Register item registry normalization
This commit is contained in:
350
internal/modules/dnd/normalize/itemregistry/normalizer.go
Normal file
350
internal/modules/dnd/normalize/itemregistry/normalizer.go
Normal file
@@ -0,0 +1,350 @@
|
||||
// Package itemregistry normalizes merged D&D item-registry candidates conservatively.
|
||||
package itemregistry
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"fmt"
|
||||
"reflect"
|
||||
"sort"
|
||||
"strconv"
|
||||
"strings"
|
||||
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/core/source"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
|
||||
"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"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared/diagnostics"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared/entityreconcile"
|
||||
)
|
||||
|
||||
const (
|
||||
Key = "dnd/item-registry"
|
||||
PromptID = "dnd.item_registry.normalize"
|
||||
normalizationPolicy = "dnd.item_registry.normalize.v1"
|
||||
semanticContextPolicy = "dnd.entity_reconcile.context.v1"
|
||||
semanticContextRadius = 2
|
||||
NormalizationPolicy = normalizationPolicy
|
||||
|
||||
ReasonCodeItemFieldsNormalized = "item_fields_normalized"
|
||||
ReasonCodeItemIDRecomputed = "item_id_recomputed"
|
||||
ReasonCodeSourceReferencesNormalized = "source_references_normalized"
|
||||
ReasonCodeDuplicateItemCollapsed = "duplicate_item_collapsed"
|
||||
ReasonCodeItemSemanticProposalInvalid = "item_semantic_proposal_invalid"
|
||||
ReasonCodeItemSemanticReconciliationExhausted = "item_semantic_reconciliation_exhausted"
|
||||
ReasonCodeItemNormalizationWarningsOmitted = "item_normalization_warnings_omitted"
|
||||
)
|
||||
|
||||
var requiredCapabilities = []string{"merged"}
|
||||
var providedCapabilities = []string{"normalized"}
|
||||
|
||||
var _ contracts.Normalizer[dnd.ItemRegistry] = (*Normalizer)(nil)
|
||||
var _ contracts.ManifestMetadataProvider = (*Normalizer)(nil)
|
||||
var _ pipeline.CheckpointFingerprintProvider = (*Normalizer)(nil)
|
||||
|
||||
type Options struct{}
|
||||
|
||||
type Normalizer struct {
|
||||
llm contracts.StructuredLLMClient
|
||||
promptSHA string
|
||||
responseSchemaSHA string
|
||||
}
|
||||
|
||||
func New(llmClient contracts.StructuredLLMClient, _ Options) (*Normalizer, error) {
|
||||
if llmClient == nil {
|
||||
return nil, normalizerErrorf("LLM client must not be nil")
|
||||
}
|
||||
promptSHA, err := promptAssetMetadata()
|
||||
if err != nil {
|
||||
return nil, normalizerErrorf("load prompt metadata: %w", err)
|
||||
}
|
||||
responseSchema, err := entityreconcile.LoadResponseSchema()
|
||||
if err != nil {
|
||||
return nil, normalizerErrorf("load response schema: %w", err)
|
||||
}
|
||||
return &Normalizer{llm: llmClient, promptSHA: promptSHA, responseSchemaSHA: responseSchema.SHA256}, nil
|
||||
}
|
||||
|
||||
func (n *Normalizer) Key() string { return Key }
|
||||
func (n *Normalizer) ReferenceSlots() []contracts.ReferenceSlot { return nil }
|
||||
|
||||
func (n *Normalizer) ManifestMetadata() map[string]any {
|
||||
if n == nil {
|
||||
return nil
|
||||
}
|
||||
return map[string]any{
|
||||
"prompt_id": PromptID, "prompt_version": entityreconcile.SchemaVersion, "prompt_sha256": n.promptSHA,
|
||||
