DOCSEARCHbyAlgolia
Submit Docs

Supervaizer Documentation

website·Indexed 31 days ago

https://docs.runwaize.com/

A Python toolkit and platform for building, managing, and operating AI agents with confidence. Includes the supervaizer-controller for agent registration and flow control, a REST API, and the Supervaize Fleet for mission, case, and context management. Features a skills cookbook for building agent capabilities and a web-based admin interface for managing WorkflowEntity objects like Jobs and Cases.

Tokens
42.3K
Snippets
175
Records
312
Agent score
50%

What's inside Supervaizer

  1. Overview of Supervaizer Integration Concepts

    The Supervaizer integration maps an existing Python agent's workflow to the Supervaizer model using three core concepts:

    • Job: A single end-to-end run of the agent.
    • Case: A repeated unit of work within a job (e.g., one email, one phone number, or one lead).
    • Step: A meaningful stage within a case, which may or may not involve human-in-the-loop (HITL) interaction.

  2. Overview of Supervaize Studio platform capabilities

    Supervaize Studio is a SaaS platform designed to manage AI agent operations. It allows users to discover and onboard AI agents, create missions with specific objectives, execute jobs using those agents with defined parameters, and monitor job performance and outcomes.

  3. Overview of the Supervaizer Controller

    The Supervaizer Controller is a Python-based runtime designed to register, describe, and expose AI agents to any A2A (Agent-to-Agent) compatible client. It serves as the primary tool for managing and operating AI agents within the Supervaize Studio platform.

  4. Overview of Supervaizer Controller

    Supervaizer is a Python toolkit designed for AI agent interoperability. It implements the Agent-to-Agent (A2A) protocol, allowing agents to be discoverable and interactable across different systems. It serves as the recommended controller for integrating AI agents into the Supervaize platform, providing APIs for agent registration, job control, event handling, and telemetry.

  5. Overview of Agent-to-Agent (A2A) Protocol support in SUPERVAIZER

    SUPERVAIZER implements the Agent-to-Agent (A2A) protocol to provide standardized agent discovery and interaction, ensuring interoperability between different AI agent systems. It incorporates features from multiple standards, including the former Agent Communication Protocol (ACP).

  6. Overview of Supervaizer Deployment CLI

    The supervaizer deploy command provides automated deployment of Supervaizer agents with zero-downtime updates, health verification, and automatic rollbacks. It supports three major cloud platforms: GCP Cloud Run, AWS App Runner, and DigitalOcean App Platform. The deployment workflow typically follows these stages: Local Testing -> Planning -> Deployment -> Monitoring -> Cleanup.

  7. Overview of Supervaize Access skill

    The Supervaize Access skill enables AI assistants (e.g., Cursor, Claude CLI, Codex) to manage a Supervaize workspace. It allows users to list and edit agents, create and manage missions, and handle case/step reporting and human-in-the-loop (HITL) workflows. Permissions are based on the user's team membership and role; read-only users cannot perform write operations.

  8. Understand the Supervaize architectural mental model

    Supervaize acts as the system of record for AI-driven work, owning business-level entities, lifecycles, and visibility. It delegates the actual execution of workflows to an execution engine, such as Supervaizer or n8n.

    Key Workflow Flow:

    1. Registration: The Supervaizer Controller registers the agent and server with Supervaize. Users then configure secrets and environment variables in the Supervaize UI.
    2. Mission Setup: Objectives and success criteria are defined, and Agents are assigned to a Mission.
    3. Job Execution: A user starts a Job in Supervaize. Supervaize creates the job context and triggers the Agent. The Agent then creates Cases and Steps. Human-in-the-loop steps pause execution until a user responds via the Supervaize UI.

  9. Add context to a Supervaize mission

    To provide AI agents with the business knowledge, rules, and constraints needed for a specific mission, follow these steps:

    1. Navigate to Missions $\rightarrow$ [Your Mission Name] to access the mission overview.
    2. Use the context configuration panel to add optional context, including Business Rules, Knowledge Base links, and Operational Constraints.

  10. Configure Slack notification types and alerts

    Supervaize Studio provides several categories of notifications that can be configured to keep teams informed about AI operations:

    • Job Status: Notifications for when a job starts (includes details and agent), completes (includes metrics and quality scores), or fails (includes error details and troubleshooting steps).
    • Mission Updates: Tracking mission status changes (started, in progress, completed), team assignments, priority/deadline updates, and milestone achievements.
    • Performance Alerts: Monitoring for response time degradation, quality score drops, resource utilization warnings, and accuracy rate fluctuations.
    • System Notifications: Platform availability, maintenance updates, security updates, and agent online/offline status.

  11. Define basic performance metrics for AI agents

    Set up fundamental performance indicators to measure AI agent success. Each metric requires a Target (the goal), a Weight (relative importance), and a Threshold (minimum acceptable level).

    • Response Time: Measure speed of response (e.g., Target: < 5 minutes, Weight: 0.3, Threshold: 10 minutes).
    • Accuracy: Measure correctness of output (e.g., Target: > 90%, Weight: 0.4, Threshold: 80%).
    • Customer Satisfaction: Measure user sentiment (e.g., Target: > 4.5/5, Weight: 0.3, Threshold: 4.0/5).

  12. Manage AI agents with Supervaize Studio

    Supervaize Studio is a management platform that allows non-technical teams to monitor, control, and support mission-critical AI agents in real time, acting as a co-pilot for AI-powered operations.