Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions e2e-tests/local_agent_auto.spec.ts
Original file line number Diff line number Diff line change
@@ -1,7 +1,17 @@
import { expect } from "@playwright/test";
import { testSkipIfWindows } from "./helpers/test_helper";

testSkipIfWindows("local-agent - auto model", async ({ po }) => {
await po.setUpDyadPro({ localAgent: true, localAgentUseAutoModel: true });
await po.page.evaluate(async () => {
await (window as any).electron.ipcRenderer.invoke("set-user-settings", {
enableCodeExplorer: false,
});
});
await expect
.poll(() => po.settings.recordSettings().enableCodeExplorer)
.toBe(false);

await po.importApp("minimal");

await po.sendPrompt("[dump]");
Expand Down
1 change: 1 addition & 0 deletions rules/e2e-testing.md
Original file line number Diff line number Diff line change
Expand Up @@ -196,6 +196,7 @@ If `npm run build` / Electron Forge packaging fails with `Failed to locate modul
- **Filesystem-heavy IPC assertions**: Operations that delete or copy whole app directories (e.g. bulk app delete) can exceed the 5s default expect timeout on CI runners. Give the post-operation assertion an explicit `{ timeout: 30_000 }`.
- **Fake Anthropic engine routes**: When app code uses Anthropic direct passthrough, the fake LLM server must handle `/v1/messages` (and provider-prefixed variants like `/engine/v1/messages`), not just `/chat/completions`. Anthropic tool results come back as user messages with `tool_result` content blocks, so fixture turn counting must skip those as user prompts.
- **Legacy local-agent `code_search` coverage**: `code_search` is hidden when code explorer is enabled and ready, so specs that explicitly test `code_search` should set `enableCodeExplorer: false` and poll persisted settings before sending the fixture prompt. Otherwise local and CI can render different snapshots depending on code-explorer readiness.
- **Local-agent request-dump snapshots and code explorer**: If a request-dump spec is not testing `explore_code`, pin `enableCodeExplorer: false` and poll `po.settings.recordSettings().enableCodeExplorer` before sending the prompt. Otherwise the serialized prompt/tool list can flip between `code_search` and `explore_code` depending on whether indexing finished first.
- **Supabase destructive SQL migration proposals**: Auto-approve does not apply destructive SQL changes. If an E2E prompt returns a destructive SQL proposal (for example `DROP TABLE`), call `po.approveProposal()` before asserting that a migration file or app-file change exists.

## Triaging failures from a CI run's html-report
Expand Down
Loading