"response_schema_key": string(entityreconcile.ResponseSchemaKey), "response_schema_id": entityreconcile.ResponseSchemaID,
|
||||
"response_schema_name": entityreconcile.ResponseSchemaName, "response_schema_version": entityreconcile.SchemaVersion,
|
||||
"response_schema_sha256": n.responseSchemaSHA, "identity_policy": identity.Policy,
|
||||
"normalization_policy": normalizationPolicy, "semantic_context_policy": semanticContextPolicy, "semantic_context_radius": semanticContextRadius,
|
||||
}
|
||||
}
|
||||
|
||||
func (n *Normalizer) CheckpointFingerprints() []pipeline.CheckpointFingerprint {
|
||||
if n == nil {
|
||||
return nil
|
||||
}
|
||||
return []pipeline.CheckpointFingerprint{
|
||||
{Name: "prompt", Value: n.promptSHA}, {Name: "response_schema", Value: n.responseSchemaSHA},
|
||||
{Name: "identity_policy", Value: identity.Policy}, {Name: "normalization_policy", Value: normalizationPolicy},
|
||||
{Name: "semantic_context_policy", Value: fmt.Sprintf("%s:%d", semanticContextPolicy, semanticContextRadius)},
|
||||
}
|
||||
}
|
||||
|
||||
func (n *Normalizer) Normalize(ctx context.Context, req contracts.TypedNormalizeRequest[dnd.ItemRegistry]) (contracts.TypedNormalizeResult[dnd.ItemRegistry], error) {
|
||||
if n == nil {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{}, normalizerErrorf("normalizer must not be nil")
|
||||
}
|
||||
if n.llm == nil {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{}, normalizerErrorf("LLM client must not be nil")
|
||||
}
|
||||
if ctx == nil {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{}, normalizerErrorf("context must not be nil")
|
||||
}
|
||||
if err := ctx.Err(); err != nil {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{}, normalizerErrorf("context error before normalize: %w", err)
|
||||
}
|
||||
|
||||
order := shared.NewSourceRefOrder(req.Source)
|
||||
records, warnings := preprocessRecords(req.MergeOutput.Value, order)
|
||||
deterministic := recordList(records)
|
||||
materials, ready, err := entityreconcile.BuildContext(req.Source, reconciliationCandidates(records), semanticContextRadius)
|
||||
if err != nil {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{}, normalizerErrorf("build semantic context: %w", err)
|
||||
}
|
||||
if !ready {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{Value: deterministic, Warnings: limitWarnings(warnings)}, nil
|
||||
}
|
||||
|
||||
var response entityreconcile.ProposalResponse
|
||||
if _, err := n.llm.CompleteStructured(ctx, contracts.StructuredCompletionRequest{
|
||||
StageName: Key, PromptID: PromptID, PromptVersion: entityreconcile.SchemaVersion,
|
||||
ProfileID: req.LLMProfile, SessionID: req.SessionID,
|
||||
Inputs: contracts.LLMInputSet{"candidates": materials.Candidates, "transcript": materials.Transcript},
|
||||
}, &response); err != nil {
|
||||
if errors.Is(err, contracts.ErrInvalidStructuredOutput) {
|
||||
return n.invalidStructuredResult(deterministic, warnings), nil
|
||||
}
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{}, normalizerErrorf("complete structured output: %w", err)
|
||||
}
|
||||
assessment := materials.Assess(response)
|
||||
applied, semanticWarnings, rejectedGroups := applySafeGroups(records, reconciliationGroups(assessment, materials.CandidateKeys()), order)
|
||||
warnings = append(warnings, semanticWarnings...)
