nova-ai-pro 1.2.7

Personal AI, On Personal Devices.

Complete Setup & User Guide

Documentation

Roadmap

Prerequisites

NOVA AI is local-first: the only hard requirement is Python. Everything else is optional, depending on how you want to run models.

Hardware: any 64-bit Windows 10/11, macOS 12+, or modern Linux. 4 GB RAM minimum, 16 GB+ recommended. A CUDA/Metal/ROCm GPU or Apple Silicon is optional but makes 7–8B models much faster. Small models (0.5–3B) run fine on CPU.

Quick Start (from source)

The fastest way to a running system from a fresh clone:

git clone https://github.com/Hamza35779/NOVA-AI.git
cd NOVA-AI
./start.sh          # Linux / macOS — interactive launcher (Windows: start.bat)

start.sh / start.bat check Python, then give you a menu: chat, voice, web server, memory wiki, diagnostics, and (on Windows) the opencode wizard. If you prefer to set things up manually, see Install from source below.

One-command everything (installs uv/Ollama if missing, pulls a starter model, builds the UI, starts backend + frontend, opens the browser): ./scripts/quickstart.sh

Install from source

1. Get the code and Python deps

git clone https://github.com/Hamza35779/NOVA-AI.git
cd NOVA-AI

# With uv (recommended):
uv sync --extra server --extra tools-search

# Or with plain pip (virtualenv recommended):
python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -e ".[server,tools-search]"

Windows note on inference-gguf: the in-process GGUF engine extra (--extra inference-gguf) compiles llama-cpp-python from source and needs Visual Studio Build Tools (C++ workload) + CMake. Without them uv sync / pip install fails inside its CMake step. Skip that extra (use Ollama or the cloud engines), or install the build tools first — prebuilt Windows wheels are not published.

Extra groups (install only what you need): dev (pytest/ruff/mypy/maturin), desktop (server + speech), server (FastAPI backend), inference-gguf (in-process GGUF, no Ollama needed — see the Windows note above), inference-vllm, inference-cloud, inference-litellm, memory-faiss / memory-pdf / memory-bm25, voice, screen, channel-telegram, and more — see [project.optional-dependencies] in pyproject.toml.

2. Build the native Rust extension (recommended)

The storage/memory layer is faster with the compiled nova_ai_rust extension and some server APIs require it:

# Needs Rust 1.90+ (rustup installs it automatically from rust/rust-toolchain.toml)
uv run maturin develop --manifest-path rust/crates/nova_ai-python/Cargo.toml

3. Build the web UI

The built UI is not committed — nova serve serves it from src/nova_ai/server/static/, which is generated by the frontend build:

cd frontend
npm install
npm run build
cd ..

4. Verify and run

uv run nova doctor      # hardware, GPU, and engine health check
uv run nova serve       # web UI + API at http://localhost:8000

Then open http://localhost:8000. The in-app GGUF Hub can download a starter model with one click — no Ollama required.

5. Connect a model engine (pick one)

Running nova init writes ~/.nova_ai/config.toml with hardware-detected defaults (or start from a template in configs/nova_ai/config.toml).

Uninstalling

Source installs live entirely inside the cloned folder plus ~/.nova_ai/:

# Remove the checkout (deactivates the CLI / uv environments)
cd .. && rm -rf NOVA-AI

# Remove user data: config, databases, downloaded GGUF models
rm -rf ~/.nova_ai        # Windows (PowerShell): Remove-Item -Recurse ~\.nova_ai

# Optional extras you may have installed along the way:
#   uv:        rm -rf ~/.local/share/uv ~/.cargo/bin/uv
#   Ollama:    use the platform uninstaller, or `brew uninstall ollama`
#   Rust:      rustup self uninstall

The curl/one-line installer variant ships an uninstaller: nova-uninstall (sees scripts/install/nova-uninstall.sh). There is no pip uninstall nova-ai step needed for source installs; for the PyPI package use pip uninstall nova-ai-pro.

Why NOVA AI?

Personal AI agents are exploding in popularity, but nearly all of them still route intelligence through cloud APIs. Your "personal" AI continues to depend on someone else's server. Local language models already handle most single-turn chat and reasoning queries, and their capability per watt keeps improving year over year. What has been missing is the software stack to make local-first personal AI practical.

NOVA AI is that stack. It is a framework for local-first personal AI, built around three core ideas: shared primitives for building on-device agents; evaluations that treat energy, FLOPs, latency, and dollar cost as first-class constraints alongside accuracy; and a learning loop that improves models using local trace data. The goal is simple: make it possible to build personal AI agents that run locally by default, calling the cloud only when truly necessary. NOVA AI aims to be both a research platform and a production foundation for local AI, in the spirit of PyTorch.

