Introduce the InferenceChatModel for langchain

[pgayvallet]

Summary

Part of #206710

This PR introduces the InferenceChatModel class, which is a langchain chatModel utilizing the inference APIs (chatComplete) under the hood.

Creating instances of InferenceChatModel can either be done by manually importing the class from the new @kbn/inference-langchain package, or by using the new createChatModel API exposes from the inference plugin's start contract.

The main upside of using this chatModel is that the unification and normalization layers are already being taken care of by the inference plugin, making sure that the underlying models are being used with the exact same capabilities. More details on the upsides and reasoning in the associated issue.

Usage

Usage is very straightforward

const chatModel = await inferenceStart.getChatModel({
  request,
  connectorId: myInferenceConnectorId,
  chatModelOptions: {
    temperature: 0.2,
  },
});

// just use it as another langchain chatModel, e.g.
const response = await chatModel.stream('What is Kibana?');
for await (const chunk of response) {
     // do something with the chunk
}

Important

This PR is only adding the implementation, and not wiring it anywhere or using it in any existing code. This is meant to be done in a later stage. Merging that implementation first will allow to have distinct PRs for the integration with search (playground) and security (assistant + other workflows), with proper testing

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[pgayvallet]

/ci

[elasticmachine]

💚 Build Succeeded

• Buildkite Build
• Commit: 9acb9f7

Metrics [docs]

Module Count

Fewer modules leads to a faster build time

id	before	after	diff
inference	28	34	+6
observabilityAIAssistantApp	508	514	+6
observabilityAiAssistantManagement	385	391	+6
searchAssistant	265	271	+6
searchPlayground	285	291	+6
total			+30

Public APIs missing comments

Total count of every public API that lacks a comment. Target amount is 0. Run node scripts/build_api_docs --plugin [yourplugin] --stats comments for more detailed information.

id	before	after	diff
@kbn/inference-common	54	59	+5
@kbn/inference-langchain	-	47	+47
inference	28	29	+1
total			+53

Public APIs missing exports

Total count of every type that is part of your API that should be exported but is not. This will cause broken links in the API documentation system. Target amount is 0. Run node scripts/build_api_docs --plugin [yourplugin] --stats exports for more detailed information.

id	before	after	diff
inference	5	6	+1

Page load bundle

Size of the bundles that are downloaded on every page load. Target size is below 100kb

id	before	after	diff
inference	7.5KB	7.6KB	+85.0B

Unknown metric groups

API count

id	before	after	diff
@kbn/inference-common	166	174	+8
@kbn/inference-langchain	-	48	+48
inference	40	42	+2
total			+58

ESLint disabled line counts

id	before	after	diff
@kbn/router-to-openapispec	3	2	-1

Total ESLint disabled count

id	before	after	diff
@kbn/router-to-openapispec	3	2	-1

History

• 💔 Build #271672 failed 2a7ef42
• 💔 Build #271536 failed 3e5cd91
• 💔 Build #271418 failed f9442c1
• 💔 Build #271403 failed 66cfefc
• 💔 Build #270521 failed 0a33655

[elasticmachine]

Pinging @elastic/appex-ai-infra (Team:AI Infra)

[gsoldevila]

Core changes LGTM (review only): Removing an eslint exception is usually good news!

[KDKHD]

Looks good from GenAi side. Just one comment but it is a nit pick.

[KDKHD]

Are these extra comments? Or can you clarify why the same text is repeated below.

[pgayvallet]

nice catch thanks, yeah those were copy pasted to create the consts in the test file, will remove the comments