|
||||
if rejectedGroups > 0 {
|
||||
return currencyRetryResult(recordList(applied), warnings, rejectedGroups), nil
|
||||
}
|
||||
if assessment.DiscardedGroups() == 0 {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{Value: recordList(applied), Warnings: limitWarnings(warnings)}, nil
|
||||
}
|
||||
return retryResult(recordList(applied), warnings, assessment), nil
|
||||
}
|
||||
|
||||
func (n *Normalizer) invalidStructuredResult(value dnd.ItemRegistry, warnings []contracts.Warning) contracts.TypedNormalizeResult[dnd.ItemRegistry] {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{Value: value, Warnings: limitWarningsForRetry(warnings), Retry: &contracts.NormalizeRetry{
|
||||
ReasonCode: ReasonCodeItemSemanticProposalInvalid, Message: "semantic proposal requires retry: invalid structured output",
|
||||
FallbackWarnings: []contracts.Warning{semanticFallbackWarning(-1)},
|
||||
}}
|
||||
}
|
||||
|
||||
func retryResult(value dnd.ItemRegistry, warnings []contracts.Warning, assessment entityreconcile.Assessment) contracts.TypedNormalizeResult[dnd.ItemRegistry] {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{Value: value, Warnings: limitWarningsForRetry(warnings), Retry: &contracts.NormalizeRetry{
|
||||
ReasonCode: ReasonCodeItemSemanticProposalInvalid, Message: diagnostics.Aggregate("semantic proposal requires retry", reconciliationIssues(assessment)),
|
||||
FallbackWarnings: []contracts.Warning{semanticFallbackWarning(assessment.DiscardedGroups())},
|
||||
}}
|
||||
}
|
||||
|
||||
func currencyRetryResult(value dnd.ItemRegistry, warnings []contracts.Warning, rejectedGroups int) contracts.TypedNormalizeResult[dnd.ItemRegistry] {
|
||||
return contracts.TypedNormalizeResult[dnd.ItemRegistry]{Value: value, Warnings: limitWarningsForRetry(warnings), Retry: &contracts.NormalizeRetry{
|
||||
ReasonCode: ReasonCodeItemSemanticProposalInvalid, Message: "semantic proposal requires retry: currency denominations must remain distinct",
|
||||
FallbackWarnings: []contracts.Warning{semanticFallbackWarning(rejectedGroups)},
|
||||
}}
|
||||
}
|
||||
|
||||
func semanticFallbackWarning(discarded int) contracts.Warning {
|
||||
message := "semantic proposal could not be applied"
|
||||
if discarded >= 0 {
|
||||
message = fmt.Sprintf("%d proposal group(s) omitted after semantic proposal retry exhaustion", discarded)
|
||||
}
|
||||
return contracts.Warning{Scope: "items", ReasonCode: ReasonCodeItemSemanticReconciliationExhausted, Message: message}
|
||||
}
|
||||
|
||||
func limitWarnings(warnings []contracts.Warning) []contracts.Warning {
|
||||
return diagnostics.LimitWarnings(warnings, "items", ReasonCodeItemNormalizationWarningsOmitted)
|
||||
}
|
||||
|
||||
func limitWarningsForRetry(warnings []contracts.Warning) []contracts.Warning {
|
||||
if warnings == nil {
|
||||
return nil
|
||||
}
|
||||
if len(warnings) < diagnostics.MaxWarnings {
|
||||
return append([]contracts.Warning(nil), warnings...)
|
||||
}
|
||||
displayed := diagnostics.MaxWarnings - 2
|
||||
bounded := append([]contracts.Warning(nil), warnings[:displayed]...)
|
||||
return append(bounded, contracts.Warning{Scope: "items", ReasonCode: ReasonCodeItemNormalizationWarningsOmitted, Message: fmt.Sprintf("%d additional warning(s) omitted", len(warnings)-displayed)})
|
||||
}
|
||||
|
||||
type normalizedRecord struct {
|
||||
item dnd.Item
|
||||
inputIndexes []int
|
||||
earliest int
|
||||
}
|
||||
|
||||
func preprocessRecords(input dnd.ItemRegistry, order shared.SourceRefOrder) ([]normalizedRecord, []contracts.Warning) {
|
||||
if input.Items == nil {
|
||||
return nil, nil
|
||||
}
|
||||
records := make([]normalizedRecord, len(input.Items))
|
||||
warnings := make([]contracts.Warning, 0)
|
||||
for index, inputItem := range input.Items {
|
||||
item, fieldsChanged, refsChanged := normalizeRecord(inputItem, order)
|
||||
records[index] = normalizedRecord{item: item, inputIndexes: []int{index}, earliest: index}
|
||||
if fieldsChanged {
|
||||
warnings = append(warnings, contracts.Warning{Scope: itemScope(index), ReasonCode: ReasonCodeItemFieldsNormalized, Message: fmt.Sprintf("input index %d: item name normalized for %s", index, diagnostics.Quote(inputItem.Name))})
|
||||
}
|
||||
if refsChanged {
|
||||
warnings = append(warnings, contracts.Warning{Scope: itemScope(index), ReasonCode: ReasonCodeSourceReferencesNormalized, Message: fmt.Sprintf("input index %d: source references normalized (original count %d, final count %d)", index, len(inputItem.SourceRefs), len(item.SourceRefs))})
|
||||
}
|
||||
if inputItem.ID != item.ID {
|
||||
warnings = append(warnings, contracts.Warning{Scope: itemScope(index), ReasonCode: ReasonCodeItemIDRecomputed, Message: fmt.Sprintf("input index %d: item ID recomputed from %s", index, diagnostics.Quote(item.Name))})
|
||||
}
|
||||
}
|
||||
groups := exactDuplicateGroups(records)
|
||||
output := make([]normalizedRecord, 0, len(groups))
|
||||
for _, members := range groups {
|
||||
retained := cloneRecord(records[members[0]])
|
||||
for _, member := range members[1:] {
|
||||
retained.inputIndexes = append(retained.inputIndexes, records[member].inputIndexes...)
