PUBLIC ARCHITECTURE

The engine for AI systems that remember, adapt, and stay portable.

Seitiate is the open-source layer behind long-lived AI workspaces. It gives builders a way to combine memory, reusable capabilities, human review, and data portability without tying the whole system to a single model or vendor.

What this page is: a public map of the engine. It shows the responsibilities of the system and how the major pieces fit together, without exposing internal implementation details or deployment-specific internals.

The high-level flow

At a public level, Seitiate can be understood as a loop: bring in sources, build durable memory, run capabilities against real work, keep humans in charge, and preserve portability as the system grows.

Sources Memory Capabilities Human Review Reusable Knowledge └─────────────── feeds ───────────────┴──────── supports ────────┴──────── teaches ────────┘ Principles: continuity, governance, portability, and reusable composition

Five public building blocks

These are the major responsibilities Seitiate exposes to developers and contributors. They describe what the engine does, not the private internals of how each deployment chooses to implement it.

01. Memory That Persists

Seitiate is built around the idea that useful AI work should not reset every session. The engine accumulates structured memory from sources, prior work, and human decisions so future sessions can begin with continuity instead of amnesia.

  • Source-aware: work can start from imported material, project context, and prior decisions.
  • Durable: useful context survives past a single chat or prompt window.
  • User-centered: memory is organized around the creator or operator, not the platform.

02. Reusable Capabilities

Seitiate treats AI behavior as something composable. Builders can package workflows, prompts, tools, and policies into reusable capabilities that can be improved, shared, and assembled into larger systems.

  • Composable: small capabilities can be combined into larger flows.
  • Portable: useful work should move between projects and deployments.
  • Extensible: contributors can add new capabilities without rewriting the whole engine.

03. Human-Guided Adaptation

The engine learns from selection, correction, and approval rather than treating AI output as final. Human choice is part of the architecture, not a cleanup step added afterward.

  • Choice matters: accepted and rejected outputs become meaningful signals.
  • Quality compounds: the system improves by observing real preferences over time.
  • No blind autopilot: AI explores, humans decide.

04. Governance and Review

Seitiate is designed for systems where quality, safety, and review are first-class concerns. The engine expects explicit gates, checks, and standards around consequential work.

  • Review-friendly: important actions are meant to be inspectable.
  • Policy-aware: deployments can layer rules and controls around workflows.
  • Built for real use: governance is part of the system contract, not an afterthought.

05. Portability and Open Boundaries

Seitiate is the engine, not the whole product story. Managed products can add experience, hosting, and specialized interfaces on top of it, but the underlying engine is intended to remain open, portable, and usable across different models and environments.

  • Model-agnostic: the engine is designed to outlast any one model provider.
  • Deployment-flexible: self-hosted and managed experiences can share the same core ideas.
  • Exit-friendly: portability is a design goal, not a premium feature.

How Seitiate relates to products

One of the most important architectural distinctions in this repo is the line between the open engine, the managed product experience, and governance-heavy deployment layers.

ENGINE

Seitiate

The open-source composition and memory layer. This is where reusable capabilities, continuity, and contributor-facing architecture belong.

PRODUCT

Sakme

A managed product built on the engine. It turns the underlying capabilities into a curated end-user experience, with hosting, workflow design, and product UX on top.

GOVERNANCE LAYER

ClosedClaw

A security and enforcement-oriented deployment layer for higher-trust environments. It is not the public entry point for contributor architecture, but it informs how governance can be layered around the engine.

Want the code instead of the marketing story?

Start with the repository if you want implementation details, contribution paths, or the current state of the engine. This page is the public overview; the repo remains the authoritative source for contributors.