[ML] Add Groq inference service

[bamchip]

Summary

• Add Groq as a chat completion inference provider, wiring up service registration, action creators, settings validation, and REST surface.
• Extend tests and docs to cover the Groq-specific behaviors and configuration knobs.

Testing

Create Groq chat completion endpoint

{
  "service": "groq",
  "service_settings": {
    "model_id": "llama-3.3-70b-versatile",
    "api_key": "<GROQ_API_KEY>"
  }
}

Stream chat completion

{
  "messages": [
    { "role": "user", "content": "Explain Groq in one sentence." }
  ]
}

Stream chat completion with tool calling

{
  "messages": [
    {
      "role": "user",
      "content": [
        { "type": "text", "text": "What is the current price of item 123?" }
      ]
    }
  ],
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_current_price",
        "description": "Look up the latest price for a catalog item.",
        "parameters": {
          "type": "object",
          "properties": {
            "item": { "type": "string", "description": "Catalog identifier (e.g. SKU)" },
            "currency": {
              "type": "string",
              "description": "ISO currency code",
              "enum": ["USD", "EUR", "GBP"]
            }
          },
          "required": ["item"]
        }
      }
    }
  ],
  "tool_choice": {
    "type": "function",
    "function": { "name": "get_current_price" }
  }
}

[github-actions]

🔍 Preview links for changed docs

• docs/reference/elasticsearch/rest-apis/index.md

[github-actions]

ℹ️ Important: Docs version tagging

👋 Thanks for updating the docs! Just a friendly reminder that our docs are now cumulative. This means all 9.x versions are documented on the same page and published off of the main branch, instead of creating separate pages for each minor version.

We use applies_to tags to mark version-specific features and changes.

Expand for a quick overview

When to use applies_to tags:

✅ At the page level to indicate which products/deployments the content applies to (mandatory)
✅ When features change state (e.g. preview, ga) in a specific version
✅ When availability differs across deployments and environments

What NOT to do:

❌ Don't remove or replace information that applies to an older version
❌ Don't add new information that applies to a specific version without an applies_to tag
❌ Don't forget that applies_to tags can be used at the page, section, and inline level

🤔 Need help?

• Check out the cumulative docs guidelines
• Reach out in the #docs Slack channel

[elasticsearchmachine]

Pinging @elastic/ml-core (Team:ML)

[davidkyle]

@elasticmachine test this please

[davidkyle]

@elasticmachine test this please

[davidkyle]

Looks great I left a couple of minor comments.

CI failed due to the use of a forbidden reflection API call in a test I pushed a change to remove that. You can run the precommit target locally to check for these sorts of problems with

./gradlew :x-pack:plugin:inference:spotlessApply :x-pack:plugin:inference:precommit

To run the tests its

./gradlew :x-pack:plugin:inference:test

[davidkyle]

Related Kibana issue for the icon elastic/kibana#243696

[bamchip]

@elasticmachine test this please

[bamchip]

@davidkyle I see a conflict in server/src/main/resources/transport/upper_bounds/9.3.csv but not sure if I should overwrite it. Will let you guide me there :)

[davidkyle]

@elasticmachine test this please

[davidkyle]

LGTM

[davidkyle]

I see a conflict in server/src/main/resources/transport/upper_bounds/9.3.csv

I resolved it by accepting the incoming change then ran ./gradlew generateTransportVersion again. However, since this morning it has happened again and it will keep on happening.

The PR is approved, I'll merge it once the UI change elastic/kibana#243696 is underway so that the icon is not missing when groq appears in Elastic serverless.

[davidkyle]

@elasticmachine test this please

[davidkyle]

@elasticmachine run elasticsearch-ci/bwc-snapshots-part3

[davidkyle]

@elasticmachine test this please

[davidkyle]

@elasticmachine test this please