|
||||
}
|
||||
retained.inputIndexes = sortedUniqueIndexes(retained.inputIndexes)
|
||||
output = append(output, retained)
|
||||
if len(members) > 1 {
|
||||
warnings = append(warnings, duplicateWarning(retained.earliest, memberInputIndexes(records, members[1:])))
|
||||
}
|
||||
}
|
||||
return output, warnings
|
||||
}
|
||||
|
||||
func normalizeRecord(input dnd.Item, order shared.SourceRefOrder) (dnd.Item, bool, bool) {
|
||||
output := cloneItem(input)
|
||||
output.Name = identity.NormalizeDisplay(input.Name)
|
||||
output.SourceRefs = order.Canonicalize(input.SourceRefs)
|
||||
output.ID = identity.DeriveID(output.Name)
|
||||
return output, input.Name != output.Name, !reflect.DeepEqual(input.SourceRefs, output.SourceRefs)
|
||||
}
|
||||
|
||||
func exactDuplicateGroups(records []normalizedRecord) [][]int {
|
||||
groups := make([][]int, 0, len(records))
|
||||
for index, record := range records {
|
||||
key := identity.ComparisonKey(record.item.Name)
|
||||
found := false
|
||||
for groupIndex, members := range groups {
|
||||
first := records[members[0]]
|
||||
if identity.ComparisonKey(first.item.Name) == key && reflect.DeepEqual(first.item.SourceRefs, record.item.SourceRefs) {
|
||||
groups[groupIndex] = append(groups[groupIndex], index)
|
||||
found = true
|
||||
break
|
||||
}
|
||||
}
|
||||
if !found {
|
||||
groups = append(groups, []int{index})
|
||||
}
|
||||
}
|
||||
return groups
|
||||
}
|
||||
|
||||
func cloneItem(input dnd.Item) dnd.Item {
|
||||
input.SourceRefs = cloneSourceRefs(input.SourceRefs)
|
||||
return input
|
||||
}
|
||||
func cloneRecord(input normalizedRecord) normalizedRecord {
|
||||
input.item = cloneItem(input.item)
|
||||
input.inputIndexes = append([]int(nil), input.inputIndexes...)
|
||||
return input
|
||||
}
|
||||
func cloneSourceRefs(input []source.SourceRef) []source.SourceRef {
|
||||
return append([]source.SourceRef(nil), input...)
|
||||
}
|
||||
|
||||
func sortedUniqueIndexes(indexes []int) []int {
|
||||
if len(indexes) == 0 {
|
||||
return nil
|
||||
}
|
||||
out := append([]int(nil), indexes...)
|
||||
sort.Ints(out)
|
||||
write := 1
|
||||
for _, index := range out[1:] {
|
||||
if index != out[write-1] {
|
||||
out[write] = index
|
||||
write++
|
||||
}
|
||||
}
|
||||
return out[:write]
|
||||
}
|
||||
func memberInputIndexes(records []normalizedRecord, members []int) []int {
|
||||
indexes := make([]int, 0, len(members))
|
||||
for _, member := range members {
|
||||
indexes = append(indexes, records[member].inputIndexes...)
