Bind extraction reuse to transcript identity
This commit is contained in:
@@ -1491,6 +1491,10 @@ func configureAnalyzeExtractionFixture(t *testing.T, env *Env, m *manifest.Manif
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
input, err := artifacts.ResolveExtractionInputIdentity(paths, m)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
outputs = append(outputs, manifest.ArtifactRecord{
|
||||
Kind: "notarius_index", LocalPath: indexPath, ProducerRunID: producerRunID, Checksum: indexChecksum,
|
||||
})
|
||||
@@ -1498,7 +1502,8 @@ func configureAnalyzeExtractionFixture(t *testing.T, env *Env, m *manifest.Manif
|
||||
Name: "extract", Status: manifest.StatusSucceeded, Outputs: outputs,
|
||||
Metadata: map[string]any{
|
||||
"narratio_run_id": producerRunID, "bundle_root": bundleRoot,
|
||||
"receipt": map[string]any{"run_id": "notarius-run-1", "pipeline_id": "campaign.extract"},
|
||||
"receipt": map[string]any{"run_id": "notarius-run-1", "pipeline_id": "campaign.extract"},
|
||||
"direct_input": input.Metadata(),
|
||||
},
|
||||
}
|
||||
return outputPaths
|
||||
|
||||
@@ -91,9 +91,9 @@ func (extractStage) Run(ctx context.Context, env *Env, m *manifest.Manifest) (*S
|
||||
}
|
||||
|
||||
paths := sessionPathsForEnv(env, sessionID)
|
||||
input, err := artifacts.ResolveSessionArtifact(paths, m, artifacts.ArtifactTranscriptFinalTrimmed)
|
||||
input, err := artifacts.ResolveExtractionInputIdentity(paths, m)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("extract: resolve final-trimmed transcript: %w", err)
|
||||
return nil, fmt.Errorf("extract: resolve final-trimmed transcript identity: %w", err)
|
||||
}
|
||||
|
||||
timeout, err := time.ParseDuration(strings.TrimSpace(notariusConfig.Timeout))
|
||||
@@ -108,10 +108,7 @@ func (extractStage) Run(ctx context.Context, env *Env, m *manifest.Manifest) (*S
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("extract: resolve notarius config path: %w", err)
|
||||
}
|
||||
inputPath, err := absolutePath(input.Path)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("extract: resolve transcript input path: %w", err)
|
||||
}
|
||||
inputPath := input.Path
|
||||
workingDirectory, err := absolutePath(notariusConfig.WorkingDirectory)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("extract: resolve notarius working directory: %w", err)
|
||||
@@ -143,7 +140,7 @@ func (extractStage) Run(ctx context.Context, env *Env, m *manifest.Manifest) (*S
|
||||
}
|
||||
}
|
||||
|
||||
fingerprint, err := extractionFingerprint(resolvedBinary, configPath, notariusConfig, timeout, workingDirectory)
|
||||
fingerprint, err := extractionFingerprint(resolvedBinary, configPath, notariusConfig, timeout, workingDirectory, input)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("extract: build configuration fingerprint: %w", err)
|
||||
}
|
||||
@@ -257,6 +254,7 @@ func (extractStage) Run(ctx context.Context, env *Env, m *manifest.Manifest) (*S
|
||||
"warnings_path": promotedWarningsPath,
|
||||
"narratio_run_id": runID,
|
||||
"configuration_fingerprint": fingerprint,
|
||||
"direct_input": input.Metadata(),
|
||||
"receipt": map[string]any{
|
||||
"run_id": adapterResult.Receipt.RunID, "pipeline_id": adapterResult.Receipt.PipelineID,
|
||||
"normalized_output_count": adapterResult.Receipt.NormalizedOutputCount,
|
||||
@@ -370,12 +368,13 @@ type fingerprintOutput struct {
|
||||
}
|
||||
|
||||
type fingerprintDocument struct {
|
||||
Binary string `json:"binary"`
|
||||
ConfigPath string `json:"config_path"`
|
||||
PipelineID string `json:"pipeline_id"`
|
||||
Timeout string `json:"timeout"`
|
||||
WorkingDirectory string `json:"working_directory"`
|
||||
Outputs []fingerprintOutput `json:"outputs"`
|
||||
Binary string `json:"binary"`
|
||||
ConfigPath string `json:"config_path"`
|
||||
PipelineID string `json:"pipeline_id"`
|
||||
Timeout string `json:"timeout"`
|
||||
WorkingDirectory string `json:"working_directory"`
|
||||
Input artifacts.ExtractionInputIdentity `json:"input"`
|
||||
Outputs []fingerprintOutput `json:"outputs"`
|
||||
}
|
