Implement raw module output contracts
This commit is contained in:
@@ -4,10 +4,7 @@ import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"strings"
|
||||
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/core/artifacts"
|
||||
"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/sharedassets/dnd"
|
||||
@@ -45,14 +42,6 @@ func (e *Extractor) Key() string {
|
||||
return Key
|
||||
}
|
||||
|
||||
func (e *Extractor) ArtifactType() string {
|
||||
return ArtifactType
|
||||
}
|
||||
|
||||
func (e *Extractor) SchemaVersion() string {
|
||||
return SchemaVersion
|
||||
}
|
||||
|
||||
func (e *Extractor) ReferenceSlots() []contracts.ReferenceSlot {
|
||||
return dnd.ReferenceSlots(referenceSlotDescriptions)
|
||||
}
|
||||
@@ -77,13 +66,6 @@ func (e *Extractor) ManifestMetadata() map[string]any {
|
||||
return metadata
|
||||
}
|
||||
|
||||
func (e *Extractor) Validators() []contracts.Validator {
|
||||
return []contracts.Validator{
|
||||
ShapeValidator{},
|
||||
SourceRefValidator{},
|
||||
}
|
||||
}
|
||||
|
||||
func (e *Extractor) Extract(ctx context.Context, req contracts.ExtractionRequest) (contracts.ExtractionResult, error) {
|
||||
if e == nil {
|
||||
return contracts.ExtractionResult{}, extractorErrorf("extractor must not be nil")
|
||||
@@ -121,22 +103,26 @@ func (e *Extractor) Extract(ctx context.Context, req contracts.ExtractionRequest
|
||||
if response.SpellCasts == nil {
|
||||
return contracts.ExtractionResult{}, extractorErrorf("malformed structured output: spell_casts must be present")
|
||||
}
|
||||
if len(response.SpellCasts) == 0 {
|
||||
return contracts.ExtractionResult{}, nil
|
||||
content, err := json.Marshal(response)
|
||||
if err != nil {
|
||||
return contracts.ExtractionResult{}, extractorErrorf("marshal raw output: %w", err)
|
||||
}
|
||||
|
||||
candidates := make([]artifacts.ArtifactCandidate, 0, len(response.SpellCasts))
|
||||
for i, spellCast := range response.SpellCasts {
|
||||
payload, err := spellCastPayload(spellCast)
|
||||
if err != nil {
|
||||
return contracts.ExtractionResult{}, extractorErrorf("marshal spell cast[%d]: %w", i, err)
|
||||
}
|
||||
candidates = append(candidates, artifacts.ArtifactCandidate{
|
||||
Payload: payload,
|
||||
SourceRefs: sourceRefCandidates(req.Source, spellCast.SourceRefs),
|
||||
})
|
||||
}
|
||||
return contracts.ExtractionResult{Candidates: candidates}, nil
|
||||
return contracts.ExtractionResult{
|
||||
Output: contracts.ExtractOutput{
|
||||
Schema: contracts.ResponseSchema{
|
||||
ID: ResponseSchemaID,
|
||||
Name: ResponseSchemaName,
|
||||
Version: SchemaVersion,
|
||||
},
|
||||
Payload: contracts.RawPayload{
|
||||
Content: content,
|
||||
MediaType: "application/json",
|
||||
Metadata: map[string]any{
|
||||
"spell_cast_count": len(response.SpellCasts),
|
||||
},
|
||||
},
|
||||
},
|
||||
}, nil
|
||||
}
|
||||
|
||||
func ModuleSpec() pipeline.ModuleSpec {
|
||||
@@ -155,26 +141,6 @@ func Register(registry *pipeline.ExtractorRegistry) error {
|
||||
})
|
||||
}
|
||||
|
||||
func spellCastPayload(spellCast spellCastResponse) (json.RawMessage, error) {
|
||||
return json.Marshal(SpellCast{
|
||||
Caster: strings.TrimSpace(spellCast.Caster),
|
||||
Spell: strings.TrimSpace(spellCast.Spell),
|
||||
Effect: strings.TrimSpace(spellCast.Effect),
|
||||
NarrativeDescription: strings.TrimSpace(spellCast.NarrativeDescription),
|
||||
})
|
||||
}
|
||||
|
||||
func sourceRefCandidates(doc *source.SourceDocument, refs []dnd.SourceRefResponse) []source.SourceRef {
|
||||
if len(refs) == 0 {
|
||||
return nil
|
||||
}
|
||||
out := make([]source.SourceRef, 0, len(refs))
|
||||
for _, ref := range refs {
|
||||
out = append(out, dnd.SourceRefCandidate(doc, ref))
|
||||
}
|
||||
return out
|
||||
}
|
||||
|
||||
func extractorErrorf(format string, args ...any) error {
|
||||
return fmt.Errorf("dnd spells extractor: "+format, args...)
