Drivers
What is a driver?
A driver is a way to make a given information source available within your Retrieval Agent.
Its name (and description when applicable) is important because in some cases, the Retrieval Agent will use it to evaluate if the corresponding source is valuable for the query it is currently processing.
How to create a driver?
On the left menu, click on Drivers, then use the pull-down menu called Add driver on the right to select the kind of driver you need.
Depending on the drivers, you will need to provide different information, including crednetials or API keys, see below.
Supported drivers
NucliaDB
This driver gives access to your Agentic RAG Knowledge Boxes.
You need to select a Knowledge Box among the ones you have access to in your account, and enter the search configuration you wish to apply.
If the Knowledge Box is located in the same region as the RAO workflow, an API Key will automatically be created, you will just be prompted to enter its expiration date. If not, you will have to provide the URL and key yourself.
Google
This driver allows calling Gemini hosted by Google through the Google API.
Without VertexAI
Go to https://aistudio.google.com/app/api-keys, create an API key and copy it.
Paste the key in the API key field and save the driver.
With VertexAI
See https://cloud.google.com/docs/authentication/provide-credentials-adc to create credentials.
SQL
This driver enables connecting to a SQL database.
You must provide the connection URI (e.g., postgresql://user:password@localhost:5432/mydb). Additionally, it requires the main database schema and a description providing context on the tables and relations so that the agent can properly interpret the information.
Snowflake
This driver enables connecting a Retrieval Agent to a Snowflake database.
You must provide the Snowflake account, user, warehouse, and database. You can optionally define a default schema and role. Authentication can be configured either with a password or with a PEM-encoded private key and optional passphrase.
The driver also lets you configure connection and execution settings such as the connection pool size, query timeout, query tag, and session parameters.
Perplexity
Go to https://www.perplexity.ai/account/api/group to create an API group in your Perplexity account, then create an API ket and copy it in the API key field and save the driver.
MCP server
This driver allows to use any MCP server. You need to provide the endpoint URL, and the needed authentication parameters (depending on the MCP server you are using).