[kbn-evals] Fix APM/OTel tracing conflict and inference endpoint connector resolution by viduni94 · Pull Request #259446 · elastic/kibana

viduni94 · 2026-03-24T18:45:02Z

Summary

Fixes two issues breaking kbn-evals runs (both local and CI):

1. APM / OpenTelemetry tracing conflict

A recent validation in initTelemetry (#258303, #258663) throws when Elastic APM and OpenTelemetry tracing are both active. The evals_tracing Scout config enables OTel tracing but didn't explicitly disable APM, causing Kibana (and the Playwright worker) to crash on startup.

Fix:

Added a coerceCliValue helper in applyConfigOverrides (kbn-apm-config-loader) that converts 'true'/'false' to booleans and numeric strings to numbers before they're set in the config object.
Added --elastic.apm.active=false and --elastic.apm.contextPropagationOnly=false to the evals_tracing Scout server config and to require_init_apm.js (for the Playwright worker when TRACING_EXPORTERS is set).
Updated the kbn-evals README to document the required APM settings when configuring tracing in kibana.dev.yml.

2. Inference endpoint connector resolution

#258530 consolidated LLM connector listing through the inference plugin's getConnectorList(), which now returns inference endpoint IDs (e.g.: .anthropic-claude-4.6-opus-chat_completion) instead of Kibana stack connector keys (e.g.: elastic-llm-claude-46-opus). kbn-evals was still passing the stack connector key to the inference API, which then tried to execute it as a Kibana action - resulting in "Saved object [action/.anthropic-claude-4.6-opus-chat_completion] not found".

Fix:

createConnectorFixture now detects .inference-type connectors and extracts their inferenceId from the config, using the ES inference endpoint ID directly. This bypasses the Kibana actions framework and aligns with the unified connector model from #258530.

Checklist

Unit or functional tests were updated or added to match the most common scenarios
The PR description includes the appropriate Release Notes section, and the correct release_note:* label is applied per the guidelines
Review the backport guidelines and apply applicable backport:* labels.

…tion

elasticmachine · 2026-03-24T18:45:07Z

Pinging @elastic/obs-ai-team (Team:obs-ai)

spong · 2026-03-24T20:10:52Z

x-pack/platform/packages/shared/kbn-evals/README.md

+elastic.apm.active: false
+elastic.apm.contextPropagationOnly: false


Some additional doc references we'll probably want to update along with this one:

This file:

Lines ~356-358 — the "configure the HTTP exporter in kibana.dev.yml" block only shows exporters, but the note about APM flags is added nearby (good).

Lines ~423-434 — the trace-based evaluators prerequisite section also shows exporter config without APM flags.
These should either include the flags or cross-reference the note about them.

Evals plugin readme: x-pack/platform/plugins/shared/evals/README.md

Missing elastic.apm.active: false and elastic.apm.contextPropagationOnly: false before the telemetry lines. Users following this README to configure their kibana.dev.yml will hit the same APM/OTel conflict.

Agent Builder readme: x-pack/platform/plugins/shared/agent_builder/README.md

First example on lines 21-28 and the second example at lines 38-44. Neither snippet includes the APM-disable flags. Users enabling tracing for agent_builder will hit the same conflict.

Entity Analytics readme: kbn-evals-suite-entity-analytics/README.md

This config is contradictory -- it sets sample_rate and exporters but disables tracing. This looks like an oversight and should probably enable tracing (in which case it also needs the APM flags). Not our config, but might as well cleanup. cc @ymao1

Thanks @spong
Updated here - 751b67c

spong

Code reviewed and LGTM! 👍 Thanks for these fixes @viduni94 🙏

Left one comment about some further documentation updates if you don't mind, and also confirmed the new Scout server config fix propagates to all 11 eval suites, so that all LGTM.

viduni94 · 2026-03-24T21:48:20Z

Code reviewed and LGTM! 👍 Thanks for these fixes @viduni94 🙏

Thanks @spong

Left one comment about some further documentation updates if you don't mind

Yes, of course. I updated them

and also confirmed the new Scout server config fix propagates to all 11 eval suites, so that all LGTM.

