notarius/internal/modules/dnd/normalize/itemregistry/normalizer_test.go

package itemregistry

import (
	"context"
	"encoding/json"
	"errors"
	"reflect"
	"strconv"
	"strings"
	"testing"
	"time"

	"gitea.maximumdirect.net/eric/notarius/internal/core/source"
	"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
	"gitea.maximumdirect.net/eric/notarius/internal/framework/llm"
	"gitea.maximumdirect.net/eric/notarius/internal/framework/pipeline"
	"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd"
	"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/items/identity"
	"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared/diagnostics"
	"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared/entityreconcile"
	identityvalidator "gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/validate/itemregistry/identity"
	"gitea.maximumdirect.net/eric/promptkit"
)

func TestModuleContractAndMetadata(t *testing.T) {
	want := pipeline.ModuleSpec{Key: Key, Stage: pipeline.StageNormalize, ExecutionClass: contracts.ExecutionClassLLMBacked, Requires: []string{"merged"}, Provides: []string{"normalized"}, ArtifactKind: dnd.ItemRegistryKind}
	if got := ModuleSpec(); !reflect.DeepEqual(got, want) {
		t.Fatalf("ModuleSpec() = %#v, want %#v", got, want)
	}
	registry := pipeline.NewNormalizerRegistry()
	if err := Register(registry); err != nil {
		t.Fatal(err)
	}
	if _, err := New(nil, Options{}); err == nil {
		t.Fatal("New() accepted nil client")
	}
	metadata := newNormalizer(t, &recordingNormalizerClient{}).ManifestMetadata()
	if metadata["identity_policy"] != identity.Policy || metadata["response_schema_id"] != entityreconcile.ResponseSchemaID || metadata["normalization_policy"] != normalizationPolicy || metadata["semantic_context_radius"] != semanticContextRadius {
		t.Fatalf("metadata = %#v", metadata)
	}
}

func TestNormalizeConsolidatesEqualNamesAcrossEvidenceWithoutMutation(t *testing.T) {
	doc := &source.SourceDocument{ID: "session", Units: []source.SourceUnit{
		{ID: 1, Text: "The rope is secured."},
		{ID: 2, Text: "The party takes the rope."},
		{ID: 3, Text: "The rope is packed away."},
		{ID: 4, Text: "A lantern lights the path."},
	}}
	input := dnd.ItemRegistry{Items: []dnd.Item{
		{Name: "  Rope  ", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 3, EndUnitID: 3}, {SourceID: "session", StartUnitID: 1, EndUnitID: 1}}},
		{Name: "rope", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 2, EndUnitID: 2}, {SourceID: "session", StartUnitID: 3, EndUnitID: 3}}},
		{Name: "Lantern", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 4, EndUnitID: 4}}},
	}}
	before := dnd.ItemRegistry{Items: []dnd.Item{
		{Name: "  Rope  ", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 3, EndUnitID: 3}, {SourceID: "session", StartUnitID: 1, EndUnitID: 1}}},
		{Name: "rope", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 2, EndUnitID: 2}, {SourceID: "session", StartUnitID: 3, EndUnitID: 3}}},
		{Name: "Lantern", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 4, EndUnitID: 4}}},
	}}
	client := &recordingNormalizerClient{}
	result, err := newNormalizer(t, client).Normalize(context.Background(), normalizeRequestWithSource(input, doc))
	if err != nil || len(result.Value.Items) != 2 || !reflect.DeepEqual(input, before) || len(client.requests) != 1 {
		t.Fatalf("Normalize() = %#v, %v; want deterministic non-mutating consolidation", result, err)
	}
	rope := result.Value.Items[0]
	wantRefs := []source.SourceRef{{SourceID: "session", StartUnitID: 1, EndUnitID: 1}, {SourceID: "session", StartUnitID: 2, EndUnitID: 2}, {SourceID: "session", StartUnitID: 3, EndUnitID: 3}}
	if rope.Name != "Rope" || rope.ID != identity.DeriveID("Rope") || !reflect.DeepEqual(rope.SourceRefs, wantRefs) || result.Value.Items[1].Name != "Lantern" || !hasWarning(result.Warnings, ReasonCodeDuplicateItemCollapsed) {
		t.Fatalf("items = %#v, warnings = %#v; want earliest display name, canonical evidence union, and stable placement", result.Value.Items, result.Warnings)
	}
	validation, validationErr := identityvalidator.New(identityvalidator.Options{}).Validate(context.Background(), contracts.TypedValidationRequest[dnd.ItemRegistry]{Value: result.Value})
	if validationErr != nil || !validation.Approved {
		t.Fatalf("default item-registry identity validation = %#v, %v; want normalized cross-chunk candidates accepted", validation, validationErr)
	}
	var candidates struct {
		Candidates []struct {
			Name string `json:"name"`
		} `json:"candidates"`
	}
	if err := json.Unmarshal(client.requests[0].Inputs["candidates"].Content, &candidates); err != nil || len(candidates.Candidates) != 2 || candidates.Candidates[0].Name != "Rope" || candidates.Candidates[1].Name != "Lantern" {
		t.Fatalf("semantic candidates = %#v, %v; want one candidate per comparison name", candidates, err)
	}
	repeated, repeatErr := newNormalizer(t, &recordingNormalizerClient{}).Normalize(context.Background(), normalizeRequestWithSource(result.Value, doc))
	if repeatErr != nil || !reflect.DeepEqual(repeated.Value, result.Value) {
		t.Fatalf("repeated normalization = %#v, %v; want stable output", repeated, repeatErr)
	}
}