|
||||
}
|
||||
|
||||
@@ -7,12 +7,11 @@ import (
|
||||
"strings"
|
||||
"testing"
|
||||
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/core/source"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/modules/sharedassets/dnd"
|
||||
)
|
||||
|
||||
func TestExtractReturnsSpellCandidateFromStructuredOutput(t *testing.T) {
|
||||
func TestExtractReturnsRawOutputFromStructuredResponse(t *testing.T) {
|
||||
client := &fakeSpellsLLMClient{
|
||||
response: extractionResponse{
|
||||
SpellCasts: []spellCastResponse{
|
||||
@@ -53,29 +52,19 @@ func TestExtractReturnsSpellCandidateFromStructuredOutput(t *testing.T) {
|
||||
t.Fatalf("transcript content = %q, want original source input", got)
|
||||
}
|
||||
|
||||
if len(result.Candidates) != 1 {
|
||||
t.Fatalf("len(Candidates) = %d, want 1", len(result.Candidates))
|
||||
if result.Output.Payload.MediaType != "application/json" {
|
||||
t.Fatalf("MediaType = %q, want application/json", result.Output.Payload.MediaType)
|
||||
}
|
||||
candidate := result.Candidates[0]
|
||||
if candidate.Index != 0 || candidate.ExtractorKey != "" || candidate.ArtifactType != "" || candidate.SchemaVersion != "" {
|
||||
t.Fatalf("candidate envelope fields = %#v, want runner-normalized zero values", candidate)
|
||||
if result.Output.Schema.ID != ResponseSchemaID || result.Output.Schema.Name != ResponseSchemaName || result.Output.Schema.Version != SchemaVersion {
|
||||
t.Fatalf("schema = %#v, want response schema provenance", result.Output.Schema)
|
||||
}
|
||||
var payload SpellCast
|
||||
if err := json.Unmarshal(candidate.Payload, &payload); err != nil {
|
||||
t.Fatalf("Unmarshal(Payload) error = %v, want nil", err)
|
||||
|
||||
var payload extractionResponse
|
||||
if err := json.Unmarshal(result.Output.Payload.Content, &payload); err != nil {
|
||||
t.Fatalf("Unmarshal(Content) error = %v, want nil", err)
|
||||
}
|
||||
wantPayload := SpellCast{
|
||||
Caster: "Aria",
|
||||
Spell: "Cure Wounds",
|
||||
Effect: "Heals an injured ally.",
|
||||
NarrativeDescription: "Aria restores the fighter after the fight.",
|
||||
}
|
||||
if payload != wantPayload {
|
||||
t.Fatalf("payload = %#v, want %#v", payload, wantPayload)
|
||||
}
|
||||
wantRef := source.SourceRef{SourceID: "session-alpha", StartUnitID: 1, EndUnitID: 2}
|
||||
if len(candidate.SourceRefs) != 1 || candidate.SourceRefs[0] != wantRef {
|
||||
t.Fatalf("SourceRefs = %#v, want %#v", candidate.SourceRefs, []source.SourceRef{wantRef})
|
||||
if len(payload.SpellCasts) != 1 || payload.SpellCasts[0].Spell != " Cure Wounds " {
|
||||
t.Fatalf("payload = %#v, want raw structured response", payload)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -174,15 +163,19 @@ func TestPromptInputsMapLegacyRosterReferenceToParty(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
func TestExtractReturnsNoCandidatesForEmptyResponse(t *testing.T) {
|
||||
func TestExtractReturnsRawOutputForEmptyResponse(t *testing.T) {
|
||||
client := &fakeSpellsLLMClient{response: extractionResponse{SpellCasts: []spellCastResponse{}}}
|
||||
|
||||
result, err := New().Extract(context.Background(), extractionRequestWithClient(client))
|
||||
if err != nil {
|
||||
t.Fatalf("Extract() error = %v, want nil", err)
|
||||
}
|
||||
if len(result.Candidates) != 0 {
|
||||
t.Fatalf("Candidates = %#v, want none", result.Candidates)
|
||||
var payload extractionResponse
|
||||
if err := json.Unmarshal(result.Output.Payload.Content, &payload); err != nil {
|
||||
t.Fatalf("Unmarshal(Content) error = %v, want nil", err)
|
||||
}
|
||||
if len(payload.SpellCasts) != 0 {
|
