Agent Configuration

Datalinc's agentic layer (Eve) allows you to deploy AI agents that can securely manage your data, automate workflows, and answer questions. These agents are not just conversational chatbots; they are equipped with functional tools that let them query databases, read files, and send emails on your behalf.

Choosing a Model (BYOK)

Datalinc operates on a Bring-Your-Own-Key (BYOK) model for Large Language Models (LLMs). This ensures that you have complete control over your LLM billing and data privacy agreements.

1. Navigate to AI Agent > Configuration.
2. Select your preferred provider (e.g., openai, anthropic).
3. Enter the specific model name (e.g., gpt-4o, claude-3-5-sonnet-20240620).
4. Provide your secure API Key.
5. Click Save Configuration.

Equipping Tools

Agents operate based on the tools you provision to them. By default, agents have access to basic conversational abilities, but you can explicitly enable integrations.

Available Tools

• Virtual Database Querying: Allows the agent to run SELECT statements against your provisioned virtual databases.
• File & Web Explorer: Permits the agent to read HTML contents from public websites or extract text from your Vault documents.
• SMTP Email (Human-in-the-Loop): Allows the agent to draft and send emails using your SMTP credentials.

Human-in-the-Loop (HITL) Approvals

Certain sensitive tools (like dropping database tables or sending outbound emails) require explicit human approval.

When an agent hits a gated tool, execution pauses and you will receive a prompt in the chat interface to approve or reject the action.

Setting the Default Configuration

If you have multiple configurations, you must set one as the Default to ensure the chat interface knows which model to spin up when a user types a message. Click the Make Default star icon next to your preferred configuration.