|
||||
}
|
||||
return sortedUniqueIndexes(indexes)
|
||||
}
|
||||
func recordValues(records []normalizedRecord) []dnd.Item {
|
||||
if records == nil {
|
||||
return nil
|
||||
}
|
||||
values := make([]dnd.Item, len(records))
|
||||
for index, record := range records {
|
||||
values[index] = cloneItem(record.item)
|
||||
}
|
||||
return values
|
||||
}
|
||||
func recordList(records []normalizedRecord) dnd.ItemRegistry {
|
||||
if records == nil {
|
||||
return dnd.ItemRegistry{}
|
||||
}
|
||||
return dnd.ItemRegistry{Items: recordValues(records)}
|
||||
}
|
||||
|
||||
func duplicateWarning(retainedIndex int, removed []int) contracts.Warning {
|
||||
const maxDisplayedIndices = 20
|
||||
displayed := removed
|
||||
if len(displayed) > maxDisplayedIndices {
|
||||
displayed = displayed[:maxDisplayedIndices]
|
||||
}
|
||||
indices := make([]string, len(displayed))
|
||||
for index, removedIndex := range displayed {
|
||||
indices[index] = strconv.Itoa(removedIndex)
|
||||
}
|
||||
message := fmt.Sprintf("retained input index %d; removed input indices [%s]", retainedIndex, strings.Join(indices, ", "))
|
||||
if omitted := len(removed) - len(displayed); omitted > 0 {
|
||||
message += fmt.Sprintf("; %d additional removed input indices omitted", omitted)
|
||||
}
|
||||
return contracts.Warning{Scope: itemScope(retainedIndex), ReasonCode: ReasonCodeDuplicateItemCollapsed, Message: message}
|
||||
}
|
||||
func itemScope(index int) string { return fmt.Sprintf("items[%d]", index) }
|
||||
|
||||
func ModuleSpec() pipeline.ModuleSpec {
|
||||
return pipeline.ModuleSpec{Key: Key, Stage: pipeline.StageNormalize, ExecutionClass: contracts.ExecutionClassLLMBacked, Requires: append([]string(nil), requiredCapabilities...), Provides: append([]string(nil), providedCapabilities...), ArtifactKind: dnd.ItemRegistryKind}
|
||||
}
|
||||
func Register(registry *pipeline.NormalizerRegistry) error {
|
||||
return pipeline.RegisterNormalizerBuilder(registry, ModuleSpec(), validateOptions, func(request pipeline.BuildRequest) (contracts.Normalizer[dnd.ItemRegistry], error) {
|
||||
options, err := DecodeOptions(request.Options)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return New(request.Dependencies.LLM, options)
|
||||
})
|
||||
}
|
||||
func validateOptions(options map[string]any) error { _, err := DecodeOptions(options); return err }
|
||||
func DecodeOptions(options map[string]any) (Options, error) {
|
||||
if err := pipeline.RejectUnknownOptions(options); err != nil {
|
||||
return Options{}, normalizerErrorf("%w", err)
|
||||
}
|
||||
return Options{}, nil
|
||||
}
|
||||
func normalizerErrorf(format string, args ...any) error {
|
||||
return fmt.Errorf("dnd item registry normalizer: "+format, args...)
