Using Ollama / local embedding models with the Servoy AI plugin

Questions
1

Hi all,

we are currently working on a semantic search feature in Servoy using PostgreSQL with pgvector.Right now we generate the embeddings ourselves by calling a local Ollama instance via plugins.http.
The model we use is bge-m3, which gives us 1024-dimensional embeddings.
After generating the embedding, we store the vector in PostgreSQL using plugins.rawSQL.This works fine so far.Now we are looking at the Servoy AI plugin and wondering if there is a more native way to do this.

In the documentation I found the OpenAI embedding model builder, for example:

plugins.ai.createOpenAiEmbeddingModelBuilder()

I also saw that there is a Servoy embedding store builder, but I am not fully sure if this can only be used together with the embedding models provided by the plugin.

So my questions are:

  1. Does the Servoy AI plugin currently support Ollama or other local embedding models?

  2. Can createServoyEmbeddingStoreBuilder() be used with embeddings generated by a non-OpenAI model?

  3. Is there any way to provide a custom embedding model/provider to the Servoy AI plugin?

  4. If this is not supported at the moment, would the recommended approach be to continue calling Ollama manually via plugins.http and handle the pgvector storage/search ourselves?

Thanks in advance.

2

Thanks for posting on the forum and for using the AI Runtime plugin.

The plugin is model/vendor-agnostic. It should be pretty easy to add support for Ollama to the plugin.
Please file a case at support.servoy.com and we’ll take a look.

Best,
Sean