func TestNormalizeAppliesSafeAliasProposal(t *testing.T) {
	doc := semanticDocument()
	input := dnd.ItemRegistry{Items: []dnd.Item{
		{Name: "Star Compass", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 10, EndUnitID: 10}}},
		{Name: "Compass of the Stars", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 20, EndUnitID: 20}}},
		{Name: "Gold Pieces", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 30, EndUnitID: 30}}},
	}}
	client := &recordingNormalizerClient{response: `{"duplicate_groups":[{"members":["candidate-000001","candidate-000002"],"canonical":"candidate-000002"}]}`}
	result, err := newNormalizer(t, client).Normalize(context.Background(), normalizeRequestWithSource(input, doc))
	if err != nil || result.Retry != nil || len(result.Value.Items) != 2 {
		t.Fatalf("Normalize() = %#v, %v", result, err)
	}
	merged := result.Value.Items[0]
	if merged.Name != "Compass of the Stars" || merged.ID != identity.DeriveID(merged.Name) || len(merged.SourceRefs) != 2 || !hasWarning(result.Warnings, ReasonCodeDuplicateItemCollapsed) {
		t.Fatalf("merged item = %#v, warnings = %#v", merged, result.Warnings)
	}
	encoded := string(client.requests[0].Inputs["candidates"].Content) + string(client.requests[0].Inputs["transcript"].Content)
	if strings.Contains(encoded, doc.ID) || strings.Contains(encoded, "candidate-") || strings.Contains(encoded, merged.ID) || !strings.Contains(encoded, `"source_refs"`) {
		t.Fatalf("private inputs = %s", encoded)
	}
}

