Submit to save the VectorDB.

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 VectorDB

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

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

2. Select the VectorDB 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 VectorDB name; can be customized if desired.
Docker Image	Auto-filled for managed VectorDBs. For third-party VectorDBs, confirm or provide the correct image if applicable.
Deployment Environment Variables	Define any environment variables required for your VectorDB.
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.

Deployment tab for the VectorDB 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 VectorDB.

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

2. Select the VectorDB 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 VectorDB’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.

   

   Ingested Jobs tab in the DuploCloud Platform

Viewing Ingestion Jobs

After uploading and ingesting documents into a VectorDB, 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 VectorDB 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 Agents

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