[aws-amplify] Can't use native node:sqlite implementation on Amplify

[antalasztrik]

Is your feature request related to a problem? Please describe

When trying to statically generate a lot of pages (350+) we ran into an issue, where some pages wouldn't get pre-rendered. In the build logs we found some lines that indicated that there is a problem with our SQLite db:

• 2025-10-30T10:09:46.951Z [WARNING]: [error] Failed to execute SQL INSERT INTO _content_info VALUES ('checksum_XYZ', false, 'Pa7CrgJJaD4QDj2tA5Aa1fBsFeNWQoZj6Qg3fp83yrc', 'v3.5.0--7-zYv-DKz3BK19q0UotPW_pKZn8L7iqWiEW9NESIvoU', 'LrPDg5JwY_7Sv0VaNBgS0amTMfOq9glrFxZh5Ge2HCQ'); -- meta: SQLITE_BUSY: database is locked
• 2025-10-30T10:09:47.514Z [WARNING]: [error] Failed to execute SQL CREATE TABLE IF NOT EXISTS _content_XYZ (id TEXT PRIMARY KEY, "extension" VARCHAR, "meta" TEXT, "stem" VARCHAR, "__hash__" TEXT UNIQUE); -- structure: SQLITE_BUSY: database is locked
• and then a bunch of [WARNING]: [error] [unhandledRejection] SQLITE_BUSY: database is locked when trying to pre-render some of the pages.

This only happens with prod data, where some of the JSON files are over 1.7MB in size, with much smaller test data we get no errors, so most likely the problem is not in our code. We were also able to reproduce these errors in a local environment, so it's not just Amplify.

Describe the solution you'd like

Right now there is no way to use the native node:sqlite implementation in an Amplify environment, because of this line: https://github.com/nuxt/content/blob/main/src/presets/aws-amplify.ts#L20.

We haven't found any documentation on why sqlite3 is a hard requirement for Amplify and why the native node implementation wouldn't work. Since Amplify supports Node version 22, would it make sense to re-visit if this dependency is really needed for those environments?

Describe alternatives you've considered

We manually throttled our pre-render speed in the Nitro config, which seemed to help. We still get some SQLITE_BUSY errors, but we assumed it works on one of the re-tries Nitro does when the initial pre-render fails. So with this option all of our pages are properly generated, but manually throttling the rendering doesn't seem like the best solution.

We also tried the native node:sqlite implementation with the experimental.sqliteConnector option in our local environment and the errors were completely gone. So it looks like the issue comes from the sqlite3 dependency.

Additional context

If sqlite3 is not installed, we have this error when building the project:
2025-10-31T13:11:24.376Z [WARNING]: [error] [@nuxt/content] Nuxt Content requires sqlite3module to work in AWS Amplify environment. Please runnpm install sqlite3 to install it and try again.

[farnabaz]

This is expected, sqlite3 is peer dependency, sadly there is not single sqlite package which could work on all providers. AWS works with sqlite3, Cloudflare and Vercle works with better-sqlite3 and modern node environments can use node:sqlite.
Therefore adapters are peer dependency in Nuxt Content, and based on your deployment environment you need to install them manually.

Checkout https://content.nuxt.com/docs/deploy/aws-amplify

[antalasztrik]

Thank you for the reply @farnabaz.

I understand that sqlite3 is a dependency on Amplify environments for Nuxt content, we had it installed, that's where the issues came from, that's why we have been looking for an alternative.

My question still stands, does it have to be a peer dependency? Because my logic is that since Amplify environments support Node 22, we should be able to use the native Node implementation instead of sqlite3. Or are there any other requirements for it to work, that Amplify is missing?

Also can you link me to a documentation detailing why we need this specific module? I couldn't find anything in their docs and Google wasn't helpful either.

[farnabaz]

Also can you link me to a documentation detailing why we need this specific module? I couldn't find anything in their docs and Google wasn't helpful either.

There isn’t any official documentation on this. Based on our tests and user feedback, other adapters don’t work reliably on Amplify (likely because those adapters depend on native OS libraries).

Regarding native Node SQLite support, I haven’t tested it fully yet, which is why the modules currently enforce sqlite3.

I’d appreciate your help testing native SQLite. You can open a PR to comment out the lines that enforce sqlite3, then use a commit release to deploy your application. It may take me a while to test on Amplify, so since you already have a project set up, your testing could help us improve the module if applicable. 🙏

[antalasztrik]

I commented out the import from the preset file and can confirm that it works. sqlite3 removed from our project, experimental.sqliteConnector set to "native" and using Node v22.21.1 on Amplify. No more SQLITE_BUSY or other SQL errors during build and all pages work as expected.

[farnabaz]

I commented out the import from the preset file and can confirm that it works. sqlite3 removed from our project, experimental.sqliteConnector set to "native" and using Node v22.21.1 on Amplify. No more SQLITE_BUSY or other SQL errors during build and all pages work as expected.

Thanks for fast check. I appreciate it

Could you upgrade the Amplify docs in your PR, and describe how to change Node version? (If 22 is node default version)

[antalasztrik]

Unfortunately, version 22 is not the default Node version on Amplify, so my existing PR would break builds for all deployments that are not using the native connector option and are not on Node v22.5+. There are different ways to set the Node version in Amplify, so I'd let the developers decide on how they do it - but we could give them documentation pages that detail on how to do it.

I can try to update the PR when I have some free time, but I might need some pointers. My thoughts are for the preset:

• Check if the user has the experimental.sqliteConnector option set to "native".
  • If it is set to "native", it should use the Node implementation. Does the native path handle Node version checks already or would it be something that also needs to be done in the preset?
  • What happens if the user sets it to better-sqlite3? Should we give them the option to install it in an Amplify environment or nudge them to using sqlite3 instead?
  • If it is unset or set to sqlite3, use the current flow that imports sqlite3.

I think the default option should still be the one using sqlite3, because you need additional work for the native Node connector to actually work on Amplify. However I think if it is documented properly, developers could consider opting in to this feature.