Thank you 🙏🏻

viduni94 · 2026-03-25T15:46:47Z

@spong @csr I had to push 2 more fixes to override env variables to disable APM

Thank you

Let me know whether these look okay to you. With these fixes, the evals are running without any config errors.

spong · 2026-03-25T15:51:52Z

x-pack/platform/packages/shared/kbn-evals/src/config/require_init_apm.js

+  // CI sets ELASTIC_APM_ACTIVE=true globally. The APM config loader merges env vars
+  // after Kibana config (CLI args), so the env var would override our --elastic.apm.active=false.


TIL, thank you!

csr

Recent custom servers config changes LGTM

sphilipse · 2026-03-25T16:09:30Z

x-pack/platform/packages/shared/kbn-evals/src/utils/create_connector_fixture.ts

    }
  }

+  // .inference connectors are backed by ES inference endpoints.


I think this is okay for now, but the better solution would be to adapt use to accept both. We shouldn't need to do this transformation, as the inference plugin should accept both connector Ids and inference Ids. If they don't then that's a bug we need to address on the inference plugin.

Thanks @sphilipse
My initial thought was to add the fix here -

kibana/x-pack/platform/plugins/shared/inference/server/chat_complete/callback_api.ts

Lines 322 to 356 in b791960

: async () => {

const executor = await getInferenceExecutor({

connectorId,

request,

actions,

esClient,

logger,

});

const connector = executor.getConnector();

const connectorType = connector.type;

const inferenceAdapter = getInferenceAdapter(connectorType);

if (!inferenceAdapter) {

throw createInferenceRequestError(

`Adapter for type ${connectorType} not implemented`,

400

);

}

return {

callbackContext: {

model: {

family: getConnectorFamily(connector),

provider: getConnectorProvider(connector),

id: getConnectorDefaultModel(connector),

},

},

getSpanModel: (modelName) => ({

id: modelName ?? getConnectorDefaultModel(connector),

family: getConnectorFamily(connector),

provider: getConnectorProvider(connector),

}),

chatComplete: (options) => inferenceAdapter.chatComplete({ ...options, executor }),

};

};

(But wasn't sure whether that would affect anything else as we are moving away of from stack connectors AFAIK)

The problem flow was:

resolveAndCreatePipeline checks endpointIdCache.has('elastic-llm-claude-46-opus') → false (cache only has ES endpoint IDs)

Falls into the stack-connector path, calls getInferenceExecutor

Inside, getConnectorById resolves the stack connector to an inference endpoint with connectorId: '.anthropic-claude-4.6-opus-chat_completion' and isInferenceEndpoint: true

But createInferenceExecutor still calls actionsClient.execute() with the endpoint ID → "Saved object not found"

Proposed fix:

: async () => { const executor = await getInferenceExecutor({ connectorId, request, actions, esClient, logger, }); const connector = executor.getConnector(); // If the stack connector resolved to an inference endpoint, // use the endpoint executor instead of the actions-based one. if (connector.isInferenceEndpoint) { const inferenceId = connector.connectorId; const endpointMeta = await resolveInferenceEndpoint({ inferenceId, esClient }); const endpointExecutor = createInferenceEndpointExecutor({ inferenceId, esClient }); return { callbackContext: { model: endpointMeta.modelId ? { id: endpointMeta.modelId } : undefined, }, getSpanModel: (modelName) => endpointMeta.provider ? { id: modelName ?? endpointMeta.modelId, provider: endpointMeta.provider } : undefined, chatComplete: (options) => inferenceEndpointAdapter.chatComplete({ ...options, executor: endpointExecutor }), }; } // ... rest of existing stack connector logic

Does this make sense to you? Or do you have any suggestions to resolve it in a better way?
Appreciate any thoughts 🙏🏻

Yes, I think this makes sense! Thanks @viduni94, you caught a bug for us :)

Thanks @sphilipse
I'll open a follow up PR with the fix.

viduni94 · 2026-03-25T17:33:59Z

/ci

viduni94 · 2026-03-25T17:34:42Z

run docs-build

viduni94 · 2026-03-25T19:11:00Z

/sync-ci

elasticmachine · 2026-03-25T19:13:01Z

💚 Build Succeeded

Buildkite Build
Commit: a84f58a

Metrics [docs]