||||
t.Fatalf("SpellCasts = %#v, want none", payload.SpellCasts)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -271,23 +264,16 @@ func TestExtractPreservesResponseOrder(t *testing.T) {
|
||||
if err != nil {
|
||||
t.Fatalf("Extract() error = %v, want nil", err)
|
||||
}
|
||||
if len(result.Candidates) != 2 {
|
||||
t.Fatalf("len(Candidates) = %d, want 2", len(result.Candidates))
|
||||
var payload extractionResponse
|
||||
if err := json.Unmarshal(result.Output.Payload.Content, &payload); err != nil {
|
||||
t.Fatalf("Unmarshal(Content) error = %v, want nil", err)
|
||||
}
|
||||
|
||||
var first, second SpellCast
|
||||
if err := json.Unmarshal(result.Candidates[0].Payload, &first); err != nil {
|
||||
t.Fatalf("Unmarshal(first) error = %v, want nil", err)
|
||||
}
|
||||
if err := json.Unmarshal(result.Candidates[1].Payload, &second); err != nil {
|
||||
t.Fatalf("Unmarshal(second) error = %v, want nil", err)
|
||||
}
|
||||
if first.Spell != "Cure Wounds" || second.Spell != "Fire Bolt" {
|
||||
t.Fatalf("candidate order = %q, %q; want response order", first.Spell, second.Spell)
|
||||
if len(payload.SpellCasts) != 2 || payload.SpellCasts[0].Spell != "Cure Wounds" || payload.SpellCasts[1].Spell != "Fire Bolt" {
|
||||
t.Fatalf("spell order = %#v, want response order", payload.SpellCasts)
|
||||
}
|
||||
}
|
||||
|
||||
func TestExtractCopiesCandidateSourceRefs(t *testing.T) {
|
||||
func TestExtractDefensivelyCopiesRawContent(t *testing.T) {
|
||||
client := &fakeSpellsLLMClient{
|
||||
response: extractionResponse{
|
||||
SpellCasts: []spellCastResponse{
|
||||
@@ -308,8 +294,12 @@ func TestExtractCopiesCandidateSourceRefs(t *testing.T) {
|
||||
}
|
||||
client.response.SpellCasts[0].SourceRefs[0].StartUnitID = dnd.UnitRefFromInt(99)
|
||||
|
||||
if got := result.Candidates[0].SourceRefs[0].StartUnitID; got != 1 {
|
||||
t.Fatalf("candidate source ref start = %d, want copied 1", got)
|
||||
var payload extractionResponse
|
||||
if err := json.Unmarshal(result.Output.Payload.Content, &payload); err != nil {
|
||||
t.Fatalf("Unmarshal(Content) error = %v, want nil", err)
|
||||
}
|
||||
if got := payload.SpellCasts[0].SourceRefs[0].StartUnitID.String(); got != "1" {
|
||||
t.Fatalf("source ref start = %q, want copied 1", got)
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -17,12 +17,6 @@ func TestNewReturnsExtractorWithMetadata(t *testing.T) {
|
||||
if extractor.Key() != Key {
|
||||
t.Fatalf("extractor.Key() = %q, want %q", extractor.Key(), Key)
|
||||
}
|
||||
if extractor.ArtifactType() != ArtifactType {
|
||||
t.Fatalf("extractor.ArtifactType() = %q, want %q", extractor.ArtifactType(), ArtifactType)
|
||||
}
|
||||
if extractor.SchemaVersion() != SchemaVersion {
|
||||
t.Fatalf("extractor.SchemaVersion() = %q, want %q", extractor.SchemaVersion(), SchemaVersion)
|
||||
}
|
||||
}
|
||||
|
||||
func TestModuleSpec(t *testing.T) {
|
||||
|
||||
@@ -48,31 +48,30 @@ func TestRunnerProcessesSeriatimInputWithDNDSpellsExtractor(t *testing.T) {
|
||||
t.Fatalf("Run() error = %v, want nil", err)
|
||||
}
|
||||
|
||||
if len(output.Approved) != 2 {
|
||||
t.Fatalf("len(Approved) = %d, want 2", len(output.Approved))
|
||||
if len(output.NormalizeOutputs) != 1 {
|
||||
t.Fatalf("len(NormalizeOutputs) = %d, want 1", len(output.NormalizeOutputs))
|
||||
}
|
||||
var first, second SpellCast
|
||||
if err := json.Unmarshal(output.Approved[0].Payload, &first); err != nil {
|
||||
t.Fatalf("Unmarshal(first payload) error = %v, want nil", err)
|
||||
rawOutput := output.NormalizeOutputs[0]
|
||||