func TestNormalizeAppliesCurrencyReconciliationSafely(t *testing.T) {
	doc := semanticDocument()
	ref := func(unitID int) []source.SourceRef {
		return []source.SourceRef{{SourceID: doc.ID, StartUnitID: unitID, EndUnitID: unitID}}
	}
	for _, test := range []struct {
		name          string
		items         []dnd.Item
		response      string
		wantNames     []string
		wantRefCounts []int
		wantRetry     bool
		warning       string
	}{
		{
			name: "same denomination aliases",
			items: []dnd.Item{
				{Name: "GP", SourceRefs: ref(10)},
				{Name: "Gold Piece", SourceRefs: ref(20)},
				{Name: "Gold Pieces", SourceRefs: ref(30)},
			},
			response:      `{"duplicate_groups":[{"members":["candidate-000001","candidate-000002","candidate-000003"],"canonical":"candidate-000002"}]}`,
			wantNames:     []string{"Gold Piece"},
			wantRefCounts: []int{3},
			warning:       ReasonCodeDuplicateItemCollapsed,
		},
		{
			name: "different denominations",
			items: []dnd.Item{
				{Name: "Gold Pieces", SourceRefs: ref(10)},
				{Name: "Silver Pieces", SourceRefs: ref(20)},
			},
			response:      `{"duplicate_groups":[{"members":["candidate-000001","candidate-000002"],"canonical":"candidate-000001"}]}`,
			wantNames:     []string{"Gold Pieces", "Silver Pieces"},
			wantRefCounts: []int{1, 1},
			wantRetry:     true,
			warning:       ReasonCodeItemSemanticProposalInvalid,
		},
		{
			name: "currency plus ordinary item",
			items: []dnd.Item{
				{Name: "Gold Pieces", SourceRefs: ref(10)},
				{Name: "Longsword", SourceRefs: ref(20)},
			},
			response:      `{"duplicate_groups":[{"members":["candidate-000001","candidate-000002"],"canonical":"candidate-000001"}]}`,
			wantNames:     []string{"Gold Pieces", "Longsword"},
			wantRefCounts: []int{1, 1},
			wantRetry:     true,
			warning:       ReasonCodeItemSemanticProposalInvalid,
		},
		{
			name: "ordinary items",
			items: []dnd.Item{
				{Name: "Star Compass", SourceRefs: ref(10)},
				{Name: "Compass of the Stars", SourceRefs: ref(20)},
			},
			response:      `{"duplicate_groups":[{"members":["candidate-000001","candidate-000002"],"canonical":"candidate-000002"}]}`,
			wantNames:     []string{"Compass of the Stars"},
			wantRefCounts: []int{2},
			warning:       ReasonCodeDuplicateItemCollapsed,
		},
		{
			name: "currency plus ordinary canonical item",
			items: []dnd.Item{
				{Name: "Gold Pieces", SourceRefs: ref(10)},
				{Name: "Longsword", SourceRefs: ref(20)},
			},
			response:      `{"duplicate_groups":[{"members":["candidate-000001","candidate-000002"],"canonical":"candidate-000002"}]}`,
			wantNames:     []string{"Gold Pieces", "Longsword"},
			wantRefCounts: []int{1, 1},
			wantRetry:     true,
			warning:       ReasonCodeItemSemanticProposalInvalid,
		},
	} {
		t.Run(test.name, func(t *testing.T) {
			input := dnd.ItemRegistry{Items: make([]dnd.Item, len(test.items))}
			for index, item := range test.items {
				input.Items[index] = cloneItem(item)
			}
			before := dnd.ItemRegistry{Items: make([]dnd.Item, len(input.Items))}
			for index, item := range input.Items {
				before.Items[index] = cloneItem(item)
			}

			result, err := newNormalizer(t, &recordingNormalizerClient{response: test.response}).Normalize(context.Background(), normalizeRequestWithSource(input, doc))
			if err != nil || (result.Retry != nil) != test.wantRetry || !reflect.DeepEqual(input, before) || !hasWarning(result.Warnings, test.warning) {
				t.Fatalf("Normalize() = %#v, %v", result, err)
			}
			if len(result.Value.Items) != len(test.wantNames) {
				t.Fatalf("items = %#v, want names %#v", result.Value.Items, test.wantNames)
			}
			for index, item := range result.Value.Items {
				if item.Name != test.wantNames[index] || len(item.SourceRefs) != test.wantRefCounts[index] {
					t.Fatalf("item %d = %#v, want name %q with %d source refs", index, item, test.wantNames[index], test.wantRefCounts[index])
				}
			}
			if test.wantRetry && (len(result.Retry.FallbackWarnings) != 1 || result.Retry.FallbackWarnings[0].ReasonCode != ReasonCodeItemSemanticReconciliationExhausted) {
				t.Fatalf("retry = %#v, want preserved-group fallback", result.Retry)
			}
		})
	}
}

