Skip to content
Uvilo OS
archived Visibility internal Owner erik@uvilo.com Approver _ Created _ Updated _

Uvilo OS Requirements

Uvilo OS is a multi-purpose workflow and document creation, management, and publishing system for Uvilo.

Purpose

  • Provide a single source of truth for
    • Our company vision/mission/values
    • Our processes and automations
    • Our documentation
    • Strategic planning
  • Allow team members to easily access and modify company information.
  • Provide public-facing documentation and help to customers and partners.
  • Provide canonical information to AI agents to perform tasks and help customers.
  • Automate the design, running, and evolution of our company.

Requirements

Human-friendly workflows

  • Easily editable by team members
  • Usable by non-tech people
  • Easily generate multiple websites using different subsets of the content:
    • Marketing site
    • Internal-facing dashboard, documentation, and collaboration
    • External-facing support and documentation
  • Fully indexed, quickly searchable, and queryable by people and chatbots
  • Provides version control so mistakes can be easily reverted (maybe via git source control)

Agentic workflows

  • Easily editable by LLMs
  • Uses a standard AI-friendly file format like markdown
  • Includes a shared LLM-agnostic work environment based on LibreChat, where team members can work with multiple skilled agents to build, manage, and evolve Uvilo (the company), and Uvilo (the product).
  • Supports automated agentic execution: AI agents can autonomously implement large projects across multiple sessions with minimal human-in-the-loop interaction
  • Project-based execution: each architecture project (e.g., Taxonomy, Onboarding Quiz) is implemented by a sequence of sessions, each picking up where the last left off
  • Cross-session continuity: machine-readable state files, session logs, and working logs allow any agent to resume work from any prior session
  • Resilient to failures: if a session crashes, times out, or is interrupted, the state on disk reflects actual progress and the next session can recover gracefully
  • Interchangeable LLMs: the system supports swapping between different LLM providers (Anthropic, OpenAI, etc.) via LibreChat, paying by token rather than monthly subscription

Technical requirements

  • Preferably based on an open-source platform or custom vibe-coded, so not expensive to maintain as we grow
  • If there is code to maintain, it should be typescript-based