Skip to content
View VeyrNova's full-sized avatar
  • AURA
  • France

Block or report VeyrNova

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
VeyrNova/README.md
Veyr — Creator of AURA

AURA Local First GitHub YouTube


About

I'm Veyr, an independent creator building AURA — a local-first AI desktop environment focused on natural interaction, contextual memory, intelligent tool routing and modular AI orchestration.

AURA is not designed as another chat window. The goal is a desktop assistant that can understand intent, keep useful context, select the right intelligence or tool, and act through one coherent interface.


AURA

🧠 Intelligence
Local + cloud model orchestration
🎙 Voice
Natural speech input and output
💾 Memory
Persistent contextual recall
🛠 Tools
Files, web, system and automation

One interface. Multiple intelligences. One assistant.


Current AURA checkpoint

23 September 2026 — AURA is currently in a stabilization and performance phase.

  • Sanitized validation branch: validation/aura-current-pre-performance-safe-20260923
  • Validated SHA: 61a4d4d5302846942921d05e457ab2a08590e365
  • Active Windows Chromium surface: v0.7.2.2-rc4.2
  • Immediate work: dead/redundant-code audit → UI patch consolidation → startup/performance optimization
  • Public-repository policy: no local databases, browser profiles, secrets, logs, private media or temporary runtime state

The next technical milestone is to reduce startup cost without removing features or changing validated behavior.

Latest validated milestone — R21

On 23 September 2026, AURA validated lazy loading for the global cartography dataset. The change removes 7.32 MB from the initial blocking JavaScript path while preserving the Weather map and loading global cartography only when needed.

Current focus: startup optimization, dead-code verification and careful consolidation of accumulated UI patch layers.

R23 runtime result

R23 was tested on 23 September 2026 and rejected after a UI regression affecting side panels and the Memory, Tasks, Agenda and Modules surfaces. The local rollback restored normal behavior.

R21 remains the current validated performance checkpoint.

R25 CSS audit

R25 found no fully redundant stylesheet. The main consolidation candidate is aura-dev-ui-screen-fit-r3-fix2.css, where 85.56% of declarations are reproduced exactly later in the cascade. A declaration-level audit is required before any cleanup.

R27 runtime result

R27 was rejected after a runtime panel regression affecting Memory, Tasks, Agenda and Modules. The exact pre-R27 CSS was restored and normal behavior returned. R21 remains the validated performance baseline; optimization now shifts to runtime lifecycle profiling rather than static cascade trimming.

Latest validated runtime optimization — R30

R30 FIX1 is validated. AURA now keeps the legacy P042 orb as a fallback but avoids its full render workload while that canvas is hidden/collapsed. The visible P0436 orb and the rest of the UI remain unchanged.

This marks the shift from risky static removal toward runtime lifecycle optimization.

Current Focus

Area Status Direction
AI orchestration ● Active Model and provider routing
Voice interaction ● Active Speech input, output and live routing
Contextual memory ● Active Relevant long-term project context
Adaptive interface ● Active Task-specific UI and reactive states
Desktop automation ◐ Building Reliable real-world actions
Developer Fabric ◐ Building Multi-agent development orchestration
Public Alpha ○ Planned Controlled external testing

Architecture

AURA architecture


Tech

Python JavaScript HTML5 CSS3 Git GitHub Windows

AI orchestration • Local LLMs • Voice • Memory • Tool calling • Desktop apps • Automation • Developer agents

AURA Developer Fabric

The Developer Fabric is AURA's development orchestration layer. Its purpose is to maintain a shared project context while routing development work toward the most suitable coding model, agent or tool.

                         AURA
                           │
                    Developer Fabric
                           │
               ┌───────────┼───────────┐
               │           │           │
          Coding Agent  Coding Agent  Coding Agent
               │           │           │
               └───────────┼───────────┘
                           │
                    Shared Context
                           │
                    Project + Tools

Principles

Local first

Privacy-sensitive and latency-sensitive workloads should be able to run locally whenever practical.

Provider independent

Models and providers should remain replaceable rather than becoming hard dependencies.

Context matters

The assistant should retain useful project context without treating every interaction as isolated.

Tools over text

When a task needs an action, AURA should route to a capability instead of only generating an answer.


Roadmap

Foundation → Intelligence → Memory → Voice → Tools → Automation → Developer Fabric → Alpha

Current stage: active private development and reliability work.


Public Development

The public side of AURA will focus on architecture, demonstrations, screenshots, technical notes and development milestones. Private runtime components, credentials and sensitive configuration remain outside public repositories.


Follow the build

GitHub YouTube



VEYR × AURA

Think. Route. Remember. Act.

Popular repositories Loading

  1. AURA AURA Public

    AURA Native — local-first Windows AI assistant built with Qt/QML and PySide

    Python 1

  2. VeyrNova VeyrNova Public

    Creator of AURA — building a local-first AI desktop assistant with voice, memory, tools and an adaptive UI.