Documentation

Complete Documentation

Guides, API references, and tutorials for every component of the SE-AI architecture.

Getting Started

Installation

Install SE-AI CLI and dependencies

Quick Start

Run your first Mind in 5 minutes

Configuration

Configure hardware, models, and privacy

Doctor Check

Verify system health and dependencies

Core Concepts

Model vs Mind

The fundamental distinction

Cognitive Compiler

Goal → Task Graph optimization

Evolution Loop

12-gate governed evolution

Intelligence Efficiency

Metric: Outcome / Resources

Privacy Gates

Default deny, local-first

Minds

Mind Specification

Complete Mind configuration schema

Mind Lifecycle

Create, run, evolve, export, import

PAIOS Reference

First reference implementation

Custom Minds

Build your own Mind

Memory

Memory Types

10 memory types explained

Lifecycle

Capture → Consolidate → Archive

Consolidation

Compression, decay, forgetting

Retrieval

Query, ranking, embeddings

Skills

Skill Anatomy

Inputs, outputs, procedure, evaluator

Built-in Skills

Reasoning, coding, extraction, planning

Skill Composition

Chaining and parallel execution

Custom Skills

Create and register skills

Validation

Test cases and benchmarks

Tools

Tool Capabilities

Web, file, shell, code, HTTP, database

Toolchains

Research, coding, data analysis

Sandboxing

Isolation, permissions, limits

Custom Tools

Models & Routing

Model Registry

Runtime Adapters

Ollama, MLX, llama.cpp, Local

Routing Engine

Policy-driven selection

Provider Registry

Lifecycle: discovered → active

Hardware Awareness

Real-time capacity detection

Evolution

Evolution Layers

10 layers from prompts to architecture

Candidate Generation

Generators per layer

Sandbox & Benchmark

Isolation, resource limits, MindBench

Reviews & Promotion

Security, privacy, cost, human approval

Rollback & Monitoring

Instant rollback, canary deployment

Genome

Genome Schema

Versioning, branching, lineage

Diff & Compare

Visual genome comparison

Rollback & Reproduce

Instant rollback, reproduction

Branching

Experiment branches

MindBench

Benchmark Suite

Standard task suite

Running Experiments

Create, run, compare

Metrics Reference

20 metrics explained

Reproducibility

Seeds, configs, hardware fingerprints

Security & Privacy

Threat Model

Injection, exfiltration, escalation, poisoning

Capability Permissions

Least privilege enforcement

Privacy Gates

Classification, redaction, audit

Audit Logging

Decision records, provenance

SDK & CLI

TypeScript SDK

High-level client API

CLI Reference

All seai commands

Plugin Development

Extend SE-AI

Start Building Now

Install the CLI, run the doctor, and create your first Mind in minutes.

Get StartedView Darwin SpecView Architecture

One Command

seai init → seai run

Local First

Runs on your hardware

Evolves

Improves through experience