Apex

A Next.js application with Prisma and a Docker-based scanner.

Prerequisites

• Node.js 20+
• Docker (for the scanner)
• npm, yarn, pnpm, or bun

Quick start

Follow these steps in order to run the app locally.

1. Install dependencies

npm install

2. Set up the database (Prisma)

Generate the Prisma client (required before running the app):

npx prisma generate

Optional: if you need a fresh database or schema changes, run migrations:

npx prisma migrate dev

Ensure a .env file exists in the project root. At minimum:

DATABASE_URL="file:./dev.db"

For paying for scans (Wallet tab): Apex needs a wallet to receive XRP. Set XRPL_WALLET_SEED (or APEX_XRPL_SEED) to that wallet’s secret in .env or .env.local. Optionally set XRPL_WALLET_ADDRESS for reference. Restart the dev server after changing.

3. Build the scanner Docker image

The app uses a scanner service that runs in Docker. From the project root run:

PowerShell (Windows):

docker build -f docker/Dockerfile.scanner -t apex-scanner:latest . --progress=plain

Or: .\docker\build-scanner.ps1

Bash (Linux/macOS):

docker build -f docker/Dockerfile.scanner -t apex-scanner:latest . --progress=plain

The image is based on the official Playwright image (Chromium pre-installed), so the first build is usually 2–5 minutes. If the build fails, ensure Docker Desktop is running and you have enough disk space.

4. Run the development server

npm run dev

Open http://localhost:3000 in your browser.

---

Scanning repos

Scans run in the apex-scanner Docker container. Ensure Docker Desktop is running before you start a scan. The app connects to Docker automatically (Windows: named pipe; Linux/macOS: Unix socket). To use a custom Docker endpoint, set DOCKER_SOCKET in your environment.

If you see connect ENOENT /var/run/docker.sock on Windows (or when running from WSL), add to .env.local:

DOCKER_SOCKET=//./pipe/docker_engine

Then restart the dev server.

---

Summary checklist

Step	Command
Install deps	npm install
Prisma client	npx prisma generate
Scanner image	docker build -f docker/Dockerfile.scanner -t apex-scanner:latest .
Start app	npm run dev

AI fixer tuning (optional)

If you want longer-running or more incremental AI fixing, add these to .env:

OPENCODE_MODEL="opencode/glm-5-free"
OPENCODE_TIMEOUT_SECONDS=90
OPENCODE_TOTAL_TIMEOUT_SECONDS=240
OPENCODE_PROMPT_BATCH_SIZE=1
OPENCODE_PROMPT_MAX_BATCHES=50
OPENCODE_WORKERS=1
OPENCODE_CONTRAST_BATCH_SIZE=1
OPENCODE_CONTRAST_THINKING=true
OPENCODE_ALL_THINKING=false
# optional: minimal | high | max (provider-specific)
OPENCODE_THINKING_VARIANT=

• OPENCODE_PROMPT_BATCH_SIZE=1 processes one violation at a time.
• OPENCODE_CONTRAST_BATCH_SIZE=1 splits contrast violations into smaller batches instead of one giant prompt.
• Fixes are upserted to the database after each batch, so they appear during fixing.
• OPENCODE_TOTAL_TIMEOUT_SECONDS caps total fixer runtime for MVP responsiveness.
• Contrast violations are grouped first and can run in thinking mode with OPENCODE_CONTRAST_THINKING=true.
• OPENCODE_WORKERS splits prompt batches across isolated subagent workers (max 6).
• OPENCODE_PARALLEL_GROUPS is accepted as a legacy alias for OPENCODE_WORKERS.
• Set OPENCODE_ALL_THINKING=true to run every fix batch in thinking mode.
• Increase timeouts and max batches only when you need deeper remediation.

Learn more

• Next.js Documentation
• Prisma Documentation
• Next.js deployment (e.g. Vercel)