Sections

NOVA Documentation

An AI-native workspace that bridges the gap between high-level planning and autonomous execution. From idea to production — all inside a single terminal.

Currently running: openai/gpt-oss-120b · Switch with :model

1. Getting Started

NOVA is an AI-native workspace designed to bridge the gap between high-level planning and autonomous execution. It combines a conversational AI engine with real file system access, live code execution, and built-in Git integration — all from the Spark Terminal.

Plan & Validate

Describe your goal. The Prompt Enhancer refines it, and the Plan Validator ensures it's execution-ready.

Build

Run 'build it'. NOVA autonomously writes and implements the code using top-tier models.

Heal & Clean

Healer intercepts crashes and patches code. Janitor refactors for readability and performance.

Repo Mapping

NOVA parses your AST to understand project-wide context and dependencies automatically.

Installation & Setup

Install NOVA

The quickest way to get NOVA is via pip:

pip install nova-bridgeye

Initial Setup

Authentication

Launch the Spark Terminal

After authentication, open the Spark Terminal — this is NOVA's core interaction environment. Type :help at any time to view all available commands.

Project Initialization

For new projects, run the :wizard command to launch the Project Creation Wizard. It scaffolds your project structure and configures NOVA's context automatically.

Build — The Execution Engine

Build is the evolution of chat. It turns abstract prompts into production-ready code through a structured three-phase process.

Planning — Describe your goal

Start by telling NOVA what you want to build. If your prompt is sparse, the Prompt Enhancer will help flesh out the business logic and structure before generating the PLAN.md.

"Plan me a simple crypto price tracker using free APIs."

Verification — Review the plan

NOVA generates a detailed PLAN.md. The built-in Plan Validator ensures the architecture and implementation steps are complete before allowing the build to proceed.

Implementation — Run it

NOVA autonomously writes and implements all files sequentially using specialized coding models (like Kimi k2.6). If the connection drops, simply type :continue to resume the build queue.

build it

Auto-Healing — The Healer

Healer runs automatically detect and fix broken logic or runtime errors. Run your command as usual — NOVA handles the rest.

Command	Description
run <cmd>	Execute any command with Auto-Healing enabled
run app.py	Example: run a Python script
run server.js	Example: run a Node.js server

How it works: If execution fails, NOVA intercepts the stack trace, analyzes the error in context, and initializes the Healer to patch the code iteratively until the command runs successfully.

Healer supports Python, Node.js, Go, and any language that outputs a stack trace to stdout/stderr. Use run <cmd>.

Code Janitor — Refactoring

The Janitor tidies and refactors your code to maintain high health scores. It optimizes for readability, performance, and elimination of technical debt.

Load the file into context

:load <file_name>

Execute the cleanup

clean

Review the refactored output

NOVA refactors the code for readability, optimizes performance, and removes technical debt. Review the diff before saving.

Terminal Command Reference

Core Modes

Command	Action
run <cmd>	Run a command with Auto-Healing enabled (e.g., run python app.py)
clean	Run the Code Janitor to refactor and clean the loaded file
build it	Execute the verified Build plan and autonomously generate code
:continue	Resume an interrupted or dropped build queue

File & Navigation

Command	Action
ls / :map	Show the project AST blueprint and file tree
cd <path>	Change directory
pwd	Print current working directory
:load <file>	Load a file into AI context — required before clean or targeted debugging
:unload	Remove files from the current context
:paste	Input multiline text or paste external logs

System Operations

Command	Action
:gitoptions	Open the Git Operations Center (Init, Commit, Push, Create Repo)
:model	Change the active AI model
:overdrive	Toggle auto-confirm mode (bypasses [y/n] prompts for fast ops)
:makeroot / :exitroot	Lock or unlock NOVA file operations to the current directory boundary
reset	Wipe current session memory for a fresh start
exit	Shutdown the NOVA session

Git Operations Center

Access Git workflows directly from the terminal via :gitoptions.

First-Time Setup

On first run of :gitoptions, if no Git repository is detected in your project directory, NOVA will prompt you to initialize a Git repository before any other options are available.

spark terminal > :gitoptions
→ No git repo detected. Initialising…
✔ Git repository initialized successfully.
 
——— AUTOMATION SETTINGS ———
»○ Toggle Auto-Commit  [🔴 OFF]
   ○ Toggle Auto-Push   [🔴 OFF]
——— ACTIONS ———
   ○ 📝 Manual Commit (Stage & Commit all)
   ○ 🚀 Push to Origin (Creates GitHub Repo if none exists)
   ○ ⬇️ Pull from Origin
   ○ 📊 Git Status
——— EXIT ———
   ○ ← Back to Terminal

Options Explained

Option	Description
Toggle Auto-Commit	Automatically commits after each Build or Healer run
Toggle Auto-Push	Automatically pushes to origin after every commit
Manual Commit	Stages all changes and commits with an AI-generated commit message
Push to Origin	Pushes the current branch. Prompts to create a brand new GitHub Repo if no origin exists.
Pull from Origin	Pulls the latest changes from the remote repository (supports force sync)
Git Status	Displays a summary of staged, unstaged, and untracked files

Navigate the Git menu using arrow keys and press Enter to select an option.

AI Models

NOVA supports multiple underlying AI models. Switch at any time using :model in the Spark Terminal without losing your session context.

The current default model is openai/gpt-oss-120b, optimized for large-scale autonomous code generation tasks.

Subscription & Credit System

Access to NOVA's autonomous features is managed via your subscription tier. Credits are consumed based on the complexity and output of each operation.

Feature	Credit Usage
Build Actions	Consumed based on output tokens generated
Healer Runs	Charged per execution window / iteration
Janitor Runs	Charged per execution window
Navigation & Context	Free (ls, cd, :load, :unload, etc.)

If the Launch Gatekeeper is active, new registrations may be limited based on system load. View your live occupancy and credit limits on your Dashboard.