Skip to main content

Frequently asked questions

When do I need Bowrain, and when is kapi enough?​

kapi holds the context graph for one project: it reads and writes many content formats, runs checks, and draws on local memory, all from files you own, with no server or account. Bowrain is the platform that holds voice profiles, vocabulary, and content memory once, across every project, surface, and teammate, and adds real-time collaboration, connectors, automation, versioned history, and corrections that become enforced checks.

kapi on its own is enough when one person works from one checkout. Use Bowrain when:

  • content lives in systems beyond that checkout;
  • several projects or surfaces should share one memory;
  • several people work on the same content;
  • you need shared review decisions.

The full comparison is in How Bowrain and kapi fit together.

Can I self-host Bowrain?​

Yes. Bowrain is available under the AGPL-3.0 license, or under a commercial license, and runs as a set of Docker services (server, worker, PostgreSQL, a job queue, and the web UI) behind your own OIDC provider. A self-hosted deployment runs without the billing pipeline: no credit limits, no plan gates. See Self-hosting.

Do bring-your-own-key runs consume credits?​

No. When a workspace configures its own AI provider key, drafting and check operations run against that provider account, and those runs are not metered against Bowrain credits. (Usage is still counted internally, for abuse protection.) Credits are only spent when an operation uses the shared platform provider. See Security and privacy.

What happens to my data if I leave?​

You can export content to its original format and use the kapi CLI to export shared stores in open interchange formats: kapi memory export writes TMX, and kapi terms export writes TBX, CSV, or JSON. A self-hosted deployment additionally keeps everything in a PostgreSQL database you can back up directly.

Which formats are supported?​

Bowrain reads and writes the same formats the neokapi engine supports: document formats, data formats, subtitle formats, office formats, and bilingual interchange formats. The generated format reference lists the current set.

Does Bowrain train models on my content?​

No. Bowrain has no model-training pipeline. Your content is sent to an AI provider only to carry out an operation you initiated, a draft or a quality check, and, with a bring-your-own key, that request goes to your own provider account. The platform's analytics measure product usage, not content (see Security and privacy).

How do review statuses work?​

Review status is tracked per block and per locale, so reviewing the French target of a block never changes its German target. A block moves through four states, Not Started, Draft, Translated, and Reviewed, and you can filter a file by state to work through one stage at a time. In the review session, approving a block marks it Reviewed; rejecting it sends it back to Draft. See Review.

Can I use my own machine-translation provider?​

Bowrain drafts with AI (large language model) providers, not dedicated machine-translation engines. You can bring your own key for any provider the engine's translate tool supports, including a locally hosted model, and drafting runs against it. Dedicated MT providers are not part of the product.

I have a question that is not here​

Contact hello@bowrain.cloud.