✅ unchanged

History

💔 Build #416529 failed 194e534
💚 Build #416447 succeeded 9015a64
💔 Build #416381 failed 278fd59
💔 Build #416349 failed f71d629
💔 Build #415921 failed b210c7c
💚 Build #415874 succeeded 751b67c

cc @viduni94

…ector resolution (elastic#259446) Closes elastic#259472 ## Summary Fixes two issues breaking `kbn-evals` runs (both local and CI): ### 1. APM / OpenTelemetry tracing conflict A recent validation in `initTelemetry` (elastic#258303, elastic#258663) throws when Elastic APM and OpenTelemetry tracing are both active. The `evals_tracing` Scout config enables OTel tracing but didn't explicitly disable APM, causing Kibana (and the Playwright worker) to crash on startup. Fix: - Added a `coerceCliValue` helper in `applyConfigOverrides` (`kbn-apm-config-loader`) that converts 'true'/'false' to booleans and numeric strings to numbers before they're set in the config object. - Added `--elastic.apm.active=false` and `--elastic.apm.contextPropagationOnly=false` to the `evals_tracing` Scout server config and to `require_init_apm.js` (for the Playwright worker when `TRACING_EXPORTERS` is set). - Updated the `kbn-evals` README to document the required APM settings when configuring tracing in `kibana.dev.yml`. ### 2. Inference endpoint connector resolution elastic#258530 consolidated LLM connector listing through the inference plugin's `getConnectorList()`, which now returns inference endpoint IDs (e.g.: `.anthropic-claude-4.6-opus-chat_completion`) instead of Kibana stack connector keys (e.g.: `elastic-llm-claude-46-opus`). `kbn-evals` was still passing the stack connector key to the inference API, which then tried to execute it as a Kibana action - resulting in "Saved object `[action/.anthropic-claude-4.6-opus-chat_completion]` not found". Fix: - `createConnectorFixture` now detects `.inference-type` connectors and extracts their `inferenceId` from the config, using the ES inference endpoint ID directly. This bypasses the Kibana actions framework and aligns with the unified connector model from [elastic#258530](elastic#258530). ### Checklist - [x] [Unit or functional tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html) were updated or added to match the most common scenarios - [x] The PR description includes the appropriate Release Notes section, and the correct `release_note:*` label is applied per the [guidelines](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process) - [x] Review the [backport guidelines](https://docs.google.com/document/d/1VyN5k91e5OVumlc0Gb9RPa3h1ewuPE705nRtioPiTvY/edit?usp=sharing) and apply applicable `backport:*` labels.

…oints (#259656) Closes #259641 ## Summary When a stack connector ID (e.g.: a preconfigured `.inference` connector) is passed to the inference plugin's `chatComplete` API, `getConnectorById` may resolve it to an `InferenceConnector` with `isInferenceEndpoint: true`. Previously, `resolveAndCreatePipeline` only checked the `endpointIdCache` to decide whether to use the inference endpoint execution path. If the cache didn't contain the ID, it fell through to the stack connector adapter path, which then failed because there's no adapter for the .inference connector type. This caused a `Saved object [action/<inference-endpoint-id>] not found` error when callers (e.g.: `kbn-evals`) passed preconfigured `.inference` connector IDs, because the code attempted to execute the ES inference endpoint ID as a Kibana saved-object action. ### Changes - `callback_api.ts`: After `getInferenceExecutor` resolves the connector in the stack connector branch, check `connector.isInferenceEndpoint`. If true, redirect to the inference endpoint execution path (`resolveInferenceEndpoint` + `createInferenceEndpointExecutor` + `inferenceEndpointAdapter`). - `api.test.ts`: Added tests covering the "stack connector resolving to inference endpoint" path. - `create_connector_fixture.ts` (`kbn-evals`): Removed the client-side workaround that was extracting inferenceId from `.inference` connectors - no longer needed now that the inference plugin handles this server-side. - `create_connector_fixture.test.ts`: Removed corresponding workaround tests. ## Related - #258530 introduced unified connector listing via `getConnectorList()`/`getConnectorById()`, which returns inference endpoints with `isInferenceEndpoint: true` - #259446 added a temporary client-side workaround in `kbn-evals` (now removed by this PR) ### Checklist - [x] [Unit or functional tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html) were updated or added to match the most common scenarios - [x] The PR description includes the appropriate Release Notes section, and the correct `release_note:*` label is applied per the [guidelines](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process) - [x] Review the [backport guidelines](https://docs.google.com/document/d/1VyN5k91e5OVumlc0Gb9RPa3h1ewuPE705nRtioPiTvY/edit?usp=sharing) and apply applicable `backport:*` labels.

…oints (elastic#259656) Closes elastic#259641 ## Summary When a stack connector ID (e.g.: a preconfigured `.inference` connector) is passed to the inference plugin's `chatComplete` API, `getConnectorById` may resolve it to an `InferenceConnector` with `isInferenceEndpoint: true`. Previously, `resolveAndCreatePipeline` only checked the `endpointIdCache` to decide whether to use the inference endpoint execution path. If the cache didn't contain the ID, it fell through to the stack connector adapter path, which then failed because there's no adapter for the .inference connector type. This caused a `Saved object [action/<inference-endpoint-id>] not found` error when callers (e.g.: `kbn-evals`) passed preconfigured `.inference` connector IDs, because the code attempted to execute the ES inference endpoint ID as a Kibana saved-object action. ### Changes - `callback_api.ts`: After `getInferenceExecutor` resolves the connector in the stack connector branch, check `connector.isInferenceEndpoint`. If true, redirect to the inference endpoint execution path (`resolveInferenceEndpoint` + `createInferenceEndpointExecutor` + `inferenceEndpointAdapter`). - `api.test.ts`: Added tests covering the "stack connector resolving to inference endpoint" path. - `create_connector_fixture.ts` (`kbn-evals`): Removed the client-side workaround that was extracting inferenceId from `.inference` connectors - no longer needed now that the inference plugin handles this server-side. - `create_connector_fixture.test.ts`: Removed corresponding workaround tests. ## Related - elastic#258530 introduced unified connector listing via `getConnectorList()`/`getConnectorById()`, which returns inference endpoints with `isInferenceEndpoint: true` - elastic#259446 added a temporary client-side workaround in `kbn-evals` (now removed by this PR) ### Checklist - [x] [Unit or functional tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html) were updated or added to match the most common scenarios - [x] The PR description includes the appropriate Release Notes section, and the correct `release_note:*` label is applied per the [guidelines](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process) - [x] Review the [backport guidelines](https://docs.google.com/document/d/1VyN5k91e5OVumlc0Gb9RPa3h1ewuPE705nRtioPiTvY/edit?usp=sharing) and apply applicable `backport:*` labels.

…ector resolution (elastic#259446) Closes elastic#259472 ## Summary Fixes two issues breaking `kbn-evals` runs (both local and CI): ### 1. APM / OpenTelemetry tracing conflict A recent validation in `initTelemetry` (elastic#258303, elastic#258663) throws when Elastic APM and OpenTelemetry tracing are both active. The `evals_tracing` Scout config enables OTel tracing but didn't explicitly disable APM, causing Kibana (and the Playwright worker) to crash on startup. Fix: - Added a `coerceCliValue` helper in `applyConfigOverrides` (`kbn-apm-config-loader`) that converts 'true'/'false' to booleans and numeric strings to numbers before they're set in the config object. - Added `--elastic.apm.active=false` and `--elastic.apm.contextPropagationOnly=false` to the `evals_tracing` Scout server config and to `require_init_apm.js` (for the Playwright worker when `TRACING_EXPORTERS` is set). - Updated the `kbn-evals` README to document the required APM settings when configuring tracing in `kibana.dev.yml`. ### 2. Inference endpoint connector resolution elastic#258530 consolidated LLM connector listing through the inference plugin's `getConnectorList()`, which now returns inference endpoint IDs (e.g.: `.anthropic-claude-4.6-opus-chat_completion`) instead of Kibana stack connector keys (e.g.: `elastic-llm-claude-46-opus`). `kbn-evals` was still passing the stack connector key to the inference API, which then tried to execute it as a Kibana action - resulting in "Saved object `[action/.anthropic-claude-4.6-opus-chat_completion]` not found". Fix: - `createConnectorFixture` now detects `.inference-type` connectors and extracts their `inferenceId` from the config, using the ES inference endpoint ID directly. This bypasses the Kibana actions framework and aligns with the unified connector model from [elastic#258530](elastic#258530). ### Checklist - [x] [Unit or functional tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html) were updated or added to match the most common scenarios - [x] The PR description includes the appropriate Release Notes section, and the correct `release_note:*` label is applied per the [guidelines](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process) - [x] Review the [backport guidelines](https://docs.google.com/document/d/1VyN5k91e5OVumlc0Gb9RPa3h1ewuPE705nRtioPiTvY/edit?usp=sharing) and apply applicable `backport:*` labels.

…oints (elastic#259656) Closes elastic#259641 ## Summary When a stack connector ID (e.g.: a preconfigured `.inference` connector) is passed to the inference plugin's `chatComplete` API, `getConnectorById` may resolve it to an `InferenceConnector` with `isInferenceEndpoint: true`. Previously, `resolveAndCreatePipeline` only checked the `endpointIdCache` to decide whether to use the inference endpoint execution path. If the cache didn't contain the ID, it fell through to the stack connector adapter path, which then failed because there's no adapter for the .inference connector type. This caused a `Saved object [action/<inference-endpoint-id>] not found` error when callers (e.g.: `kbn-evals`) passed preconfigured `.inference` connector IDs, because the code attempted to execute the ES inference endpoint ID as a Kibana saved-object action. ### Changes - `callback_api.ts`: After `getInferenceExecutor` resolves the connector in the stack connector branch, check `connector.isInferenceEndpoint`. If true, redirect to the inference endpoint execution path (`resolveInferenceEndpoint` + `createInferenceEndpointExecutor` + `inferenceEndpointAdapter`). - `api.test.ts`: Added tests covering the "stack connector resolving to inference endpoint" path. - `create_connector_fixture.ts` (`kbn-evals`): Removed the client-side workaround that was extracting inferenceId from `.inference` connectors - no longer needed now that the inference plugin handles this server-side. - `create_connector_fixture.test.ts`: Removed corresponding workaround tests. ## Related - elastic#258530 introduced unified connector listing via `getConnectorList()`/`getConnectorById()`, which returns inference endpoints with `isInferenceEndpoint: true` - elastic#259446 added a temporary client-side workaround in `kbn-evals` (now removed by this PR) ### Checklist - [x] [Unit or functional tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html) were updated or added to match the most common scenarios - [x] The PR description includes the appropriate Release Notes section, and the correct `release_note:*` label is applied per the [guidelines](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process) - [x] Review the [backport guidelines](https://docs.google.com/document/d/1VyN5k91e5OVumlc0Gb9RPa3h1ewuPE705nRtioPiTvY/edit?usp=sharing) and apply applicable `backport:*` labels.

…ector resolution (elastic#259446) Closes elastic#259472 ## Summary Fixes two issues breaking `kbn-evals` runs (both local and CI): ### 1. APM / OpenTelemetry tracing conflict A recent validation in `initTelemetry` (elastic#258303, elastic#258663) throws when Elastic APM and OpenTelemetry tracing are both active. The `evals_tracing` Scout config enables OTel tracing but didn't explicitly disable APM, causing Kibana (and the Playwright worker) to crash on startup. Fix: - Added a `coerceCliValue` helper in `applyConfigOverrides` (`kbn-apm-config-loader`) that converts 'true'/'false' to booleans and numeric strings to numbers before they're set in the config object. - Added `--elastic.apm.active=false` and `--elastic.apm.contextPropagationOnly=false` to the `evals_tracing` Scout server config and to `require_init_apm.js` (for the Playwright worker when `TRACING_EXPORTERS` is set). - Updated the `kbn-evals` README to document the required APM settings when configuring tracing in `kibana.dev.yml`. ### 2. Inference endpoint connector resolution elastic#258530 consolidated LLM connector listing through the inference plugin's `getConnectorList()`, which now returns inference endpoint IDs (e.g.: `.anthropic-claude-4.6-opus-chat_completion`) instead of Kibana stack connector keys (e.g.: `elastic-llm-claude-46-opus`). `kbn-evals` was still passing the stack connector key to the inference API, which then tried to execute it as a Kibana action - resulting in "Saved object `[action/.anthropic-claude-4.6-opus-chat_completion]` not found". Fix: - `createConnectorFixture` now detects `.inference-type` connectors and extracts their `inferenceId` from the config, using the ES inference endpoint ID directly. This bypasses the Kibana actions framework and aligns with the unified connector model from [elastic#258530](elastic#258530). ### Checklist - [x] [Unit or functional tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html) were updated or added to match the most common scenarios - [x] The PR description includes the appropriate Release Notes section, and the correct `release_note:*` label is applied per the [guidelines](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process) - [x] Review the [backport guidelines](https://docs.google.com/document/d/1VyN5k91e5OVumlc0Gb9RPa3h1ewuPE705nRtioPiTvY/edit?usp=sharing) and apply applicable `backport:*` labels.

…oints (elastic#259656) Closes elastic#259641 ## Summary When a stack connector ID (e.g.: a preconfigured `.inference` connector) is passed to the inference plugin's `chatComplete` API, `getConnectorById` may resolve it to an `InferenceConnector` with `isInferenceEndpoint: true`. Previously, `resolveAndCreatePipeline` only checked the `endpointIdCache` to decide whether to use the inference endpoint execution path. If the cache didn't contain the ID, it fell through to the stack connector adapter path, which then failed because there's no adapter for the .inference connector type. This caused a `Saved object [action/<inference-endpoint-id>] not found` error when callers (e.g.: `kbn-evals`) passed preconfigured `.inference` connector IDs, because the code attempted to execute the ES inference endpoint ID as a Kibana saved-object action. ### Changes - `callback_api.ts`: After `getInferenceExecutor` resolves the connector in the stack connector branch, check `connector.isInferenceEndpoint`. If true, redirect to the inference endpoint execution path (`resolveInferenceEndpoint` + `createInferenceEndpointExecutor` + `inferenceEndpointAdapter`). - `api.test.ts`: Added tests covering the "stack connector resolving to inference endpoint" path. - `create_connector_fixture.ts` (`kbn-evals`): Removed the client-side workaround that was extracting inferenceId from `.inference` connectors - no longer needed now that the inference plugin handles this server-side. - `create_connector_fixture.test.ts`: Removed corresponding workaround tests. ## Related - elastic#258530 introduced unified connector listing via `getConnectorList()`/`getConnectorById()`, which returns inference endpoints with `isInferenceEndpoint: true` - elastic#259446 added a temporary client-side workaround in `kbn-evals` (now removed by this PR) ### Checklist - [x] [Unit or functional tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html) were updated or added to match the most common scenarios - [x] The PR description includes the appropriate Release Notes section, and the correct `release_note:*` label is applied per the [guidelines](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process) - [x] Review the [backport guidelines](https://docs.google.com/document/d/1VyN5k91e5OVumlc0Gb9RPa3h1ewuPE705nRtioPiTvY/edit?usp=sharing) and apply applicable `backport:*` labels.

Fix APM/OTel tracing conflict and inference endpoint connector resolu…

aa9293d

…tion

viduni94 self-assigned this Mar 24, 2026

viduni94 added release_note:skip Skip the PR/issue when compiling release notes backport:skip This PR does not require backporting Team:obs-ai Observability AI team labels Mar 24, 2026

viduni94 requested review from a team and vigneshshanmugam as code owners March 24, 2026 18:45

viduni94 changed the title ~~Fix APM/OTel tracing conflict and inference endpoint connector resolu…~~ [kbn-evals] Fix APM/OTel tracing conflict and inference endpoint connector resolution Mar 24, 2026

spong reviewed Mar 24, 2026

View reviewed changes

spong approved these changes Mar 24, 2026

View reviewed changes

Update README files

751b67c

viduni94 requested review from a team as code owners March 24, 2026 21:46

viduni94 requested a review from tiansivive March 24, 2026 21:46

neptunian approved these changes Mar 24, 2026

View reviewed changes

TinaHeiligers approved these changes Mar 24, 2026

View reviewed changes

ymao1 approved these changes Mar 24, 2026

View reviewed changes

viduni94 enabled auto-merge (squash) March 24, 2026 23:35

viduni94 disabled auto-merge March 24, 2026 23:35

viduni94 enabled auto-merge (squash) March 24, 2026 23:35

Merge branch 'main' into fix-kbn-evals

b210c7c

pgayvallet approved these changes Mar 25, 2026

View reviewed changes

spong reviewed Mar 25, 2026

View reviewed changes

Merge branch 'main' into fix-kbn-evals

9015a64

csr approved these changes Mar 25, 2026

View reviewed changes

viduni94 enabled auto-merge (squash) March 25, 2026 16:04

sphilipse reviewed Mar 25, 2026

View reviewed changes

Merge branch 'main' into fix-kbn-evals

194e534

Changes from node scripts/lint.js --fix

a84f58a

elastic-vault-github-plugin-prod bot requested a review from a team as a code owner March 25, 2026 17:42

This was referenced Mar 25, 2026

Refactor AI insights evals to use scenario-based testing #258043

Merged

[Inference] resolveAndCreatePipeline should handle stack connector IDs that resolve to inference endpoints #259641

Closed

tylersmalley approved these changes Mar 25, 2026

View reviewed changes

viduni94 merged commit 149ebc8 into elastic:main Mar 25, 2026
16 checks passed

kibanamachine added the v9.4.0 label Mar 25, 2026

viduni94 mentioned this pull request Mar 25, 2026

[Inference] Handle stack connector IDs that resolve to inference endpoints #259656

Merged

3 tasks

		elastic.apm.active: false
		elastic.apm.contextPropagationOnly: false

		// CI sets ELASTIC_APM_ACTIVE=true globally. The APM config loader merges env vars
		// after Kibana config (CLI args), so the env var would override our --elastic.apm.active=false.

	: async () => {
	const executor = await getInferenceExecutor({
	connectorId,
	request,
	actions,
	esClient,
	logger,
	});
	const connector = executor.getConnector();
	const connectorType = connector.type;
	const inferenceAdapter = getInferenceAdapter(connectorType);

	if (!inferenceAdapter) {
	throw createInferenceRequestError(
	`Adapter for type ${connectorType} not implemented`,
	400
	);
	}

	return {
	callbackContext: {
	model: {
	family: getConnectorFamily(connector),
	provider: getConnectorProvider(connector),
	id: getConnectorDefaultModel(connector),
	},
	},
	getSpanModel: (modelName) => ({
	id: modelName ?? getConnectorDefaultModel(connector),
	family: getConnectorFamily(connector),
	provider: getConnectorProvider(connector),
	}),
	chatComplete: (options) => inferenceAdapter.chatComplete({ ...options, executor }),
	};
	};

Conversation

viduni94 commented Mar 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

1. APM / OpenTelemetry tracing conflict

2. Inference endpoint connector resolution

Checklist

Uh oh!

elasticmachine commented Mar 24, 2026

Uh oh!

spong Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

viduni94 Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

spong left a comment

Choose a reason for hiding this comment

Uh oh!

viduni94 commented Mar 24, 2026

Uh oh!

viduni94 commented Mar 25, 2026

Uh oh!

spong Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

csr left a comment

Choose a reason for hiding this comment

Uh oh!

sphilipse Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

viduni94 Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

sphilipse Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

viduni94 Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

viduni94 commented Mar 25, 2026

Uh oh!

viduni94 commented Mar 25, 2026

Uh oh!

viduni94 commented Mar 25, 2026

Uh oh!

Uh oh!

elasticmachine commented Mar 25, 2026

💚 Build Succeeded

Metrics [docs]

History

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

12 participants

viduni94 commented Mar 24, 2026 •

edited

Loading