The ToolHive Operator currently requires the CRD to either use the Hugging Face Text Embeddings Inference image (ghcr.io/huggingface/text-embeddings-inference) or configure optimizer.embeddingService with an OpenAI-compatible embeddings endpoint.
For deployments that only require keyword search and do not use semantic search, this dependency on an embeddings service seems unnecessary.
Could you please provide a way to make this optional in the CRD, so that optimizer.embeddingService only needs to be configured when semantic search is explicitly enabled? This would allow keyword-only deployments without requiring an embeddings service.
The ToolHive Operator currently requires the CRD to either use the Hugging Face Text Embeddings Inference image (ghcr.io/huggingface/text-embeddings-inference) or configure optimizer.embeddingService with an OpenAI-compatible embeddings endpoint.
For deployments that only require keyword search and do not use semantic search, this dependency on an embeddings service seems unnecessary.
Could you please provide a way to make this optional in the CRD, so that optimizer.embeddingService only needs to be configured when semantic search is explicitly enabled? This would allow keyword-only deployments without requiring an embeddings service.