||||
|
||||
func extractionFingerprint(
|
||||
@@ -383,6 +382,7 @@ func extractionFingerprint(
|
||||
cfg *config.NotariusConfig,
|
||||
timeout time.Duration,
|
||||
workingDirectory string,
|
||||
input artifacts.ExtractionInputIdentity,
|
||||
) (string, error) {
|
||||
keys := make([]string, 0, len(cfg.Outputs))
|
||||
for key := range cfg.Outputs {
|
||||
@@ -399,7 +399,7 @@ func extractionFingerprint(
|
||||
}
|
||||
payload, err := json.Marshal(fingerprintDocument{
|
||||
Binary: binary, ConfigPath: configPath, PipelineID: cfg.PipelineID,
|
||||
Timeout: timeout.String(), WorkingDirectory: workingDirectory, Outputs: outputs,
|
||||
Timeout: timeout.String(), WorkingDirectory: workingDirectory, Input: input, Outputs: outputs,
|
||||
})
|
||||
if err != nil {
|
||||
return "", err
|
||||
|
||||
@@ -37,15 +37,6 @@ func (extractStage) ValidateResume(_ context.Context, env *Env, m *manifest.Mani
|
||||
if err != nil {
|
||||
return ResumeValidation{}, fmt.Errorf("extract resume: resolve Notarius working directory: %w", err)
|
||||
}
|
||||
fingerprint, err := extractionFingerprint(resolvedBinary, configPath, cfg, timeout, workingDirectory)
|
||||
if err != nil {
|
||||
return ResumeValidation{}, fmt.Errorf("extract resume: build configuration fingerprint: %w", err)
|
||||
}
|
||||
record := m.Stages[(extractStage{}).Name()]
|
||||
if record == nil || metadataString(record.Metadata, "configuration_fingerprint") != fingerprint {
|
||||
return NonResumable("Notarius invocation contract changed"), nil
|
||||
}
|
||||
|
||||
sessionID := strings.TrimSpace(m.SessionID)
|
||||
if sessionID == "" {
|
||||
sessionID = strings.TrimSpace(env.Config.Session.SessionID)
|
||||
@@ -54,6 +45,18 @@ func (extractStage) ValidateResume(_ context.Context, env *Env, m *manifest.Mani
|
||||
return ResumeValidation{}, fmt.Errorf("extract resume: session ID is required")
|
||||
}
|
||||
paths := sessionPathsForEnv(env, sessionID)
|
||||
input, err := artifacts.ResolveExtractionInputIdentity(paths, m)
|
||||
if err != nil {
|
||||
return ResumeValidation{}, fmt.Errorf("extract resume: resolve final-trimmed transcript identity: %w", err)
|
||||
}
|
||||
fingerprint, err := extractionFingerprint(resolvedBinary, configPath, cfg, timeout, workingDirectory, input)
|
||||
if err != nil {
|
||||
return ResumeValidation{}, fmt.Errorf("extract resume: build configuration fingerprint: %w", err)
|
||||
}
|
||||
record := m.Stages[(extractStage{}).Name()]
|
||||
if record == nil || metadataString(record.Metadata, "configuration_fingerprint") != fingerprint {
|
||||
return NonResumable("Notarius invocation contract changed"), nil
|
||||
}
|
||||
definitions := make(map[string]artifacts.ExtractionArtifactDefinition, len(cfg.Outputs))
|
||||
for key, output := range cfg.Outputs {
|
||||
definitions[key] = artifacts.ExtractionArtifactDefinition{
|
||||
@@ -61,7 +64,7 @@ func (extractStage) ValidateResume(_ context.Context, env *Env, m *manifest.Mani
|
||||
SchemaID: output.SchemaID, SchemaVersion: output.SchemaVersion, ModuleKey: output.ModuleKey,
|
||||
}
|
||||
}
|
||||
proof := artifacts.InspectExtractionEvidence(paths, m, definitions)
|
||||
proof := artifacts.InspectExtractionEvidence(paths, m, definitions, input)
|
||||
if proof.State == artifacts.ExtractionEvidenceUnsafe {
|
||||
return ResumeValidation{}, fmt.Errorf("extract resume: %s", proof.Reason)
|
||||
}
|
||||
|
||||
@@ -269,6 +269,11 @@ func TestExtractStageAdapterAndPromotionFailuresReturnNoOutputs(t *testing.T) {
|
||||
}
|
||||
|
||||
func TestExtractionFingerprintIsIndependentOfOutputMapOrder(t *testing.T) {
|
||||
input := artifacts.ExtractionInputIdentity{
|
||||
Checksum: "checksum", SourceID: artifacts.ArtifactTranscriptFinalTrimmed,
|
||||
ProducerStage: "trim", OutputKind: artifacts.TranscriptOutputKindFinalTrimmed,
|
||||
ProducerRunID: "trim-run", Provenance: "manifest.trim.outputs",
|
||||
}
|
||||
first := &config.NotariusConfig{PipelineID: "pipeline", Outputs: map[string]config.NotariusOutputConfig{
|
||||
"zeta": {LaneID: "z", MediaType: "application/json", SchemaID: "z", SchemaVersion: "v1"},
