Align prompts and strengthen validation context

internal/framework/validators/llm_context_test.go

@@ -0,0 +1,75 @@
+package validators
+
+import (
+	"encoding/json"
+	"strings"
+	"testing"
+
+	"gitea.maximumdirect.net/eric/audita/internal/core/chunking"
+	"gitea.maximumdirect.net/eric/audita/internal/core/schema"
+	"gitea.maximumdirect.net/eric/audita/internal/framework/proposals"
+)
+
+func TestValidationContextCarriesAdjacentEvidenceAndRelevantDefinitions(t *testing.T) {
+	req := makeReq([]proposals.EnrichedCorrectionProposal{mk(7, "gestures", "Jesters")})
+	req.WorkingTranscript = &schema.Transcript{Segments: []schema.Segment{
+		{ID: 1, Text: "The performers arrived."},
+		{ID: 2, Text: "Those gestures are back."},
+		{ID: 3, Text: "Lyra recognized the troupe."},
+		{ID: 4, Text: "Elsewhere."},
+	}}
+	req.CandidateProposal[0].TargetSegmentID = 2
+	req.Glossary = &schema.Glossary{Entries: []schema.GlossaryEntry{
+		{Name: "Jesters", Summary: "A traveling troupe"},
+		{Name: "Lyra", Summary: "The host"},
+		{Name: "Unrelated", Summary: "Not mentioned"},
+	}}
+	payload, immediate := BuildLLMValidationRequest("spoken_form_plausibility", LLMValidatorTypeSpokenFormPlausibility, req)
+	if len(immediate) != 0 || len(payload.Items) != 1 {
+		t.Fatalf("payload: %+v immediate: %+v", payload, immediate)
+	}
+	item := payload.Items[0]
+	if len(item.NeighboringSegments) != 2 || item.NeighboringSegments[0].ID != 1 || item.NeighboringSegments[1].ID != 3 || len(item.GlossaryEntries) != 2 || item.GlossaryEntries[0].Name != "Jesters" || item.GlossaryEntries[1].Name != "Lyra" {
+		t.Fatalf("context: %+v", item)
+	}
+	variables, err := BuildValidationVariables(payload.Items, "")
+	if err != nil {
+		t.Fatal(err)
+	}
+	var rendered []LLMValidationItem
+	if err := json.Unmarshal([]byte(variables["PayloadJSON"]), &rendered); err != nil {
+		t.Fatal(err)
+	}
+	if len(rendered[0].NeighboringSegments) != 2 || rendered[0].GlossaryEntries[0].Summary != "A traveling troupe" {
+		t.Fatalf("rendered context: %+v", rendered)
+	}
+	// The context-free endpoint must not reach beyond the current chunk.
+	req.WorkingTranscript.Segments = req.WorkingTranscript.Segments[1:2]
+	req.Glossary = nil
+	isolated, _ := BuildLLMValidationRequest("editorial_review", LLMValidatorTypeEditorialReview, req)
+	if len(isolated.Items[0].NeighboringSegments) != 0 || len(isolated.Items[0].GlossaryEntries) != 0 {
+		t.Fatalf("unexpected external context: %+v", isolated)
+	}
+}
+
+func TestValidationContextFitsBudgetWithoutDiscardingCandidate(t *testing.T) {
+	estimator := chunking.NewSimpleTokenEstimator()
+	base := LLMValidationItem{CorrectionIndex: 7, SegmentID: 1, OriginalText: "teh", CorrectedText: "the", OriginalSegmentText: "teh cat", CorrectedSegmentText: "the cat"}
+	required, err := estimateBatchTokens(estimator, []LLMValidationItem{base})
+	if err != nil {
+		t.Fatal(err)
+	}
+	enriched := base
+	enriched.NeighboringSegments = []schema.Segment{{ID: 2, Text: strings.Repeat("long context ", 300)}}
+	enriched.GlossaryEntries = []schema.GlossaryEntry{{Name: "cat", Summary: strings.Repeat("long definition ", 300)}}
+	fitted := fitValidationContext(enriched, required, estimator)
+	tokens, err := estimateBatchTokens(estimator, []LLMValidationItem{fitted})
+	if err != nil || tokens > required || fitted.OriginalSegmentText != base.OriginalSegmentText || fitted.CorrectedSegmentText != base.CorrectedSegmentText {
+		t.Fatalf("fitted: %+v tokens=%d err=%v", fitted, tokens, err)
+	}
+	full, _ := estimateBatchTokens(estimator, []LLMValidationItem{enriched})
+	batches, err := ChunkLLMValidationItems([]LLMValidationItem{enriched, enriched}, full, estimator)
+	if err != nil || len(batches) != 2 {
+		t.Fatalf("context must count toward batch budget: %+v %v", batches, err)
+	}
+}