func TestNormalizePreservesCandidatesForUnsafeProposalGroups(t *testing.T) {
	doc := semanticDocument()
	input := dnd.ItemRegistry{Items: []dnd.Item{
		{Name: "Star Compass", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 10, EndUnitID: 10}}},
		{Name: "Compass", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 20, EndUnitID: 20}}},
		{Name: "Rope", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 30, EndUnitID: 30}}},
	}}
	client := &recordingNormalizerClient{response: `{"duplicate_groups":[{"members":["candidate-000001","candidate-000002"],"canonical":"candidate-000001"},{"members":["candidate-000002","candidate-000003"],"canonical":"candidate-000003"},{"members":["candidate-000001","candidate-000003"],"canonical":"candidate-000099"}]}`}
	result, err := newNormalizer(t, client).Normalize(context.Background(), normalizeRequestWithSource(input, doc))
	if err != nil || result.Retry == nil || len(result.Value.Items) != 3 || !strings.Contains(result.Retry.Message, "overlapping_member") || !strings.Contains(result.Retry.Message, "canonical_unknown") || strings.Contains(result.Retry.Message, "Star Compass") || len(result.Retry.Message) > 4096 {
		t.Fatalf("Normalize() = %#v, %v; want deterministic retry fallback", result, err)
	}
}

func TestNormalizeRetryFallbackErrorsWarningsAndIdempotence(t *testing.T) {
	doc := semanticDocument()
	input := dnd.ItemRegistry{Items: []dnd.Item{{Name: "Star Compass", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 10, EndUnitID: 10}}}, {Name: "Compass", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 20, EndUnitID: 20}}}}}
	invalid, err := newNormalizer(t, &recordingNormalizerClient{err: contracts.ErrInvalidStructuredOutput}).Normalize(context.Background(), normalizeRequestWithSource(input, doc))
	if err != nil || invalid.Retry == nil || invalid.Retry.ReasonCode != ReasonCodeItemSemanticProposalInvalid {
		t.Fatalf("invalid result = %#v, %v", invalid, err)
	}
	_, err = newNormalizer(t, &recordingNormalizerClient{err: errors.New("provider unavailable")}).Normalize(context.Background(), normalizeRequestWithSource(input, doc))
	if err == nil || !strings.Contains(err.Error(), "provider unavailable") {
		t.Fatalf("provider error = %v", err)
	}
	if bounded := limitWarningsForRetry(make([]contracts.Warning, diagnostics.MaxWarnings+5)); len(bounded) != diagnostics.MaxWarnings-1 || bounded[len(bounded)-1].ReasonCode != ReasonCodeItemNormalizationWarningsOmitted {
		t.Fatalf("retry warning limit = %#v", bounded)
	}
	first, err := newNormalizer(t, &recordingNormalizerClient{}).Normalize(context.Background(), normalizeRequestWithSource(input, doc))
	second, secondErr := newNormalizer(t, &recordingNormalizerClient{}).Normalize(context.Background(), normalizeRequestWithSource(first.Value, doc))
	if err != nil || secondErr != nil || !reflect.DeepEqual(first.Value, second.Value) {
		t.Fatalf("normalization must be idempotent: first=%#v err=%v second=%#v err=%v", first, err, second, secondErr)
	}
}

func TestRegisterPromptAssetsPreparesItemNormalizationPrompt(t *testing.T) {
	registry := llm.NewAssetRegistry()
	if err := entityreconcile.RegisterSchemaAssets(registry); err != nil {
		t.Fatal(err)
	}
	if err := RegisterPromptAssets(registry); err != nil {
		t.Fatal(err)
	}
	options, err := registry.PromptKitOptions()
	if err != nil {
		t.Fatal(err)
	}
	options = append(options, promptkit.WithProfiles(promptkit.OpenAICompatibleProfile(promptkit.OpenAICompatibleProfileConfig{ID: "item-normalize-test", Endpoint: "http://127.0.0.1:1/v1", Model: "test"})))
	engine, err := promptkit.NewEngine(promptkit.Config{Timeout: time.Second}, options...)
	if err != nil {
		t.Fatal(err)
	}
	prepared, err := engine.Prepare(context.Background(), promptkit.RunRequest{PromptID: PromptID, PromptVersion: entityreconcile.SchemaVersion, ProfileID: "item-normalize-test", Inputs: map[string]promptkit.ArtifactRef{"candidates": promptkit.Inline(`{"candidates":[{"name":"Rope","source_refs":[{"start_unit_id":1,"end_unit_id":1}]}]}`), "transcript": promptkit.Inline(`{"windows":[{"units":[]}]}`)}})
	if err != nil {
		t.Fatal(err)
	}
	if prepared.OutputContract.SchemaPath != "dnd_entity_reconcile_llm.v1.json" || !strings.Contains(prepared.Messages[1].Content, "currency denominations") || !strings.Contains(prepared.Messages[1].Content, "materially different item types") {
		t.Fatalf("prepared prompt = %#v", prepared)
	}
}