|
||||
"alpha": {LaneID: "a", MediaType: "application/json", SchemaID: "a", SchemaVersion: "v1"},
|
||||
@@ -276,17 +281,39 @@ func TestExtractionFingerprintIsIndependentOfOutputMapOrder(t *testing.T) {
|
||||
second := &config.NotariusConfig{PipelineID: "pipeline", Outputs: map[string]config.NotariusOutputConfig{
|
||||
"alpha": first.Outputs["alpha"], "zeta": first.Outputs["zeta"],
|
||||
}}
|
||||
one, err := extractionFingerprint("/bin/notarius", "/etc/notarius.yml", first, time.Minute, "/work")
|
||||
one, err := extractionFingerprint("/bin/notarius", "/etc/notarius.yml", first, time.Minute, "/work", input)
|
||||
if err != nil {
|
||||
t.Fatalf("extractionFingerprint(first) error = %v", err)
|
||||
}
|
||||
two, err := extractionFingerprint("/bin/notarius", "/etc/notarius.yml", second, time.Minute, "/work")
|
||||
two, err := extractionFingerprint("/bin/notarius", "/etc/notarius.yml", second, time.Minute, "/work", input)
|
||||
if err != nil {
|
||||
t.Fatalf("extractionFingerprint(second) error = %v", err)
|
||||
}
|
||||
if one != two {
|
||||
t.Fatalf("fingerprints differ: %q != %q", one, two)
|
||||
}
|
||||
for _, test := range []struct {
|
||||
name string
|
||||
mutate func(*artifacts.ExtractionInputIdentity)
|
||||
}{
|
||||
{name: "checksum", mutate: func(identity *artifacts.ExtractionInputIdentity) { identity.Checksum = "different-checksum" }},
|
||||
{name: "source", mutate: func(identity *artifacts.ExtractionInputIdentity) {
|
||||
identity.SourceID = artifacts.ArtifactTranscriptFinal
|
||||
}},
|
||||
{name: "producer", mutate: func(identity *artifacts.ExtractionInputIdentity) { identity.ProducerRunID = "different-trim-run" }},
|
||||
} {
|
||||
t.Run(test.name, func(t *testing.T) {
|
||||
changedInput := input
|
||||
test.mutate(&changedInput)
|
||||
changed, err := extractionFingerprint("/bin/notarius", "/etc/notarius.yml", second, time.Minute, "/work", changedInput)
|
||||
if err != nil {
|
||||
t.Fatalf("extractionFingerprint(changed input) error = %v", err)
|
||||
}
|
||||
if one == changed {
|
||||
t.Fatalf("fingerprint did not change with %s identity", test.name)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestExtractStageResumeValidationAcceptsCurrentImmutableResult(t *testing.T) {
|
||||
@@ -376,6 +403,14 @@ func TestExtractStageResumeValidationRequiresFreshEvidence(t *testing.T) {
|
||||
{name: "no succeeded extract record", mutate: func(_ *testing.T, _ *Env, m *manifest.Manifest) {
|
||||
m.Stages["extract"].Status = manifest.StatusFailed
|
||||
}},
|
||||
{name: "changed direct transcript", mutate: func(t *testing.T, _ *Env, m *manifest.Manifest) {
|
||||
if err := os.WriteFile(m.Stages["trim"].Outputs[0].LocalPath, []byte(`{"segments":[{"id":"changed"}]}`), 0o644); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}},
|
||||
{name: "changed direct producer", mutate: func(_ *testing.T, _ *Env, m *manifest.Manifest) {
|
||||
m.Stages["trim"].Outputs[0].ProducerRunID = "trim-run-2"
|
||||
}},
|
||||
{name: "missing lane", mutate: func(t *testing.T, _ *Env, m *manifest.Manifest) {
|
||||
if err := os.Remove(m.Stages["extract"].Outputs[0].LocalPath); err != nil {
|
||||
t.Fatalf("Remove(lane) error = %v", err)
|
||||
|
||||
@@ -1126,11 +1126,16 @@ func configurePublishExtractionFixture(t *testing.T, env *Env, m *manifest.Manif
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
input, err := artifacts.ResolveExtractionInputIdentity(paths, m)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
m.Stages["extract"] = &manifest.StageRecord{
|
||||
Name: "extract", Status: manifest.StatusSucceeded,
|
||||
Metadata: map[string]any{
|
||||
"narratio_run_id": producerRunID, "bundle_root": bundleRoot,
|
||||
"receipt": map[string]any{"run_id": "notarius-run-1", "pipeline_id": "campaign.extract"},
|
||||
"receipt": map[string]any{"run_id": "notarius-run-1", "pipeline_id": "campaign.extract"},
|
||||
"direct_input": input.Metadata(),
|
||||
},
|
||||
Outputs: []manifest.ArtifactRecord{
|
||||
{
|
||||
|
||||
Reference in New Issue
Block a user