n8n/packages/testing/playwright/tests
Claude 1822c994aa
refactor(core): Consolidate MCP workflow validation into create_workflow_from_code
Drop the standalone validate_workflow MCP tool: create_workflow_from_code
already runs the identical parse/validate pipeline, so the pre-flight cost
an extra round trip (the full workflow code was sent twice per build) and
one more tool in every tools/list response.

To keep the validation gate as strong without the pre-flight:

- ParseValidateHandler now exposes fatal validation issues in a separate
  `errors` field (still folded into `warnings` for the code-builder agent).
- create_workflow_from_code hard-blocks on fatal validation errors and
  returns them all without persisting anything.
- Node-group violations (102_mcp_canvas_groups) are checked before the
  save path and reported all at once, instead of first-violation-only.
- MCP instructions, tool descriptions, recovery hints, and scope maps no
  longer reference the removed tool.

The in-app code-builder agent's internal validate_workflow tool is
unaffected.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EKFYgNjgLFPkft76sXghep
2026-07-31 13:49:39 +03:00
..
cli-workflows chore: Apply biome formatting to playwright package (no-changelog) (#26586) 2026-03-05 10:47:20 +00:00
dev-server-smoke test: Enforce canonical e2e spec owners and fix owner-name drift (no-changelog) (#32591) 2026-06-24 07:44:15 +01:00
e2e refactor(core): Consolidate MCP workflow validation into create_workflow_from_code 2026-07-31 13:49:39 +03:00
evals test(ai-builder): Move the workflow-eval case corpus to LangTracer (no-changelog) (#33983) 2026-07-15 08:58:49 +00:00
infrastructure fix: Chat works when running on separate webhook server (#33724) 2026-07-08 06:54:14 +00:00
performance test: Enforce canonical e2e spec owners and fix owner-name drift (no-changelog) (#32591) 2026-06-24 07:44:15 +01:00