Skip to content

test(llmops): reproduce checker JSON failure on typographic quotes #21

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
bweinzierl wants to merge 1 commit into
base: main
Choose a base branch
from
+17 −0
Draft

test(llmops): reproduce checker JSON failure on typographic quotes #21

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 2 additions & 0 deletions internal/llmops/json_response_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,6 +78,8 @@ func TestCheckersJSONResponse(t *testing.T) {
{name: "nested_code_literal", output: "Reviewed Options{Child: Child{Enabled: true}}.\n" + checker.payload, want: 1},
{name: "quoted_prose", output: "The \"review\" follows.\n" + checker.payload, want: 1},
{name: "escaped_strings", output: strings.Replace(checker.payload, "The", `Quoted \"{x}\" and C:\\tmp affect the`, 1), want: 1},
{name: "typographic_quotes", output: strings.Replace(checker.payload, "The", "„Quoted“ and «cited» affect the", 1), want: 1},
{name: "typographic_quote_normalized_unescaped", output: strings.Replace(checker.payload, "The", `„Quoted" affects the`, 1), wantErr: true},
{name: "two_results", output: checker.payload + "\n" + checker.payload, wantErr: true},
{name: "clean_result_after_findings", output: checker.payload + `{"findings":[]}`, wantErr: true},
{name: "empty_object_after_payload", output: checker.payload + "\n{}", wantErr: true},
Expand Down
15 changes: 15 additions & 0 deletions internal/llmops/writingguide_eval_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -150,3 +150,18 @@ func TestWritingGuideEval_ConflatedDraftYieldsConflation(t *testing.T) {
return hasAxis(r, "conflation")
})
}

// TestWritingGuideEval_TypographicQuotesStayValidJSON pins the JSON contract
// for drafts quoted with typographic quotation marks: German „…“ close with
// U+201C, which the model tends to echo as an unescaped straight quote when it
// cites the draft, breaking the response. Any findings are fine — the case
// fails only on unparseable output.
func TestWritingGuideEval_TypographicQuotesStayValidJSON(t *testing.T) {
draft := &model.Entry{
Type: model.TypeSignal, Kind: model.KindDone, Layer: model.LayerTactical,
Content: "Die nutzersichtbaren Bestie-Code-Texte sind umformuliert und über MR !106 nach `main` gemergt: Kein Button spricht mehr von „Code einlösen“ — die Wording-Direktive ist umgesetzt, und der Paywall-Einstieg verspricht keine Store-Code-Einlösung mehr.\n\n" +
"- Auf Bennos Wunsch wurden die begleitenden Fließtexte angeglichen: In `NoBestie.tsx` heißt es jetzt „… oder nutze einen Bestie-Code …“ statt „löse … ein“, und die Fehlermeldung spricht von „Verbindungslink“ statt „Einladungslink“.\n" +
"- Die Schreibweise ist auf „Bestie-Code“ mit Bindestrich vereinheitlicht.",
}
runGuideEvalPassRate(t, draft, blockingTier, func(*llmops.WritingGuideResult) error { return nil })
}
Loading