|
||||
}
|
||||
193
internal/modules/dnd/normalize/itemregistry/normalizer_test.go
Normal file
193
internal/modules/dnd/normalize/itemregistry/normalizer_test.go
Normal file
@@ -0,0 +1,193 @@
|
||||
package itemregistry
|
||||
|
||||
import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"errors"
|
||||
"reflect"
|
||||
"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"
|
||||
"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 TestNormalizePreprocessesExactDuplicatesWithoutMutation(t *testing.T) {
|
||||
input := dnd.ItemRegistry{Items: []dnd.Item{
|
||||
{Name: " Rope ", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 1, EndUnitID: 1}}},
|
||||
{Name: "rope", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 1, EndUnitID: 1}}},
|
||||
{Name: "Rope", SourceRefs: []source.SourceRef{{SourceID: "session", StartUnitID: 2, EndUnitID: 2}}},
|
||||
}}
|
||||
before := dnd.ItemRegistry{Items: append([]dnd.Item(nil), input.Items...)}
|
||||
result, err := newNormalizer(t, &recordingNormalizerClient{}).Normalize(context.Background(), normalizeRequest(input))
|
||||
if err != nil || len(result.Value.Items) != 2 || !reflect.DeepEqual(input, before) {
|
||||
t.Fatalf("Normalize() = %#v, %v; want non-mutating exact deduplication", result, err)
|
||||
}
|
||||
if result.Value.Items[0].ID != result.Value.Items[1].ID || !hasWarning(result.Warnings, ReasonCodeDuplicateItemCollapsed) {
|
||||
t.Fatalf("items = %#v, warnings = %#v; want type identity and duplicate warning", result.Value.Items, result.Warnings)
|
||||
}
|
||||
}
|
||||
|
||||
func TestNormalizeAppliesSafeAliasProposalAndPreservesCurrency(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-000001") || strings.Contains(encoded, merged.ID) {
|
||||
t.Fatalf("private inputs = %s", encoded)
|
||||
}
|
||||
currencyClient := &recordingNormalizerClient{response: `{"duplicate_groups":[{"members":["candidate-000001","candidate-000002"],"canonical":"candidate-000001"}]}`}
|
||||
currency := dnd.ItemRegistry{Items: []dnd.Item{{Name: "Gold Pieces", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 10, EndUnitID: 10}}}, {Name: "Silver Pieces", SourceRefs: []source.SourceRef{{SourceID: doc.ID, StartUnitID: 20, EndUnitID: 20}}}}}
|
||||
result, err = newNormalizer(t, currencyClient).Normalize(context.Background(), normalizeRequestWithSource(currency, doc))
|
||||
if err != nil || result.Retry == nil || len(result.Value.Items) != 2 || !hasWarning(result.Warnings, ReasonCodeItemSemanticProposalInvalid) {
|
||||
t.Fatalf("currency result = %#v, %v; want denominations retained", result, err)
|
||||
}
|
||||
}
|
||||
|
||||
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":[{"key":"candidate-000001","name":"Rope","source_refs":[]}]}`), "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":[]}`
|
||||
}
|
||||
if err := json.Unmarshal([]byte(response), output); err != nil {
|
||||
return contracts.StructuredCompletionResponse{}, err
|
||||
}
|
||||
return contracts.StructuredCompletionResponse{Content: json.RawMessage(response)}, nil
|
||||
}
|
||||
|
||||
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
|
||||
}
|
||||
62
internal/modules/dnd/normalize/itemregistry/prompt_assets.go
Normal file
62
internal/modules/dnd/normalize/itemregistry/prompt_assets.go
Normal file
@@ -0,0 +1,62 @@
|
||||
package itemregistry
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"io/fs"
|
||||
"sync"
|
||||
|
||||
rootassets "gitea.maximumdirect.net/eric/notarius/assets"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/framework/llm"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/framework/promptfs"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared"
|
||||
)
|
||||
|
||||
const promptAssetRoot = "assets/prompts"
|
||||
|
||||
var promptAssetManifest = shared.PromptAssetManifest{
|
||||
ModuleDir: PromptID,
|
||||
ModuleFiles: []promptfs.ModulePromptFile{
|
||||
