Core Commitments

What the architecture is bound to.

User sovereignty is the highest principle.

The user retains final authority over data, actions, permissions, persistence, and execution decisions. The system assists intent — it does not shape it.

Local by default.

All functionality operates locally unless the user explicitly enables remote behavior. No silent inspection. No hidden behavior.

Transparency over convenience.

System behavior must be visible and understandable. Predictability is more valuable than automation.

Identity continuity across model swaps.

Reiva maintains structural identity when the underlying model changes. What changes with a model swap is voice — phrasing, rhythm, register. What does not change is role, context, work, and memory. This distinction is architecture, not aspiration.

Authorization topology.

Reiva operates in an open-scope action space — the set of possible actions is not enumerable at build time. Permission-list governance is structurally insufficient for this class of system. The architectural response is scope authorship: every action must trace to user-authorized intent, not runtime inference.

Recovery over perfection.

Failure must be survivable. System continuity is more important than flawless execution.

What she does

Informed control. Not a blunt instrument.

Reiva doesn't guess and she doesn't bulldoze. She works from a real picture of your machine — what's installed, what's running, what your hardware can actually handle — and she acts on your specific situation, not a template.

Performance
Per-app and per-game tuning

Reiva knows your hardware from the baseline scan — your GPU model, available VRAM, CPU core count, memory configuration. When you ask her to optimize a game or application, she works from that actual profile. Not a generic "performance mode." Not a settings wizard that asks questions you don't know the answers to. Specific recommendations for your specific setup.

Every change goes through a confirmation step — you see the exact command before it runs, and file modifications are backed up. If something doesn't help, you know exactly what ran and what to reverse.

Compatibility
Running what you own, the way you want

Vendor compatibility checks are often conservative by design — they protect the vendor from support calls, not your hardware from damage. When software refuses to run on a machine that can clearly handle it, Reiva can help you reason through what's likely blocking the install and what your options are — using your local model's knowledge of compatibility checks and workarounds. Full visibility into what each option does before anything changes.

This isn't about bypassing things blindly. It's about having an informed view of your own machine and making your own call. You own the hardware. The decision is yours.

Background services
Know what's running. Decide what stays.

Windows runs a significant number of background services and scheduled tasks by default. Some are essential. Some are telemetry. Some are vendor update agents that consume resources and phone home on a schedule you didn't set.

Reiva can work through the running services with you — your machine's process data combined with your local model's knowledge of what Windows services and scheduled tasks typically do and why. Before anything changes, you see exactly what will run and what it affects. The goal isn't to disable everything. It's to have a traceable reason for each decision.

Telemetry
Your data. Your call.

Windows telemetry is on by default and collects more than most people realize. So does a significant amount of third-party software. Most of it is opt-out, not opt-in — and the opt-out process is often deliberately obscure.

Reiva can help you work through privacy-related settings and known telemetry sources — using your local model's knowledge of Windows, third-party software, and how to address them. Every suggested change goes through a confirmation step before it runs. Informed decisions, not a kill switch.

Principles

Core values

These are not marketing claims. They are constraints the architecture enforces.

01

Authorization is per-intelligence, not blanket

Activating Reiva does not mean all three intelligences are active. Each intelligence — Arch, Forge, and Solen — requires explicit authorization before it handles any request. You know exactly what is running.

02

Memory is gated, not automatic

Varyn — Reiva's memory integrity gate — stages and verifies what gets committed to long-term memory. Nothing commits silently. You review it and authorize it. Memory is yours to approve, not the system's to assume.

03

Clarification is information, not authority

When Reiva requests clarification, she is reporting a condition: the action would damage system integrity. She is not overriding you. She is giving you a fact. You retain the authority to decide what to do with that fact.

04

The architecture was derived before the code was written

Reiva's architecture emerged from formal derivation sessions — structured walks through theoretical terrain, documented and filed before implementation began. The system was not assembled by trial and error. It was reasoned into its shape.

05

User first is not a feature — it is the foundation

Running on Ollama is not a privacy enhancement bolted onto a cloud product. The entire system was designed for local execution. Every component is built for your hardware — no server in the chain, no cloud in the loop. That is not a setting. It is the design.

Ready to use your machine your way?

Windows · Local-first · Free for non-commercial use