Skip to content

fstech-digital/operational-ontology-framework

Repository files navigation

Operational Ontology Framework

A public reference for operating AI systems with explicit state, governance and auditability.

Operational ontology is not a vocabulary list. It is the control layer that connects Data, Logic and Action so AI can operate inside real organizational constraints.

Publisher FSTech
Canonical page https://fstech.digital/ontologia-operacional/
Technical essay https://fstech.digital/framework/
Status Public reference, not production code
License See LICENSE

🇧🇷 Leia em português


What this repository contains

This repository publishes the public surface of the Operational Ontology Framework:

  • the core concepts;
  • the D+L+A model, Data, Logic and Action;
  • governance artifacts for stateful AI work;
  • anti-patterns observed in AI automation projects;
  • generic schemas and templates that define the interface of the method.

It intentionally does not publish FSTech production code, client implementations, prompts, agents, adapters, deployment scripts or commercial playbooks.

In short: this repository publishes the map, not the machine.


Core thesis

Most AI projects fail because they automate before the organization has a precise operational structure.

A model can generate text. It cannot infer, safely and consistently, all of the following unless they are made explicit:

  1. which entities exist;
  2. what each entity means;
  3. which rules govern each decision;
  4. which actions are allowed;
  5. what must be escalated to a human;
  6. what changed since the last session;
  7. what must be written back for the next operator.

The Operational Ontology Framework treats those constraints as first-class artifacts.


D+L+A

Every operational AI system needs three layers working together.

Layer Question Failure mode when absent
Data What exists and what does each thing mean? The model guesses entities and relationships.
Logic What rules, thresholds and routes govern decisions? The model improvises process.
Action What can be executed, by whom and under which guardrails? Automation creates uncontrolled side effects.

A chatbot is not an operational system until Data, Logic and Action are explicit and connected.


Public artifact model

The framework uses five governance artifacts.

Artifact Role Volatility
Pin Invariants: identity, domain entities, boundaries and non-negotiable rules Low
Spec Current work: tasks, acceptance criteria, blockers and execution state High
Handoff Session continuity: decisions, attempts, results and next actions Append-only
Facts Long-term observed knowledge with source and confidence Medium
Skills Reusable procedures refined through execution Medium

These artifacts are not magic. Their value is operational discipline: every AI action can be traced back to explicit context.


Anti-patterns

Common failure modes this framework is designed to avoid:

  • Prompt as architecture: hiding business rules inside a giant prompt.
  • Vector memory as governance: assuming retrieval equals accountability.
  • Automation before simplification: scaling a broken process.
  • Chatbot theater: conversational UI without operational authority or state.
  • No write-back: every session starts from scratch because the system learned nothing durable.
  • Unbounded agency: allowing an agent to act without clear escalation and action limits.

Repository structure

docs/       Public explanation of the framework
schemas/    Generic, non-client JSON schema examples
templates/  Minimal public governance artifact templates
LICENSE     Usage terms
NOTICE.md   Attribution and trademark notice
SECURITY.md Security and disclosure policy

No production implementation is included.


When to use the framework

Use it when an AI system must:

  • preserve state across sessions;
  • operate under domain-specific rules;
  • produce auditable decisions;
  • coordinate humans and agents;
  • execute actions with guardrails;
  • improve through write-back instead of prompt drift.

Do not use it for stateless text generation, one-off scripts or workflows where the cost of governance is higher than the risk of failure.


About FSTech

FSTech builds operational AI systems for companies that need more than isolated automations.

Site: https://fstech.digital
Canonical page: https://fstech.digital/ontologia-operacional/
Technical essay: https://fstech.digital/framework/

About

Operational Ontology Framework by FSTech — public reference for D+L+A, stateful AI governance and auditable operations. No production code.

Topics

Resources

License

Contributing

Security policy

Stars

11 stars

Watchers

0 watching

Forks

Packages

 
 
 

Contributors