if rawOutput.LaneID != "spells" || rawOutput.Schema.ID != ResponseSchemaID || rawOutput.Schema.Version != SchemaVersion {
|
||||
t.Fatalf("raw output envelope = %#v, want dnd spells schema on spells lane", rawOutput)
|
||||
}
|
||||
if err := json.Unmarshal(output.Approved[1].Payload, &second); err != nil {
|
||||
t.Fatalf("Unmarshal(second payload) error = %v, want nil", err)
|
||||
response := decodeRunnerSpellResponse(t, rawOutput.Payload.Content)
|
||||
if len(response.SpellCasts) != 2 {
|
||||
t.Fatalf("len(spell_casts) = %d, want 2", len(response.SpellCasts))
|
||||
}
|
||||
first, second := response.SpellCasts[0], response.SpellCasts[1]
|
||||
if first.Spell != "Cure Wounds" || second.Spell != "Fire Bolt" {
|
||||
t.Fatalf("approved spell order = %q, %q; want response order", first.Spell, second.Spell)
|
||||
t.Fatalf("spell order = %q, %q; want response order", first.Spell, second.Spell)
|
||||
}
|
||||
if first.Caster != "Aria" || second.Caster != "Borin" {
|
||||
t.Fatalf("approved casters = %q, %q; want spell data", first.Caster, second.Caster)
|
||||
t.Fatalf("casters = %q, %q; want spell data", first.Caster, second.Caster)
|
||||
}
|
||||
for _, artifact := range output.Approved {
|
||||
if artifact.ExtractorKey != Key || artifact.ArtifactType != ArtifactType || artifact.SchemaVersion != SchemaVersion {
|
||||
t.Fatalf("approved artifact envelope = %#v, want dnd spells envelope", artifact)
|
||||
for _, spell := range response.SpellCasts {
|
||||
if len(spell.SourceRefs) != 1 {
|
||||
t.Fatalf("len(SourceRefs) = %d, want 1", len(spell.SourceRefs))
|
||||
}
|
||||
if len(artifact.SourceRefs) != 1 {
|
||||
t.Fatalf("len(SourceRefs) = %d, want 1", len(artifact.SourceRefs))
|
||||
}
|
||||
if err := source.ValidateRef(expectedDoc, artifact.SourceRefs[0]); err != nil {
|
||||
t.Fatalf("ValidateRef() error = %v, want nil", err)
|
||||
if spell.SourceRefs[0].SourceID != expectedDoc.ID {
|
||||
t.Fatalf("SourceID = %q, want fixture document ID", spell.SourceRefs[0].SourceID)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -137,8 +136,8 @@ func TestRunnerPassesPartyAndGlossaryReferencesToDNDSpellsPrompt(t *testing.T) {
|
||||
t.Fatalf("Run() error = %v, want nil", err)
|
||||
}
|
||||
|
||||
if len(output.Approved) != 1 {
|
||||
t.Fatalf("len(Approved) = %d, want 1", len(output.Approved))
|
||||
if len(output.NormalizeOutputs) != 1 {
|
||||
t.Fatalf("len(NormalizeOutputs) = %d, want 1", len(output.NormalizeOutputs))
|
||||
}
|
||||
if len(output.Manifest.References) != 2 {
|
||||
t.Fatalf("manifest references = %#v, want party and glossary provenance", output.Manifest.References)
|
||||
@@ -178,8 +177,12 @@ func TestRunnerDoesNotExtractSpellMentionedOnlyInPartyReference(t *testing.T) {
|
||||
t.Fatalf("Run() error = %v, want nil", err)
|
||||
}
|
||||
|
||||
if len(output.Approved) != 0 {
|
||||
t.Fatalf("approved artifacts = %#v, want no party-reference-only spell casts", output.Approved)
|
||||
if len(output.NormalizeOutputs) != 1 {
|
||||
t.Fatalf("len(NormalizeOutputs) = %d, want empty spell response output", len(output.NormalizeOutputs))
|
||||
}
|
||||
response := decodeRunnerSpellResponse(t, output.NormalizeOutputs[0].Payload.Content)
|
||||
if len(response.SpellCasts) != 0 {
|
||||
t.Fatalf("spell_casts = %#v, want no party-reference-only spell casts", response.SpellCasts)
|
||||
}
|
||||
if len(llmClient.requests) != 1 {
|
||||
t.Fatalf("LLM calls = %d, want 1", len(llmClient.requests))
|
||||
@@ -196,7 +199,7 @@ func TestRunnerDoesNotExtractSpellMentionedOnlyInPartyReference(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
