add EIS rerank default inference endpoint

[brendan-jugan-elastic]

Overview

This PR adds the rerank default inference endpoint for the Elastic Inference Service. This changes makes a few assumptions:

1. The future model ID contained in the EIS authorizations response is rerank-v1
2. The default inference endpoint ID follows existing conventions and is .rerank-v1-elastic
3. The EIS task type -> inference API task type mapping is rerank/text/text-similarity -> TaskType.RERANK
   • This was pulled from this document outlining EIS task type mappings. This information might be outdated, so happy to modify the mapping if we want.

Testing

My testing included basic chat completions and sparse embeddings requests using the existing default endpoints with eis-gateway and eis-ray running locally, to ensure existing functionality works as expected. I've also modified the authorized models in my local eis-gateway, restarted ES, and verified that the list of default endpoints includes the new one for rerank.

Chat Completions:

curl -k -N --location 'http://localhost:9200/_inference/chat_completion/.rainbow-sprinkles-elastic/_stream' \
  --header 'Content-Type: application/json' \
  --header "Authorization: Basic ${ES_AUTH}" \
  --data '{
    "messages": [
        {
            "role": "user",
            "content": "In only two digits and nothing else, what is the meaning of life?"
        }
    ],
    "temperature": 0.7,
    "max_completion_tokens": 300
}'

Sparse Embeddings:

curl -k --location --request POST 'http://localhost:9200/_inference/sparse_embedding/.elser-v2-elastic' \
  --header 'Content-Type: application/json' \
  --header "Authorization: Basic ${ES_AUTH}" \
  --data '{
    "input": "A blue sky"
}'

Default Endpoints:

curl -k --location --request GET 'http://localhost:9200/_inference?pretty' \ 
  --header 'Content-Type: application/json' \
  --header "Authorization: Basic ${ES_AUTH}"
{
  "endpoints" : [
    {
      "inference_id" : ".elser-v2-elastic",
      "task_type" : "sparse_embedding",
      "service" : "elastic",
      "service_settings" : {
        "model_id" : "elser-v2",
        "rate_limit" : {
          "requests_per_minute" : 1000
        }
      }
    },
    {
      "inference_id" : ".rainbow-sprinkles-elastic",
      "task_type" : "chat_completion",
      "service" : "elastic",
      "service_settings" : {
        "model_id" : "rainbow-sprinkles",
        "rate_limit" : {
          "requests_per_minute" : 720
        }
      }
    },
    {
      "inference_id" : ".rerank-v1-elastic",
      "task_type" : "rerank",
      "service" : "elastic",
      "service_settings" : {
        "model_id" : "rerank-v1",
        "rate_limit" : {
          "requests_per_minute" : 500
        }
      }
    },
    other default inference endpoints....
  ]
}

[elasticsearchmachine]

Pinging @elastic/search-inference-team (Team:Search - Inference)

[elasticsearchmachine]

Pinging @elastic/search-eng (Team:SearchOrg)