Summary

Implements the TEXT_EMBEDDING function for ES|QL to generate dense vector embeddings from text using inference model.

Function Signature: TEXT_EMBEDDING(text: string, inference_id: string) -> dense_vector

Example Usage:

FROM documents 
| WHERE KNN(embedding_field, TEXT_EMBEDDING(content, "my-embedding-deployment"), 10)

Implementation Status

✅ Completed in this PR:

• TEXT_EMBEDDING_FUNCTION capability with snapshot build gating

• Core Function Infrastructure

  • TextEmbedding function class with proper type validation and serialization
  • InferenceFunction interface for inference-based functions
  • Function registration in EsqlFunctionRegistry

• Analysis of the inference function (validate existence and type of the inference endpoint)

  • Refactored pre-analysis, so it is able to collect inference ids form both Inference plans and inference function
  • Added validation for inference function in the analysis

• Add a pre-optimizer async phase to the ES|QL query execution

• Documentation generated from the annotations

• Execute the inference in the pre-optimizer

• Integration tests and end-to-end validation

🚧 TODO (Before Merge):

• Better CSV tests
• Integration tests

Notes

The function is enabled only in snapshot builds.
TEXT_EMBEDDING function is tracked into #131022