DB Providers

Estimated reading: 6 minutes

DB Providers are the storage configurations in Robility Flow that connect projects to vector databases. Each project corresponds to a Robility Manager project and has its own DB Provider configuration. Every supported DB Provider is shared across both Knowledge Bases and Memory:

 a. Knowledge Bases: Store document embeddings and retrieve relevant passages during searches and agent interactions.
 b. Memory: Store conversation history so workflows and AI agents can recall context across sessions.

Why Vector DB Providers Are Needed

Traditional databases rely on exact keyword matching. To enable AI agents to understand context and intent, Robility Flow converts text into numerical representations called vector embeddings and stores them in a vector database.

Configuring DB Providers makes it possible to:

i. Retrieve documents and past conversations based on meaning and contextual similarity rather than exact phrasing.

ii. Switch from a built-in local database during development to a scalable, managed cloud or enterprise database in production without rebuilding workflows. Existing Knowledge Bases and Memory Bases are not migrated when the provider changes (see Limitations).

iii. Manage database credentials and storage backends independently for each project.

Supported Providers

All available providers below support both Knowledge Bases and Memory.

Provider Description Setup Type Status
Chroma Local Lightweight vector database bundled directly with Robility Flow. Best for local development and testing. Built-in (no credentials needed) Available
Chroma Cloud Fully managed, cloud-hosted vector database. UI Configuration Available
OpenSearch Distributed search and analytics engine supporting k-NN vector storage. UI Configuration Available
Postgres pgVector PostgreSQL database extended with pgvector for similarity search. Server Environment Variables Available
Astra DB Cloud-native vector database built on Apache Cassandra. N/A Coming Soon
MongoDB Atlas Managed cloud database with integrated vector search. N/A Coming Soon

Provider Status Indicators

The DB Providers page displays provider availability, connection state, and activation status using the following indicators:

Status Description
Available The provider is supported and ready to be configured.
Coming Soon The provider is listed in the interface for visibility but is not yet available for configuration.
Configured and reachable The provider is properly configured and Robility Flow can connect to the database.
Active The provider is currently selected as the primary provider for the project.

Configure a DB Provider

Before configuring an external provider, valid authentication credentials are required, and the database endpoint must be network-accessible from the Robility Flow environment.

1. Navigate to Settings → DB Providers.
2. Select a provider from the available list.
3. Enter the required configuration details, if applicable.
4. Optionally, select Test Connection to validate connectivity.
5. Select Save and Use [Provider] to save and activate the provider for the project.

Provider-Specific Setup

Chroma Local

Chroma Local requires no external credentials and stores vector embeddings locally on the Robility Flow server. It is active by default for every project.

a. To return to Chroma Local after using another provider, select Chroma Local and then Save and Use Chroma Local.
b. Because data resides on local server storage, persistence depends on the deployment environment. Regular server-level backups are recommended to prevent data loss.

Chroma Cloud

Connects Robility Flow to a managed Chroma Cloud instance. All fields are required.

Field Required Description
API Key Yes Authentication key for accessing the Chroma Cloud account.
Tenant Yes Chroma Cloud tenant identifier.
Database Yes Name of the target database within the tenant.
Region Yes Cloud deployment region where the database is hosted.

OpenSearch

Connects to an OpenSearch cluster using k-NN vector indexes. Robility Flow creates the required vector index using the provided credentials.

Dimension Matching Requirement: The vector dimension of the OpenSearch index must match the output dimension of the embedding model used in Robility Flow (for example, 1536 dimensions). Mismatched dimensions cause vector storage and retrieval to fail.

Field Description
Cluster URL* Endpoint URL of the OpenSearch cluster (for example, https://search.example.com:9200).
Username* OpenSearch authentication username.
Password* Password associated with the account.
Default Index Name Name of the index used for vector storage.
Vector Field Index field name where vector embeddings are stored.
Text Field Index field name where original text content is stored.
Use TLS (HTTPS) Encrypts communication between Robility Flow and OpenSearch using HTTPS.
Verify TLS Certificate Validates the server’s TLS certificate against trusted authorities. Applies only when TLS is enabled. Disabling is not recommended in production.

Postgres pgVector

Connects Robility Flow to a PostgreSQL database using the pgvector extension for vector storage and similarity search. Unlike other providers, Postgres pgVector is configured at the server level rather than through the Robility Flow interface.

a. Server-side prerequisites: The PostgreSQL connection details (host, port, database name, and credentials) must be set in the Robility Flow server environment variables, and the pgvector extension must be installed and enabled on the target database.

b. Activation: Once the server is configured and Robility Flow can connect to the database, the provider appears as Configured and reachable under Settings → DB Providers. Select Postgres pgVector, optionally select Test Connection, and then select Save and Use Postgres pgVector to activate it for the project.

Switching the Active DB Provider

The active DB Provider of a project can be changed at any time from Settings → DB Providers. The selected provider applies only to that project. Other projects retain their own configurations.

When the active DB Provider is switched:

a. Newly created Knowledge Bases are assigned to the newly selected provider by default.
b. Existing Knowledge Bases are unaffected and continue using their originally assigned provider.
c. The newly selected provider is preselected during Memory Base creation, and any other configured provider can be chosen instead.
d. Existing Memory Bases are unaffected and continue using their originally assigned provider.

Best Practices

a. Use Chroma Local for local development and testing, Chroma Cloud for managed cloud storage, OpenSearch when OpenSearch infrastructure is already in place, and Postgres pgVector when PostgreSQL is part of the deployment architecture.
b. Run Test Connection to verify credentials and network reachability before saving a new provider.
c. Keep Verify TLS Certificate enabled for OpenSearch in production environments.

Limitations

a. Only one DB Provider can be active per project at a time.
b. Changing the active DB Provider does not migrate or alter existing Knowledge Bases or Memory Bases. They remain bound to their original providers.
c. Postgres pgVector cannot be configured through the UI and requires server-side environment variables.
d. Chroma Local stores data directly on the Robility Flow host server, so persistence depends on local disk management and server-level backups.

Share this Doc

DB Providers

Or copy link

CONTENTS
Robility Chatbot
Robility Assistant
Online