func TestRunnerRejectsDNDSpellCastWithInvalidSourceRef(t *testing.T) {
|
||||
func TestRunnerCarriesDNDSpellCastWithInvalidSourceRefAsRawOutput(t *testing.T) {
|
||||
raw := readDNDSpellsFixture(t)
|
||||
resolved := resolveDNDSpellsPipeline(t)
|
||||
llmClient := &fakeSpellsLLMClient{
|
||||
@@ -221,21 +224,21 @@ func TestRunnerRejectsDNDSpellCastWithInvalidSourceRef(t *testing.T) {
|
||||
if err != nil {
|
||||
t.Fatalf("Run() error = %v, want nil", err)
|
||||
}
|
||||
if len(output.Approved) != 0 {
|
||||
t.Fatalf("len(Approved) = %d, want 0", len(output.Approved))
|
||||
if len(output.NormalizeOutputs) != 1 {
|
||||
t.Fatalf("len(NormalizeOutputs) = %d, want 1", len(output.NormalizeOutputs))
|
||||
}
|
||||
if len(output.Rejected) != 1 {
|
||||
t.Fatalf("len(Rejected) = %d, want 1", len(output.Rejected))
|
||||
response := decodeRunnerSpellResponse(t, output.NormalizeOutputs[0].Payload.Content)
|
||||
if len(response.SpellCasts) != 1 {
|
||||
t.Fatalf("len(spell_casts) = %d, want 1", len(response.SpellCasts))
|
||||
}
|
||||
rejected := output.Rejected[0]
|
||||
if rejected.ValidatorName != sourceRefValidatorName {
|
||||
t.Fatalf("ValidatorName = %q, want %q", rejected.ValidatorName, sourceRefValidatorName)
|
||||
if response.SpellCasts[0].SourceRefs[0].SourceID != "spell-session" {
|
||||
t.Fatalf("SourceID = %q, want raw invalid source ref preserved", response.SpellCasts[0].SourceRefs[0].SourceID)
|
||||
}
|
||||
if rejected.ReasonCode != reasonInvalidSourceRef {
|
||||
t.Fatalf("ReasonCode = %q, want %q", rejected.ReasonCode, reasonInvalidSourceRef)
|
||||
if len(output.Rejected) != 0 {
|
||||
t.Fatalf("len(Rejected) = %d, want 0", len(output.Rejected))
|
||||
}
|
||||
if output.Manifest.ValidationStatus != "rejected" {
|
||||
t.Fatalf("ValidationStatus = %q, want rejected", output.Manifest.ValidationStatus)
|
||||
if output.Manifest.ValidationStatus != "approved" {
|
||||
t.Fatalf("ValidationStatus = %q, want approved", output.Manifest.ValidationStatus)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -345,3 +348,13 @@ func parseDNDSpellsFixture(t *testing.T, raw []byte) *source.SourceDocument {
|
||||
}
|
||||
return doc
|
||||
}
|
||||
|
||||
func decodeRunnerSpellResponse(t *testing.T, raw []byte) extractionResponse {
|
||||
t.Helper()
|
||||
|
||||
var response extractionResponse
|
||||
if err := json.Unmarshal(raw, &response); err != nil {
|
||||
t.Fatalf("Unmarshal(raw output) error = %v, want nil", err)
|
||||
}
|
||||
return response
|
||||
}
|
||||
|
||||
@@ -12,26 +12,6 @@ import (
|
||||
"gitea.maximumdirect.net/eric/notarius/internal/framework/validate"
|
||||
)
|
||||
|
||||
func TestExtractorValidatorsReturnsExpectedChain(t *testing.T) {
|
||||
validators := New().Validators()
|
||||
|
||||
if len(validators) != 2 {
|
||||
t.Fatalf("len(Validators()) = %d, want 2", len(validators))
|
||||
}
|
||||
if validators[0].Name() != shapeValidatorName {
|
||||
t.Fatalf("Validators()[0].Name() = %q, want %q", validators[0].Name(), shapeValidatorName)
|
||||
}
|
||||
if validators[1].Name() != sourceRefValidatorName {
|
||||
t.Fatalf("Validators()[1].Name() = %q, want %q", validators[1].Name(), sourceRefValidatorName)
|
||||
}
|
||||
|
||||
validators[0] = nil
|
||||
again := New().Validators()
|
||||
if len(again) != 2 || again[0] == nil || again[0].Name() != shapeValidatorName {
|
||||
t.Fatalf("Validators() after caller mutation = %#v, want fresh validators", again)
|
||||
}
|
||||
}
|
||||
|
||||
func TestValidatorsApproveValidCandidate(t *testing.T) {
|
||||
candidate := validSpellCandidate(7)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user