> ## Documentation Index
> Fetch the complete documentation index at: https://docs.withnubo.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Build cache

> What Raypacks remembers between builds

Every Frame gets its own build cache that survives between deploys. It's how a rebuild of an app you tweaked one line of finishes in seconds instead of starting from scratch.

## What's cached

| Cache      | Path inside the build      | What it holds                                                    |
| ---------- | -------------------------- | ---------------------------------------------------------------- |
| Go modules | `/raypacks-cache/go-mod`   | `GOMODCACHE`: downloaded module sources                          |
| Go build   | `/raypacks-cache/go-build` | `GOCACHE`: compiled package objects                              |
| npm        | `/raypacks-cache/npm`      | `npm`'s download cache (covers pnpm and yarn install layers too) |
| OCI layers | `/raypacks-cache/layers`   | Per-layer digest map so unchanged layers skip gzip + upload      |

The cache is a single persistent volume per Frame, 5Gi by default, mounted into the build. You don't manage it; Nubo creates it on first deploy.

## Cache hits in the wild

A cold build of a simple Go service: 30 to 60 seconds.

A warm rebuild of the same service after a one-line change: usually under 15 seconds, often single-digit. Go modules don't re-download, packages don't re-compile from source, and unchanged image layers don't re-upload to the registry.

## When the cache misses

A cache miss is fine; the build still works. Common reasons it happens:

* **First build.** The cache is empty.
* **Lockfile changed.** Added a new dependency? Its sources need to be fetched once.
* **Manual reset.** Delete the Frame and start over and you get a fresh cache.

## Concurrent builds

Your plan sets how many builds can run at once. Push while you're already at the limit and the new build is turned away with a "concurrent build limit reached" message (HTTP 429); it isn't queued, so wait for a running build to finish and push again.

| Plan     | Builds at once |
| -------- | -------------- |
| Trial    | 2              |
| Basic    | 2              |
| Silver   | 4              |
| Platinum | 6              |

A build that's been running for more than 60 minutes is treated as stale and stops counting against your limit, so a stuck build can never block your next push forever.

## Custom caching needs

You can't configure the cache today (its 5Gi size and paths are fixed). If you have a use case that doesn't fit the defaults, let us know at [support@withnubo.com](mailto:support@withnubo.com).