[9.2] Update langchain (main) (#240493)

[patrykkopycinski]

Backport

This is a backport of PR #240493 to the 9.2 branch.

Summary

Backport of the langchain major version update (v0.x → v1.x) including:

• @langchain/core 0.3.80 → 1.1.31
• langchain 0.3.35 → 1.2.30
• @langchain/langgraph 0.4.9 → 1.1.5
• @langchain/openai 0.6.14 → 1.2.8
• @langchain/aws 0.1.15 → 1.3.1
• @langchain/community replaced by @langchain/classic
• @langchain/google-genai / @langchain/google-vertexai / @langchain/google-common 0.2.x → 2.1.x
• AWS SDK packages updated to 3.994.0
• Import paths migrated from langchain/* to @langchain/classic/* and @langchain/core/*
• Langchain Annotation.Root state updated to use channelDefault() helper

Made with Cursor

[macroscopeapp]

🟡 Medium utils/enrich_graph_input_messages.ts:58

The spread ...lastMessage.additional_kwargs places metadata at the top level of the HumanMessage config instead of nesting it under additional_kwargs. This causes properties like name or function_call to be lost or misinterpreted by downstream consumers expecting LangChain's standard message structure. Use additional_kwargs: lastMessage.additional_kwargs to preserve the proper nesting.

      messagesCopy[messagesCopy.length - 1] = new HumanMessage({
        content: `${userPrompt} ${lastMessage.content}`.trim(),
-        ...lastMessage.additional_kwargs,
+        additional_kwargs: lastMessage.additional_kwargs,
      });

🤖 Copy this AI Prompt to have your agent fix this:

In file x-pack/solutions/security/plugins/elastic_assistant/server/lib/langchain/utils/enrich_graph_input_messages.ts around lines 58-61:

The spread `...lastMessage.additional_kwargs` places metadata at the top level of the `HumanMessage` config instead of nesting it under `additional_kwargs`. This causes properties like `name` or `function_call` to be lost or misinterpreted by downstream consumers expecting LangChain's standard message structure. Use `additional_kwargs: lastMessage.additional_kwargs` to preserve the proper nesting.

Evidence trail:
1. Code under review: x-pack/solutions/security/plugins/elastic_assistant/server/lib/langchain/utils/enrich_graph_input_messages.ts lines 57-60 (REVIEWED_COMMIT)
2. LangChain HumanMessage constructor signature showing `additional_kwargs` as a distinct named field: https://github.com/langchain-ai/langchain/blob/master/libs/core/langchain_core/messages/human.py
3. LangChain test showing correct usage: `HumanMessage(content="human", additional_kwargs={"key": "value"}, name="human erick")` in https://github.com/langchain-ai/langchain/blob/master/libs/core/tests/unit_tests/test_messages.py

[macroscopeapp]

🟢 Low elastic-search-checkpoint-saver/index.ts:384

The deleteByQuery calls do not check the response for partial failures. If some documents fail to delete due to version conflicts or other issues, those failures are silently ignored and the method completes successfully despite incomplete deletion. Consider checking result.failures.length and throwing an error if any deletions failed.

🤖 Copy this AI Prompt to have your agent fix this:

In file x-pack/platform/packages/shared/kbn-langgraph-checkpoint-saver/server/elastic-search-checkpoint-saver/index.ts around line 384:

The `deleteByQuery` calls do not check the response for partial failures. If some documents fail to delete due to version conflicts or other issues, those failures are silently ignored and the method completes successfully despite incomplete deletion. Consider checking `result.failures.length` and throwing an error if any deletions failed.

Evidence trail:
x-pack/platform/packages/shared/kbn-langgraph-checkpoint-saver/server/elastic-search-checkpoint-saver/index.ts lines 383-392 (viewed at REVIEWED_COMMIT) - shows `deleteThread` method with two `deleteByQuery` calls that don't capture or check the response.

Elasticsearch documentation: https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-delete-by-query - confirms `deleteByQuery` response includes `failures` array, `version_conflicts`, `deleted`, `total`, and `timed_out` fields that indicate partial failures.

[macroscopeapp]

🟡 Medium elastic-search-checkpoint-saver/index.ts:384

deleteThread uses hardcoded refresh: true instead of this.refreshPolicy, so it ignores the configurable refresh policy set via constructor. This is inconsistent with put and putWrites, which both respect this.refreshPolicy.

-      refresh: true,
+      refresh: this.refreshPolicy,

🤖 Copy this AI Prompt to have your agent fix this:

In file x-pack/platform/packages/shared/kbn-langgraph-checkpoint-saver/server/elastic-search-checkpoint-saver/index.ts around lines 384-387:

`deleteThread` uses hardcoded `refresh: true` instead of `this.refreshPolicy`, so it ignores the configurable refresh policy set via constructor. This is inconsistent with `put` and `putWrites`, which both respect `this.refreshPolicy`.

Evidence trail:
x-pack/platform/packages/shared/kbn-langgraph-checkpoint-saver/server/elastic-search-checkpoint-saver/index.ts lines 384-392 (deleteThread with hardcoded refresh: true), lines 369-371 (put uses this.refreshPolicy), lines 446-449 (putWrites uses this.refreshPolicy), line 50 (refreshPolicy in constructor params)

[patrykkopycinski]

/ci

[patrykkopycinski]

/ci

[patrykkopycinski]

/ci

[patrykkopycinski]

/ci

[patrykkopycinski]

/ci

[patrykkopycinski]

/ci

[elasticmachine]

💛 Build succeeded, but was flaky

• Buildkite Build
• Commit: 722b3b6

Failed CI Steps

• AppEx AI-Infra Inference APIs FTR tests #1

Test Failures

• [job] [logs] AppEx AI-Infra Inference APIs FTR tests #1 / Inference plugin - API integration tests product docs base installs the ELSER product docs

Metrics [docs]

Async chunks

Total size of all lazy-loaded chunks that will be downloaded as the user navigates the app

id	before	after	diff
aiops	528.7KB	528.7KB	+3.0B
apm	2.7MB	2.7MB	-3.0B
controls	488.6KB	488.6KB	+3.0B
dashboard	662.8KB	662.8KB	+3.0B
dashboardMarkdown	71.6KB	71.6KB	+3.0B
dataVisualizer	596.5KB	596.5KB	+3.0B
discover	1.2MB	1.2MB	+3.0B
embeddableAlertsTable	1000.2KB	1000.2KB	+3.0B
imageEmbeddable	111.9KB	111.9KB	+3.0B
links	99.1KB	99.1KB	+3.0B
maps	3.1MB	3.1MB	-3.0B
ml	5.4MB	5.4MB	-3.0B
slo	982.3KB	982.3KB	+3.0B
synthetics	1.0MB	1.0MB	+3.0B
visualizations	351.1KB	351.1KB	+3.0B
total			+27.0B

Page load bundle

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

id	before	after	diff
cloudExperiments	92.8KB	94.7KB	+1.9KB
kbnUiSharedDeps-npmDll	6.5MB	6.5MB	+66.0B
kbnUiSharedDeps-srcJs	3.8MB	3.9MB	+53.7KB
total			+55.7KB

History

• 💔 Build #418373 failed eaf7b54
• 💚 Build #416480 succeeded 858637f
• 💔 Build #416369 failed f00ab7b
• 💔 Build #416248 failed 4f22f62