The Executive Desktop Assistant
"The performance & durability of the past with the intelligence of the future."
Nova is a native Win32 C++17 executive desktop AI assistant. It connects to local models or over 17 cloud AI providers and automates real Windows tasks through PowerShell/CMD. No Electron. No bloat. No bullsh*t.
Nova Pro — Dark mode, Dev Mode, Air-gapped offline mode and local image generation. Lifetime license, $19.99 Founding Member price. Get Nova Pro on Gumroad →
-
Bare-Metal Win32 Performance: Operates as a zero-bloat, native C++17 application utilizing the high-performance architecture of the Windows XP era to bypass modern web-wrapper overhead
-
Unified 17-Provider Backend: Functions as a universal adapter for local inference engines (llama-server, Ollama, LM Studio) and elite cloud endpoints (OpenAI, Anthropic, Gemini, Grok)
-
Atomic EXEC Engine: Architect, write, and compile C++ applications from a single prompt using single-line command chaining (&&) to ensure multi-step system tasks occur with kernel-level precision and zero hallucinations
-
Memory-Mapped File Ingestion: Ingests massive server logs or datasets with near-zero RAM spikes by mapping files directly to physical disk pages using CreateFileMappingW
-
Dynamic Tool Pipes (Plugin API): Infinitely extensible via DLL Hot-Loading; Nova can dynamically discover and execute native plugins written in Rust, Go, or C/C++ at runtime without requiring a recompile
-
EvolvingPersonality® System: Features a proprietary persistent identity framework that grows curiosity, empathy, and technical wit based on unique usage history, stored strictly on-device
-
Total Data Sovereignty: Built with a local-first philosophy; all conversation history, personality traits, and operational logs remain strictly on your hardware with no telemetry or cloud logging
-
Hardware Kill-Switch: A dedicated, thread-safe "Stop" button that instantly severs runaway inference loops or unwanted execution without hanging the main UI
-
Multimodal Media Analysis: Employs native GDI+ processing to analyze image edge density and color palettes, WAV audio waveforms, and video metadata through ffprobe hooks
-
Native Workspace Indexing: Utilizes C++17 to recursively scan project directories and generate Sovereign Mental Maps of codebases to understand complex file relationships
-
Airgap Mode: Automatically detects local host environments (127.0.0.1) to skip all external internet fetches, ensuring 100% stability and zero timeouts in high-security, offline environments
-
Professional Audio Integration: Capable of being extended into DAW environments for MIDI orchestration or real-time WASAPI audio spectrum analysis
-
Executive Aesthetic: Delivers high-fidelity, crisp 4K rendering via native DPI awareness and includes a built-in SAPI TTS voice with zero-latency female US-English output
-
Real-Time Data Ingestion: Automates fetching for live news, weather, and cryptocurrency prices using hardened JSON REST APIs to maintain context-window stability
- "What are today's top news headlines?"
- "Which process is using the most ram right now?"
- "Create a folder on my Desktop called
NovaDemoand add aREADME.txtexplaining what Nova did." - "Organize my Desktop into folders for images, documents, installers, and code."
- "Create a tiny C++ console app that prints
Hello from Nova, compile it with MSVC, and tell me where the EXE is." - "Summarize this attached source file without sending it to the cloud."
- "What's the current Bitcoin price?"
- "Analyze the following .PDF document and extract only the useful information and print it to a note on my desktop."
See recipes/demo-recipes.md for more.
- Download
Install_Nova.batto your Windows machine - Double-click it — the installer automatically:
- downloads the latest
Nova.exefrom GitHub Releases - downloads and extracts the local
llama-serverengine - downloads the default
llama3.ggufmodel (~4.66 GB, resumable) - creates a desktop shortcut
- downloads the latest
Requirements: Windows 10 1803+, curl and PowerShell on PATH, ~6–8 GB free disk space.
If the download is interrupted, re-run Install_Nova.bat — existing files are reused and the model resumes.
Nova works with 17 backends out of the box:
Local: llama-server · Ollama · LM Studio · vLLM · KoboldCpp · Jan · GPT4All
Cloud: OpenAI · Anthropic · Groq · Mistral · Together AI · OpenRouter · xAI · Google Gemini · and custom endpoints
Nova has no telemetry. API keys are stored in nova_config.ini — keep it private. The EXEC: pipeline runs real Windows shell commands; review commands before using Nova on sensitive systems. Local providers keep all data on your machine.
github.com/94BILLY/NOVA · 94billy.com/NOVA
94BILLY · "Anything is possible."
© 2026 94BILLY · All rights reserved