{Name: "prompt.yaml", Path: "prompts/prompt.yaml"},
|
||||
{Name: "instructions.md", Path: "prompts/instructions.md"},
|
||||
{Name: "candidates.md", Path: "prompts/candidates.md"},
|
||||
},
|
||||
SharedFiles: []string{"common-dnd-system.md", "common-dnd-entity-reconciliation.md", "common-dnd-transcript-windows.md"},
|
||||
}
|
||||
|
||||
func moduleAssetFS() (fs.FS, error) {
|
||||
assets, err := fs.Sub(rootassets.FS(), "dnd/item-registry/normalize")
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("scope item normalization assets: %w", err)
|
||||
}
|
||||
return assets, nil
|
||||
}
|
||||
|
||||
func RegisterPromptAssets(registry *llm.AssetRegistry) error {
|
||||
assets, err := moduleAssetFS()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
promptFS, err := promptAssetManifest.PromptFS(assets)
|
||||
if err != nil {
|
||||
return fmt.Errorf("prepare item normalization prompt assets: %w", err)
|
||||
}
|
||||
return registry.RegisterPromptFS(promptFS, promptAssetRoot)
|
||||
}
|
||||
|
||||
func promptAssetMetadata() (string, error) {
|
||||
promptAssetHashOnce.Do(func() {
|
||||
assets, err := moduleAssetFS()
|
||||
if err != nil {
|
||||
promptAssetHashErr = err
|
||||
return
|
||||
}
|
||||
promptAssetHash, promptAssetHashErr = promptAssetManifest.Hash(assets)
|
||||
})
|
||||
return promptAssetHash, promptAssetHashErr
|
||||
}
|
||||
|
||||
var (
|
||||
promptAssetHashOnce sync.Once
|
||||
promptAssetHash string
|
||||
promptAssetHashErr error
|
||||
)
|
||||
153
internal/modules/dnd/normalize/itemregistry/reconciliation.go
Normal file
153
internal/modules/dnd/normalize/itemregistry/reconciliation.go
Normal file
@@ -0,0 +1,153 @@
|
||||
package itemregistry
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/items/identity"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared/diagnostics"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/shared/entityreconcile"
|
||||
)
|
||||
|
||||
type safeReconciliationGroup struct {
|
||||
members []int
|
||||
canonical int
|
||||
}
|
||||
|
||||
func reconciliationCandidates(records []normalizedRecord) []entityreconcile.Candidate {
|
||||
candidates := make([]entityreconcile.Candidate, len(records))
|
||||
for index, record := range records {
|
||||
candidates[index] = entityreconcile.Candidate{Name: record.item.Name, SourceRefs: cloneSourceRefs(record.item.SourceRefs)}
|
||||
}
|
||||
return candidates
|
||||
}
|
||||
|
||||
func reconciliationGroups(assessment entityreconcile.Assessment, candidateKeys []string) []safeReconciliationGroup {
|
||||
positions := make(map[string]int, len(candidateKeys))
|
||||
for index, key := range candidateKeys {
|
||||
positions[key] = index
|
||||
}
|
||||
safeGroups := assessment.SafeGroups()
|
||||
groups := make([]safeReconciliationGroup, 0, len(safeGroups))
|
||||
for _, group := range safeGroups {
|
||||
members := group.Members()
|
||||
memberPositions := make([]int, len(members))
|
||||
valid := true
|
||||
for index, key := range members {
|
||||
position, ok := positions[key]
|
||||
if !ok {
|
||||
valid = false
|
||||
break
|
||||
}
|
||||
memberPositions[index] = position
|
||||
}
|
||||
canonical, ok := positions[group.Canonical()]
|
||||
if valid && ok {
|
||||
groups = append(groups, safeReconciliationGroup{members: memberPositions, canonical: canonical})
|
||||
}
|
||||
}
|
||||
return groups
|
||||
}
|
||||
|
||||
func reconciliationIssues(assessment entityreconcile.Assessment) []string {
|
||||
issues := assessment.Issues()
|
||||
details := make([]string, len(issues))
|
||||
for index, issue := range issues {
|
||||
details[index] = fmt.Sprintf("group %d: %s", issue.GroupIndex, issue.Category)
|
||||
}
|
||||
return details
|
||||
}
|
||||
|
||||
func applySafeGroups(records []normalizedRecord, groups []safeReconciliationGroup, order shared.SourceRefOrder) ([]normalizedRecord, []contracts.Warning, int) {
|
||||
