Skip to content

warheart1984-ctrl/Project-Infinity

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

63 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

If you want to understand how my mind works, check out my writing. I’ve published 26 books under two names — my own, Jon Halstead, and my pen name, Adam Wolfe. Jon Halstead is my worldbuilding and systems side. Adam Wolfe is the darker, human edge of those systems. I’m not only an architect — I’m a writer, too.

AAIS — Adaptive Assistant Intelligence System

Behavior enforced, not implied.

AAIS is a local-first, law-governed assistant runtime. Every request, every decision, every reroute is explicit, visible, and accountable.

Built for environments where behavior matters more than output.


Why AAIS Exists

Most assistant systems optimize for answers. AAIS optimizes for behavior.

  • One clear operating contract per turn
  • No silent reroutes or hidden fallbacks
  • Risky and experimental work isolated from normal work
  • Operator control preserved, not abstracted away
  • Every decision leaves a signed, time-bound trace

Core doctrine: Stabilize and Free. Stability before freedom. The system earns more responsibility by staying inside clear rules, explaining its behavior, and failing in a controlled way. If it cannot do that, it slows down, asks for confirmation, or stops.


Architecture

Every request moves through a fixed, law-bound path. Nothing bypasses the chain.

flowchart TD
    Client(["Client / Operator"])

    subgraph Ingress ["Ingress Layer"]
        Bridge["Bridge\n─────────────\nFail-closed by default\nClassifies & checks\nagainst project law\nSigns every decision"]
        Jarvis["Jarvis\n─────────────\nMain authority lane\nIngress + runtime control\nOperator-facing"]
    end

    subgraph Dispatch ["Dispatch Layer"]
        Forge["Forge\n─────────────\nIsolated contractor lane\nCode execution only\nBounded authority"]
        OTEM["OTEM\n─────────────\nTask + memory support\nBounded execution lane"]
        Workflows["Workflows\n─────────────\nPackaged app routes"]
    end

    subgraph Core ["Core Runtime"]
        Runtime["aais/ engine/\n─────────────\nTurn contracts\nInvariants\nRole boundaries\nTraceability"]
        Evolve["Evolve Engine\n─────────────\nLearns from outcomes\nCannot rewrite roles\nor law"]
    end

    subgraph Providers ["Storage & Providers"]
        Memory["Governed Memory"]
        LLMs["LLM Providers\n(Claude, etc.)"]
        APIs["External APIs"]
    end

    Client --> Bridge
    Bridge --> Jarvis
    Jarvis --> Forge
    Jarvis --> OTEM
    Jarvis --> Workflows
    Forge --> Runtime
    OTEM --> Runtime
    Workflows --> Runtime
    Runtime --> Evolve
    Evolve --> Memory
    Evolve --> LLMs
    Evolve --> APIs

    style Ingress fill:#1a1a2e,stroke:#4a4a8a,color:#e0e0ff
    style Dispatch fill:#0d1117,stroke:#3a5a3a,color:#d0ffd0
    style Core fill:#1a0d0d,stroke:#8a3a3a,color:#ffd0d0
    style Providers fill:#0d0d1a,stroke:#5a5a8a,color:#d0d0ff
Loading

Layer Roles

Layer Component Role
Ingress Bridge Front door and primary safety boundary. Fail-closed. Signs every decision.
Ingress Jarvis Main authority lane. Operator-facing ingress and runtime control.
Dispatch Forge Isolated contractor lane. Code execution only. Bounded authority.
Dispatch OTEM Bounded task and memory support.
Dispatch Workflows Packaged app route layer.
Core Runtime Enforces turn contracts, invariants, role boundaries, continuity, traceability.
Core Evolve Engine Learns from outcomes within bounds. Cannot alter role definitions or law.

Quick Start

pip install -e .
python -m aais start --data-dir ./.runtime/aais-data
Surface URL
App http://127.0.0.1:8000/app
Jarvis Console http://127.0.0.1:8000/app/jarvis
Health http://127.0.0.1:8000/health

Optional preflight:

python -m aais prepare --force-build --data-dir ./.runtime/aais-data
python -m aais doctor --data-dir ./.runtime/aais-data

Frontend dev server:

cd frontend
npm install && npm run dev

Surfaces: localhost:3000/jarvis · localhost:3000/workbench · localhost:3000/memory


Requirements

Use requirements.txt for standard local setup.

File Purpose
requirements.txt Standard local
requirements-local.txt Local dev extras
requirements-laptop.txt Constrained/laptop env
requirements-advanced.txt Full feature set
requirements-training.txt Training pipeline only

Optional: Claude Provider

export ANTHROPIC_API_KEY=your_key
export AAIS_CLAUDE_MODEL=claude-sonnet-4-20250514
export AAIS_ENABLE_CLAUDE_AUTO_ROUTING=true

Or pin Claude via provider_mode=claude_first in the Jarvis Console.


Repository Structure

aais/                  Core runtime
api/                   API surface
app/                   Packaged shell + workflows
src/                   Entry points (jarvis_operator.py, api.py)
docs/
  spine/               Canonical reading path
  runtime/             System references
  contracts/           Laws + doctrine
  subsystems/          Admitted subsystem packs
  audit/               Coverage + status
  _archive/            Lineage — not authoritative
  _future/             Planned — not live
engine/                Foundation layer
forge/                 Bounded contractor lane
evolve_engine/         Outcome-based adaptation
evals/                 Evaluation harness
tests/                 Full test suite
frontend/              Web app
mobile/                Expo mobile app
training/              Training pipeline

Only docs/ (excluding _archive/ and _future/) is authoritative for runtime understanding.


Documentation

Document Governs
AAIS Human Guide System overview
AAIS AI Operating Contract Runtime behavior contract
AAIS Master Spec Full specification
REPO_LAWBOOK.md Full repo operating law

Project Laws:

Document Governs
README Law v1 Documentation rules
External Suggestion Admission Rule How external input enters the system
ARIS Runtime Contract Embedded repo-intelligence law
Cognitive Bridge Runtime Law Ingress + attestation rules

Cognitive Architecture

Unified Architectural Hyper-Systemizer — Formal specification of the cognitive engine behind Project Infinity (May 5, 2026).


Points of Interest

AAIS contains deeper layers for those who explore:

  • Internal architecture layers — nested subsystems, lineage, early doctrine
  • Foundation artifacts — structural invariants, long-term stability markers in engine/
  • Historical documentsdocs/_archive/ shows the system's evolution
  • Introspective traces — certain components maintain narrative metadata as subsystems are added

These are optional and not required for running the system.


Security

See SECURITY.md for the disclosure policy.

License

Apache 2.0

About

Main Project

Resources

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors