refactor(workflow-executor): drop direct @langchain dependency

[Scra3]

Why

The executor declared and imported @langchain/openai directly, while @forestadmin/ai-proxy already owns the langchain coupling (model construction, message/tool types). This removes that duplicated dependency edge.

What

• ServerAiAdapter no longer instantiates ChatOpenAI itself. It builds a single synthetic openai AiConfiguration whose configuration.fetch is rewritten to hit the Forest AI proxy (/liana/v1/ai-proxy) with forest-secret-key, hands it to the AiClient it already owns, and delegates getModel() to AiClient.getModel() (→ createBaseChatModel). Runtime behavior is identical; the provider/model is still chosen server-side.
• The remaining BaseChatModel type imports (server-ai-adapter, ai-client-adapter, always-error-ai-model-port, ports/ai-model-port) now come from the @forestadmin/ai-proxy façade, which already re-exports it.
• @langchain/openai removed from packages/workflow-executor/package.json. The executor now imports from @langchain/* zero times — grep -rn "@langchain" src is empty.

Notes

• yarn.lock is unchanged: @langchain/openai is still pulled transitively by ai-proxy, so only the workspace dependency edge is dropped.
• server-ai-adapter.test.ts reworked: it now asserts the config handed to AiClient (provider/model/maxRetries + the fetch redirect to the proxy with forest-secret-key) and that getModel() delegates to the client.

Verification

• yarn workspace @forestadmin/workflow-executor build ✅
• yarn workspace @forestadmin/workflow-executor lint ✅ (7 pre-existing warnings)
• yarn workspace @forestadmin/workflow-executor test → 967 passed ✅
• yarn workspace @forestadmin/ai-proxy test → 443 passed ✅

🤖 Generated with Claude Code

Note

Remove direct @langchain/openai dependency from workflow-executor in favor of @forestadmin/ai-proxy

• Removes @langchain/openai from package.json and re-sources BaseChatModel and related types through @forestadmin/ai-proxy.
• Refactors ServerAiAdapter to instantiate AiClient (from ai-proxy) with a pre-built proxy configuration via a new buildProxyConfiguration static method, replacing direct ChatOpenAI construction.
• getModel now delegates to AiClient.getModel() rather than constructing the model inline; the proxy URL rewriting and header swap (Authorization → forest-secret-key) move into the configuration builder.

Changes since #1638 opened

• Modified StepExecutorFactory.create factory method to return executors that surface WorkflowExecutorError.userMessage in stepOutcome.error when factory construction fails with a WorkflowExecutorError, replacing the previous generic error message [45c47af]
• Refactored AiClient mocking infrastructure in test files to use explicit jest.fn mocks for constructor argument capture and method delegation [45c47af]
• Added test coverage verifying that StepExecutorFactory.create surfaces userMessage from WorkflowExecutorError instances in stepOutcome.error [45c47af]

^{Macroscope summarized b5cbae2.}

[qltysh]

Coverage Impact

This PR will not change total coverage.

Modified Files with Diff Coverage (2)

Rating	File	% Diff	Uncovered Line #s
	packages/workflow-executor/src/executors/step-executor-factory.ts	100.0%
	packages/workflow-executor/src/adapters/server-ai-adapter.ts	100.0%
	Total	100.0%

🚦 See full report on Qlty Cloud »

🛟 Help

• Diff Coverage: Coverage for added or modified lines of code (excludes deleted files). Learn more.

• Total Coverage: Coverage for the whole repository, calculated as the sum of all File Coverage. Learn more.

• File Coverage: Covered Lines divided by Covered Lines plus Missed Lines. (Excludes non-executable lines including blank lines and comments.)

  • Indirect Changes: Changes to File Coverage for files that were not modified in this PR. Learn more.

[PMerlet]

LGTM

[qltysh]

1 new issue

Tool	Category	Rule	Count
qlty	Structure	Function with many returns (count = 8): create	1