Add support for alpine image

[PylotLight]

Currently getting errors around @anush008/tokenizers-linux-x64-musl.
I imagine this is because of using nod:alpine image?

024-12-07 04:32:53 [ERROR]: ❌ [server]: Error during initDatabase with file /usr/src/packages/server/node_modules/flowise-components/dist/nodes/embeddings/FastEmbed/FastEmbedEmbedding.js: Cannot find module '@anush008/tokenizers-linux-x64-musl'
2024-12-07 15:32:53.063Z 1:00 Require stack:
2024-12-07 15:32:53.063Z 1:00 - /usr/src/node_modules/.pnpm/@Anush008[email protected]/node_modules/@anush008/tokenizers/index.js
2024-12-07 15:32:53.063Z 1:00 - /usr/src/node_modules/.pnpm/[email protected]/node_modules/fastembed/lib/cjs/fastembed.js
2024-12-07 15:32:53.063Z 1:00 - /usr/src/node_modules/.pnpm/[email protected]/node_modules/fastembed/lib/cjs/index.js
2024-12-07 15:32:53.063Z 1:00 - /usr/src/packages/components/dist/nodes/embeddings/FastEmbed/FastEmbedEmbedding.js

[Anush008]

Can you try Linux x64 GNU?

[PylotLight]

I just used bun add fastembed to add to the project, how do I use that version sorry?  As far as I can tell I'm already using linux x64, I even tried adding the other GNU utils: coreutils findutils gawk This is the dockerfile I'm using + extra utils https://github.com/FlowiseAI/Flowise/blob/main/Dockerfile FastEmbed (or its dependency @anush008/tokenizers) currently only provides prebuilt binaries compatible with glibc, which is the standard C library for most Linux distributions like Debian. However, Flowise's base image is built on Alpine, which uses the musl C library. I’ve attempted to add all available gcompat packages to the Dockerfile to enable glibc compatibility on Alpine, but the errors persist, indicating that the required prebuilt binaries (@anush008/tokenizers-linux-x64-musl) are unavailable. Given that changing the entire Docker base image from Alpine to Debian or Ubuntu just to support this single library is not a viable solution, I’ll need to wait for an alternative approach—such as native support for musl or a compatible version of FastEmbed. I’m open to suggestions if anyone has ideas for resolving this without altering the base image Next steps: Can the tokenizers library can be built from source for musl/alpine?

[PylotLight]

Looks like lower level support is required for dependancy package tokenizers which fastembed uses.
Created issue here for follow up: huggingface/tokenizers#1703

[PylotLight]

given the issue is likely solely with dependency, we can probably close this one until tokenizers supports musl on alpine.