Built a free NVIDIA NIM Embeddings node for n8n — every vector DB node requires a native embedding sub-node and all existing options are paid

Built with n8n
,
1

While building a RAG pipeline I ran into a problem I couldn’t find a clean solution for.

Every vector database node in n8n — Pinecone, Qdrant, Supabase, Weaviate — only accepts a native Embedding sub-node. You can’t use an HTTP Request node as a workaround. So you’re locked into whatever paid embedding options n8n officially supports.

NVIDIA NIM offers powerful embedding models completely free, but there was no native n8n community node for them — so I built one.

GitHub: https://github.com/Sateeshreddymaddi/Custom-Nvidia-Nim-Node
npm: https://www.npmjs.com/package/@sateeshreddy/n8n-nodes-nvidia-nim

What’s included

One package, two nodes:

  • NVIDIA NIM — Chat Model sub-node (17+ free models)
  • NVIDIA NIM Embeddings — native Embedding sub-node, works with any vector DB node

Supported embedding models (all free)

  • NV-Embed-v1
  • NV-EmbedQA-E5-v5
  • Llama Nemotron Embed 1B v2
  • Llama Nemotron Embed VL 1B v2 (multimodal — text + image)
  • BGE-M3 (dense + sparse, multilingual)

Install

Settings → Community Nodes → Install → @sateeshreddy/n8n-nodes-nvidia-nim

Restart n8n and both nodes appear automatically.

Happy to answer questions or take feature requests!

2 curtidas
2

Hi @sateesh !

So cool to hear that ! Did you know that there were actually recent official nodes for this ? There is one for Nemotron Chat Model, and one for NVIDIA Nemotron Embeddings.

It got released with 2.16 version ! Do you see anything it fails to support ?

Cheers

3

Welcome @sateesh - nice work identifying the gap and shipping a node for it.

To add some context to Raphael’s question: the official n8n NVIDIA node added in v2.16 covers Nemotron Embed, but from what I see your node exposes BGE-M3 (dense + sparse, multilingual) and the VL multimodal variant - those aren’t in the official node yet. BGE-M3 in particular is useful for multilingual RAG pipelines where you need a single model that handles multiple languages without switching embedding endpoints. Worth keeping the community node around for that.

1 curtida
4

Hey @Raphael_Gomez ,good catch! I did see those 2.16 updates. However, those official nodes only support the Nemotron models themselves.

The cool thing about my custom node is that it opens the door to all the open-source models hosted on the NVIDIA free endpoint, rather than being restricted to just one family of models.

1 curtida