Installation & Quick Start

Pick your platform and start in seconds:

Start Working with NOVA AI

# Interactive Chat
nova chat

# Hands-free Voice Conversation
nova voice --push-to-talk

# Smart Model Router (Auto-selects optimal model)
nova router status
nova router test "Analyze quantum algorithms"

# Active Memory Wiki
nova memory-wiki show profile
nova memory-wiki search "preferences"

# Interactive Canvas (Open HTML/SVG/Chart visualizations)
nova canvas list

# Build/Test Diagnostics with Self-Healing Fix Suggestions
nova dev-watch -c "pytest -q"

# Presets & System Diagnostics
nova init --preset morning-digest-minimal
nova doctor

Prefix nova ... with uv run, or source .venv/bin/activate first.

Example:

nova init --preset morning-digest-mac
nova connect gdrive          # one OAuth covers Gmail / Calendar / Tasks
nova digest --fresh          # generate and play your first briefing

Per-preset deep dives: morning digest · deep research · code assistant · scheduled monitor · chat simple · or the full quickstart guide.

Skills

Skills teach agents how to better use tools and improve their reasoning. Every skill is a tool — agents discover them from a catalog and invoke them on demand.

# Install skills from public sources
nova skill install hermes:arxiv
nova skill sync hermes --category research

# Use skills with any agent
nova ask "Use the code-explainer skill to explain this Python code: for i in range(5): print(i*2)"

# Optimize skills from your trace history
nova optimize skills --policy dspy

# Benchmark the impact
nova bench skills --max-samples 5 --seeds 42

Import from Hermes Agent (~150 skills), OpenClaw (~13,700 community skills), or any GitHub repo. Skills follow the agentskills.io open standard.

See the Skills User Guide and Skills Tutorial for details.

Built-in Agents

NOVA AI ships with built-in agents across three execution modes (on-demand, scheduled, continuous):

opencode Integration

Use opencode with NOVA AI in both directions — opencode gets your local Ollama models (as the nova-ai provider) and NOVA's tool registry via MCP, so coding-agent sessions run local-first:

nova opencode install   # fetch the opencode CLI when missing
nova opencode init      # write/merge ./opencode.json (provider + MCP tools)
nova serve              # start the backend (another terminal)
nova opencode launch    # open opencode wired to NOVA AI

nova opencode status shows per-provider data residency (local vs cloud); --local-only locks opencode to local models and disables session sharing. Docs: opencode Integration.

Desktop Extras

• ⚡ Quick Capture (Windows: Alt+Space, macOS: Cmd+Shift+Space) — a Raycast-style chat popup from anywhere; the conversation syncs back to the main app automatically.
• 🔔 Morning digest notifications — a desktop notification fires the moment your briefing is stored (covers scheduler, CLI, and server delivery); nova digest plays audio on Windows via winsound/the default media player.
• 🔨 Dev-Watch — run a build/test command, get failure classification (timeout / exit code / embedded errors) plus self-healing fix suggestions; results feed the Build Diagnostics panel on the Dashboard (/api/devwatch/runs).

Complete CLI Commands Reference

NOVA AI features a comprehensive CLI suite (nova or python -m nova_ai.cli or standalone nova-ai.exe):

1. 💬 Chat, Voice & Perception

2. 🚀 Server & Background Daemons

3. 🧠 Models & Smart Routing

4. 🛠️ Autonomous Tools & Integrations

5. 📚 Long-Term Memory & Deep Research

6. ⚙️ Diagnostics, Config & Benchmarking

See the User Guide and Tutorials for detailed setup instructions.

Full documentation — including Docker deployment, cloud engines, development setup, and tutorials — at hamza35779.github.io/NOVA-AI.

Community

• GitHub: github.com/Hamza35779/NOVA-AI
• Issues: github.com/Hamza35779/NOVA-AI/issues

Contributing

We welcome contributions! See the Contributing Guide for incentives, contribution types, and the PR process.

Quick start for contributors:

git clone https://github.com/Hamza35779/NOVA-AI.git
cd NOVA-AI
uv sync --extra dev --extra framework-comparison --extra server
uv run maturin develop --manifest-path rust/crates/nova_ai-python/Cargo.toml   # native extension (CI builds it too)
cd frontend && npm ci && npm run build && cd ..                                # embeds the UI into nova serve
uv run pre-commit install
uv run pytest tests/ -v

Useful scripts: scripts/gen-cli-docs.py regenerates docs/cli/ from nova --help, and scripts/check-versions.sh catches stale installer/version strings before CI does.

Browse the Roadmap for areas where help is needed. Comment "take" on any issue to get auto-assigned.

License

Apache 2.0