FAQ

Answers to the questions that come up most often. Click any question to expand it. If yours isn’t here, the in-app AI Assistant can usually diagnose project- and config-specific issues directly.

Setup & install

Do I need to know how to code?

No. Testver works without writing any code:

• Manual test management (cases, plans, defects, reports) — zero code.
• AI Test Gen / AI Record — generate code for you.
• The only place code helps is editing automation scripts directly, and you can hand that off to the AI Assistant.

How is Testver installed?

It’s published on npm as @testver/testver and installed globally:

npm install -g @testver/testver
testver serve

Then open http://localhost:3700. Full walkthrough: Install Testver.

Note

Always install with -g (global). A non-global npm install @testver/testver only drops it in the current folder and the testver command won’t work — you’d have to use npx @testver/testver serve.

Is Docker required?

No. Docker is optional and not required for any setup. The npm package runs the exact same dashboard for both local and team-server use.

What operating systems are supported?

Windows 10/11, macOS 12+, and any modern Linux. You need Node.js 20 or newer (which also installs npm).

How much does it cost?

Testver itself is free. Cloud LLM providers (OpenAI, Anthropic, etc.) charge for API calls — typically a few dollars a month for an active team. Local LLMs (Ollama, LM Studio) are free.

Do I need an LLM at all?

No. Testver works for manual test management without any LLM. AI features simply stay hidden until you add a provider in Settings or via TESTVER_DEFAULT_PROVIDER.

Testing

My tests don’t appear in Test Explorer.

1. Confirm the project root is correct (Header → project picker).
2. The framework’s config file should be at the root (playwright.config.ts, cypress.config.js, etc.).
3. Your test files must match the framework’s discovery pattern (Playwright tests/, Cypress cypress/e2e/, etc.).
4. Make sure the tests run on their own first (npx playwright test, pytest, …) — the language runtime and project dependencies must be installed on the same machine.
5. Click Re-scan in the header.

Tests run in my terminal but fail in Testver.

Usually the run command differs. Check Settings → Run Commands and make sure it matches what you’d run yourself.

Can I use Testver without a project?

Yes. All of Test Management (cases, plans, defects, AI generation) works with no project — you just skip Test Explorer / Runner / Output.

How do I parallelize runs?

Most frameworks parallelize internally and Testver respects their config. To run two separate suites at once, use + New parallel run in the Runner.

Why is a test marked “flaky”?

It produced different results on identical code. The Dashboard surfaces flaky tests so you can triage them; Testver’s self-healing also reduces flakiness from minor locator drift.

AI features

Which LLM is best?

For test-code generation, the latest Claude Sonnet, GPT-4o, or a strong local coder model all work well. Testver defaults to Claude Sonnet because it follows complex multi-step instructions reliably. For summarisation/analysis, any capable model is fine.

Can I run everything locally with Ollama?

Yes — local LLMs power AI Assistant, AI Test Gen, and AI Record. Quality is below cloud Sonnet/GPT-4, but for routine tasks they’re plenty. Larger models (e.g. llama3:70b) are more capable but need a beefier machine. Set OLLAMA_BASE_URL if it isn’t at the default.

Do my test cases get sent to OpenAI/Anthropic?

Only when you actively use an AI feature, and only what the task needs. If that’s a concern, run a local LLM — nothing leaves your machine.

Team / multi-user

How many users can one instance handle?

No hard limit. The shared SQLite store handles dozens of concurrent users on a modest server. Testver is a single-team, self-hosted tool — one instance per team — not a multi-tenant public service.

How do roles work?

Each user has one role: Admin (manages users/settings/audit), Tester (day-to-day testing + AI), or Viewer (read-only). Manage them on the Users screen.

Can different users have different LLM keys?

LLM keys are currently instance-wide. You can configure multiple providers and let users pick at conversation time, but they share the same keys.

Can I sync data across multiple instances?

Not automatically — there’s no built-in sync. Use Import / Export to move data between instances.

Hosting & operations

How do I host Testver for my team?

Run it on one always-on machine, bound to localhost, behind a reverse proxy for HTTPS:

testver serve --host 127.0.0.1 --port 3700

Keep it running with systemd or pm2, then front it with nginx/Caddy/Traefik. The proxy must forward WebSocket Upgrade headers or live updates won’t work. Full server setup is in Install Testver.

Caution

Testver executes your test commands on the host, so anyone who can sign in can run programs on that machine. Bind to 127.0.0.1 (not 0.0.0.0) and expose it only through a secured proxy/VPN.

How do I update Testver?

npm install -g @testver/testver@latest

Then restart the service (sudo systemctl restart testver or pm2 restart testver). Your data is untouched; migrations run automatically on first start.

How do I check it’s running?

curl http://localhost:3700/api/health

A healthy server returns small JSON with "status": "healthy" and the version.

How do I back up my data?

Copy Testver’s data folder — it holds everything: test cases, plans, runs, attachments, defects, and sessions. A nightly tar/copy job does the trick. For portability between instances, also use Import / Export.

Why does my disk fill up?

Output artifacts (screenshots, videos, traces) accumulate fast. Use the age filter in Output or a cleanup Schedule to prune old files periodically.

What do common error messages mean?

Error	Cause & fix
testver: command not found	Installed without -g, or a stale terminal. Run npm install -g @testver/testver and open a new terminal (fallback: npx @testver/testver serve).
EADDRINUSE / port 3700 in use	Another process holds the port. Start elsewhere: testver serve --port 8080.
EACCES on global install	Use sudo, or better, a Node version manager (nvm) so global installs don’t need admin.
EMFILE: too many open files (macOS)	A very large project hit the watch limit. Update to the latest Testver (it self-protects) or run ulimit -n 10240 first.
Node version error	Node is below 20. Reinstall the LTS from nodejs.org.
Dashboard freezes on a multi-hour run	Update to the latest Testver — long-run output is bounded and the live socket is kept alive in current versions.

Which environment variables can I set?

Variable	Default	Purpose
TESTVER_HOST	localhost	Network address to bind. Use 0.0.0.0 only behind a proxy/VPN.
TESTVER_PORT	3700	Port the dashboard listens on.
TESTVER_DEFAULT_PROVIDER	—	Default LLM: anthropic, openai, google, xai, ollama, or lmstudio.
TESTVER_DEFAULT_MODEL	—	Optional override of the model for the default provider.

Anything else?

If you’re stuck and the docs don’t help, ask the in-app AI Assistant — it has access to your project’s structure and can often diagnose configuration or test issues directly.