AI Agents

The AIO Agents (aio_agents) framework provides an enterprise-grade management, orchestration, and execution system for deploying intelligent, autonomous AI agents in OpenEduCat.

How to Create an Agent and Activate

To create and activate a new AI agent in OpenEduCat:

  1. Navigate to AI Agents: Go to AIO ‣ Agents ‣ Agents and click the New button.

  2. Configure Basic Settings: * Name: Enter a clear descriptive title (e.g., Student Academic Advisor or Grading Assistant). * Model: Select an active LLM model to power the agent (e.g., GPT-4, Claude 3.5 Sonnet). * System Prompt / Instructions: Define the agent’s role, behavioral instructions, and output constraints.

  3. Attach Agent Components: * Link function Tools for executing operations. * Attach vector Knowledge Collections (RAG) for institutional document search. * Set Guardrails for content safety and PII masking. * Configure Approval Rules for manager verification on critical actions.

  4. Activate the Agent: * Click the Activate button in the header of the agent form view. * Draft: Initial unverified configuration state (cannot respond to automated event triggers). * Active: Operational state where the agent actively processes user prompts, event hooks, and automated workflows. * Inactive: Paused state where automated triggers skip processing.

Model Provider Configuration
AI Model Configuration
AI Agent Configuration

Tools

Function calling tools equip AI agents with the ability to retrieve database records, send notifications, run Python methods, and perform external data searches within OpenEduCat.

  • Tool Setup: Access tools under AIO ‣ Configuration ‣ Tools.

  • Predefined Tools: Pre-configured tools like odoo_record_retriever allow agents to fetch records from any Odoo model with custom filtering capabilities.

  • Basic Information & Configuration: Set Tool Code, Predefined Tool toggle, Version, Endpoint, API Keys, and Provider details.

  • Tool Capabilities: Enable agents to perform record management (Retrieve, Create, Update, Delete), execute model methods, trigger email notifications, and perform web research.

AIO Tools Configuration

Guardrails

Guardrails enforce content safety, data privacy, and behavioral policies on AI agents to ensure safe operation within educational environments.

  • Guardrails Setup: Access safety settings under AIO ‣ Guardrails ‣ Guardrails.

  • Safety Rules: Define rule categories including Personally Identifiable Information (PII) masking, toxicity filtering, prompt injection defense, and output moderation.

  • Agent Attachment: Attach guardrail policies directly to AI agents so all incoming prompts and outgoing AI responses are screened automatically before execution.

Guardrail Safety Configuration

Database Config

Database Configurations enable AI agents to connect directly with external PostgreSQL databases (including PgVector) for vector storage, RAG retrieval, and cross-database queries.

  • Configuration Access: Navigate to AIO ‣ Configuration ‣ Database Configurations.

  • Connection Parameters: Specify the Configuration Name (e.g., psql), Database Type (PostgreSQL (PgVector)), Connection URI (e.g., postgresql://username:password@localhost:8000/ai_db1), and API Key.

  • Test Connection: Click Test Connection in the form view header to run an instant connection test. Successful verification displays a status notification confirming active host and database connectivity.

Database Configuration and Connection Test

JSON Schemas

JSON Schemas provide structured data validation and input/output definitions for agent tools, survey forms, and feedback evaluation in OpenEduCat.

  • Structured Input & Output: Ensure tool calls and agent responses follow strict JSON schemas for predictable data processing.

  • Survey & Feedback Integration: Map complex response structures (e.g., course feedback surveys, question types, skipped items, and user answers) to enable automated AI evaluation and report generation.

JSON Schemas and Survey Feedback Mapping