byMember := make(map[int]safeReconciliationGroup, len(groups)*2)
|
||||
for _, group := range groups {
|
||||
for _, member := range group.members {
|
||||
byMember[member] = group
|
||||
}
|
||||
}
|
||||
output := make([]normalizedRecord, 0, len(records)-len(groups))
|
||||
warnings := make([]contracts.Warning, 0, len(groups))
|
||||
rejectedGroups := 0
|
||||
for index, record := range records {
|
||||
group, grouped := byMember[index]
|
||||
if !grouped {
|
||||
output = append(output, cloneRecord(record))
|
||||
continue
|
||||
}
|
||||
if group.members[0] != index {
|
||||
continue
|
||||
}
|
||||
if !canConsolidate(records, group) {
|
||||
for _, member := range group.members {
|
||||
output = append(output, cloneRecord(records[member]))
|
||||
}
|
||||
warnings = append(warnings, contracts.Warning{Scope: itemScope(records[group.members[0]].earliest), ReasonCode: ReasonCodeItemSemanticProposalInvalid, Message: "proposal group preserved because currency denominations must remain distinct"})
|
||||
rejectedGroups++
|
||||
continue
|
||||
}
|
||||
consolidated := consolidateSemanticGroup(records, group, order)
|
||||
output = append(output, consolidated)
|
||||
warnings = append(warnings, semanticDuplicateWarning(consolidated, records[group.canonical]))
|
||||
}
|
||||
return output, warnings, rejectedGroups
|
||||
}
|
||||
|
||||
func canConsolidate(records []normalizedRecord, group safeReconciliationGroup) bool {
|
||||
denomination := ""
|
||||
for _, member := range group.members {
|
||||
current := currencyDenomination(records[member].item.Name)
|
||||
if current == "" {
|
||||
continue
|
||||
}
|
||||
if denomination != "" && denomination != current {
|
||||
return false
|
||||
}
|
||||
denomination = current
|
||||
}
|
||||
return true
|
||||
}
|
||||
|
||||
func currencyDenomination(name string) string {
|
||||
switch identity.ComparisonKey(name) {
|
||||
case "copper piece", "copper pieces", "cp":
|
||||
return "copper"
|
||||
case "silver piece", "silver pieces", "sp":
|
||||
return "silver"
|
||||
case "electrum piece", "electrum pieces", "ep":
|
||||
return "electrum"
|
||||
case "gold piece", "gold pieces", "gp":
|
||||
return "gold"
|
||||
case "platinum piece", "platinum pieces", "pp":
|
||||
return "platinum"
|
||||
default:
|
||||
return ""
|
||||
}
|
||||
}
|
||||
|
||||
func consolidateSemanticGroup(records []normalizedRecord, group safeReconciliationGroup, order shared.SourceRefOrder) normalizedRecord {
|
||||
output := cloneRecord(records[group.members[0]])
|
||||
output.item.Name = records[group.canonical].item.Name
|
||||
for _, member := range group.members[1:] {
|
||||
output.item.SourceRefs = append(output.item.SourceRefs, records[member].item.SourceRefs...)
|
||||
output.inputIndexes = append(output.inputIndexes, records[member].inputIndexes...)
|
||||
if records[member].earliest < output.earliest {
|
||||
output.earliest = records[member].earliest
|
||||
}
|
||||
}
|
||||
output.inputIndexes = sortedUniqueIndexes(output.inputIndexes)
|
||||
output.item.SourceRefs = order.Canonicalize(output.item.SourceRefs)
|
||||
output.item.ID = identity.DeriveID(output.item.Name)
|
||||
return output
|
||||
}
|
||||
|
||||
func semanticDuplicateWarning(record normalizedRecord, canonical normalizedRecord) contracts.Warning {
|
||||
details := make([]string, 0, len(record.inputIndexes)+1)
|
||||
for _, inputIndex := range record.inputIndexes {
|
||||
details = append(details, fmt.Sprintf("input index %d", inputIndex))
|
||||
}
|
||||
if canonical.earliest != record.earliest {
|
||||
details = append(details, fmt.Sprintf("canonical display name from input index %d", canonical.earliest))
|
||||
}
|
||||
return contracts.Warning{Scope: itemScope(record.earliest), ReasonCode: ReasonCodeDuplicateItemCollapsed, Message: diagnostics.Aggregate("semantic duplicate consolidation", details)}
|
||||
}
|
||||
Reference in New Issue
Block a user