type recordingNormalizerClient struct {
	response string
	err      error
	requests []contracts.StructuredCompletionRequest
}

func (c *recordingNormalizerClient) CompleteStructured(_ context.Context, request contracts.StructuredCompletionRequest, output any) (contracts.StructuredCompletionResponse, error) {
	c.requests = append(c.requests, request)
	if c.err != nil {
		return contracts.StructuredCompletionResponse{}, c.err
	}
	response := c.response
	if response == "" {
		response = `{"duplicate_groups":[]}`
	}
	content, err := contextualProposalResponse(response, request.Inputs["candidates"].Content)
	if err != nil {
		return contracts.StructuredCompletionResponse{}, err
	}
	if err := json.Unmarshal(content, output); err != nil {
		return contracts.StructuredCompletionResponse{}, err
	}
	return contracts.StructuredCompletionResponse{Content: content}, nil
}

func contextualProposalResponse(response string, candidateContent []byte) ([]byte, error) {
	if !strings.Contains(response, "candidate-") {
		return []byte(response), nil
	}
	var selection struct {
		DuplicateGroups []struct {
			Members   []string `json:"members"`
			Canonical string   `json:"canonical"`
		} `json:"duplicate_groups"`
	}
	if err := json.Unmarshal([]byte(response), &selection); err != nil {
		return nil, err
	}
	var candidates struct {
		Candidates []entityreconcile.Selector `json:"candidates"`
	}
	if err := json.Unmarshal(candidateContent, &candidates); err != nil {
		return nil, err
	}
	selector := func(key string) entityreconcile.Selector {
		index, err := strconv.Atoi(strings.TrimPrefix(key, "candidate-"))
		if err != nil || index < 1 || index > len(candidates.Candidates) {
			return entityreconcile.Selector{Name: key, SourceRefs: []entityreconcile.SourceRange{}}
		}
		return candidates.Candidates[index-1].Clone()
	}
	proposal := entityreconcile.ProposalResponse{DuplicateGroups: make([]entityreconcile.DuplicateGroup, len(selection.DuplicateGroups))}
	for index, group := range selection.DuplicateGroups {
		members := make([]entityreconcile.Selector, len(group.Members))
		for memberIndex, key := range group.Members {
			members[memberIndex] = selector(key)
		}
		proposal.DuplicateGroups[index] = entityreconcile.DuplicateGroup{Members: members, Canonical: selector(group.Canonical)}
	}
	return json.Marshal(proposal)
}

func newNormalizer(t *testing.T, client contracts.StructuredLLMClient) *Normalizer {
	t.Helper()
	normalizer, err := New(client, Options{})
	if err != nil {
		t.Fatalf("New() error = %v", err)
	}
	return normalizer
}
func normalizeRequest(value dnd.ItemRegistry) contracts.TypedNormalizeRequest[dnd.ItemRegistry] {
	return contracts.TypedNormalizeRequest[dnd.ItemRegistry]{MergeOutput: contracts.MergeArtifact[dnd.ItemRegistry]{Value: value}}
}
func normalizeRequestWithSource(value dnd.ItemRegistry, doc *source.SourceDocument) contracts.TypedNormalizeRequest[dnd.ItemRegistry] {
	request := normalizeRequest(value)
	request.Source = doc
	return request
}
func semanticDocument() *source.SourceDocument {
	return &source.SourceDocument{ID: "item-session", Units: []source.SourceUnit{{ID: 10, Text: "The Star Compass points north."}, {ID: 20, Text: "The compass of the stars glows."}, {ID: 30, Text: "The chest holds gold pieces."}}}
}
func hasWarning(warnings []contracts.Warning, reason string) bool {
	for _, warning := range warnings {
		if warning.ReasonCode == reason {
			return true
		}
	}
	return false
}