Novua Control

AI operations command center for product and engineering teams.

Novua Control detects release bottlenecks, ownership gaps, and deployment risk across engineering workflows. It ingests signals from pull requests, deployments, and tickets, builds dependency context, scores operational risk deterministically, and surfaces explainable decision alerts with AI-assisted context.

Problem

Modern teams execute across fragmented systems: GitHub, deployment platforms, ticketing tools, and internal workflows. Important blockers are often distributed across these systems, which makes execution risk hard to see until delivery is already delayed.

Teams do not need another passive dashboard. They need a decision layer that makes operational bottlenecks visible before they become release failures, and that clearly explains what is blocked, why it escalated, and who should act next.

V1 Scope

This first version is intentionally narrow:

• GitHub pull requests
• Vercel deployments and rollout gates
• Linear-style tickets
• Dependency graph across engineering artifacts
• Deterministic risk scoring
• Explainable decision alerts with AI-readable summaries
• Audit trail for escalations, ownership, and state changes

AI is assistive in this product. It explains context and recommended action. Escalation logic remains explicit, deterministic, and auditable.

Core Capabilities

Event ingestion

Operational events enter from source systems and are normalized into one internal view:

• PR review waiting
• Deployment blocked
• Deployment failed
• Ticket blocked
• Rollout queued

Dependency graph

Artifacts are linked across systems so the engine can understand execution paths:

PR blocked -> deployment delayed -> ticket unresolved -> release risk

Deterministic risk scoring

Risk is computed through policy rules, not opaque model output. Example triggers include:

• Critical PR stale for more than 12 hours
• Production deploy blocked for more than 4 hours
• Customer-facing ticket unresolved
• Missing explicit owner
• Failed canary deployment
• Release train dependency chain blocked

Decision alerts

Alerts are not generic notifications. Each one answers:

• What is blocked
• Why it escalated
• Which artifacts are involved
• Who should act next
• What action is recommended now

AI-assisted operational briefs

Novua Control uses AI as a context layer, not as the source of truth:

• summarizes the blocked path
• explains the likely operational impact
• turns raw signals into a short decision brief
• keeps deterministic rules responsible for escalation and risk scoring

Audit trail

Every escalation keeps a trace of:

• triggering events
• ownership changes
• policy evaluations
• state changes

Architecture

Current implementation lives under src/lib/control:

• types.ts defines the domain model
• fixtures.ts contains the v1 seeded execution dataset
• engine.ts hydrates alerts, applies deterministic scoring, and prepares dashboard snapshots

The UI is intentionally a consumer of the engine, not the source of truth.

Product Positioning

Novua Control is not:

• another chatbot
• a generic AI assistant
• a passive engineering dashboard

Novua Control is:

• an AI operations command center for engineering execution
• an internal system for release bottleneck detection
• a full-stack product showing event-driven reasoning, ownership tracking, AI-assisted context, and explainable escalation

Local Development

npm install
npm run dev

Then open http://localhost:3000.

Live Connectors

The default demo keeps the seeded execution scenario so the product stays understandable on first load.

You can layer live source signals on top of that seed data by creating a local .env.local from .env.example.

GitHub

cp .env.example .env.local

Then set:

NOVUA_CONTROL_ENABLE_GITHUB_LIVE=1
GITHUB_TOKEN=your_github_token
NOVUA_CONTROL_GITHUB_REPO=owner/repository

GitHub live ingestion currently adds:

• open pull requests
• stale review detection
• missing owner detection
• requested reviewer signals
• failing checks
• merge-blocked state

Vercel

To enable live deployment ingestion:

NOVUA_CONTROL_ENABLE_VERCEL_LIVE=1
VERCEL_TOKEN=your_vercel_token
NOVUA_CONTROL_VERCEL_PROJECT_ID=your_project_id
NOVUA_CONTROL_VERCEL_TEAM_ID=optional_team_id

The public demo keeps the story focused on one incident. If you want to inspect how webhook payloads are normalized, use the ingestion preview instead of the primary landing.

Webhook Ingestion Preview

Novua Control also includes ingestion preview endpoints so the normalization layer can be exercised independently from the UI.

GitHub webhook preview

POST /api/ingest/github

Optional:

• x-github-event: pull_request
• x-github-event: pull_request_review
• x-github-event: check_run

The route returns the normalized:

• artifacts
• events
• signals

This keeps the repo legible as an operational system, not just a dashboard.

Vercel webhook preview

POST /api/ingest/vercel

Optional:

• x-vercel-event: deployment
• x-vercel-event: deployment.error
• x-vercel-event: deployment.ready

This route previews how deployment payloads become control-layer artifacts and execution events.

Durable Incident State

Novua Control can persist incident state, manual actions, and audit trail entries in Supabase Storage.

Set these environment variables to move beyond the local JSON fallback:

SUPABASE_URL=https://your-project.supabase.co
SUPABASE_SERVICE_ROLE_KEY=your_service_role_key
NOVUA_CONTROL_SUPABASE_BUCKET=novua-control
NOVUA_CONTROL_SUPABASE_OBJECT=control-store.json

With this configured:

• Assign backend owner
• Start mitigation
• Resolve incident

persist across reloads and deployments instead of living only in a local file.

Next Steps

• expand GitHub live ingestion into release-aware dependency scoring
• add Vercel deployment polling or webhook ingestion
• integrate Linear or Jira ticket events
• move incident persistence from object storage to Postgres when multi-workspace data modeling is needed
• add policy configuration and replayable incident simulations

License

This repository is source-available and proprietary.

Copyright (c) 2026 Ivete de Amorim. All rights reserved.

No permission is granted to use, copy, modify, redistribute, sell, or offer this software as a commercial service without prior written permission from the author. Commercial licensing is available on request via iveteamorim@gmail.com.