Update Notarius integration and plan references

This commit is contained in:
2026-08-29 14:41:22 +00:00
parent e433c86203
commit 42ed81cbe1
11 changed files with 1450 additions and 114 deletions

View File

@@ -15,13 +15,19 @@ import (
)
const (
maxReceiptBytes = 1 << 20
maxIndexBytes = 4 << 20
maxSummaryBytes = 4 << 20
canonicalIndexFile = "index.json"
canonicalManifestFile = "manifest.json"
canonicalRejectedFile = "rejected.json"
canonicalWarningsFile = "warnings.json"
maxReceiptBytes = 1 << 20
maxIndexBytes = 4 << 20
maxSummaryBytes = 4 << 20
canonicalIndexFile = "index.json"
canonicalManifestFile = "manifest.json"
canonicalRejectedFile = "rejected.json"
canonicalWarningsFile = "warnings.json"
canonicalDiagnosticsFile = "diagnostics.json"
warningsSchemaVersion = "notarius.warnings.v2"
diagnosticsSchemaVersion = "notarius.diagnostics.v1"
maxWarningGroups = 128
maxDiagnosticGroups = 256
maxFindingSamples = 3
)
type subprocessRun func(context.Context, subprocess.RunRequest) (subprocess.RunResult, error)
@@ -95,12 +101,30 @@ func (r *SubprocessRunner) Run(ctx context.Context, req RunRequest) (RunResult,
if err != nil {
return baseResult, err
}
diagnostics, diagnosticOccurrences, diagnosticsTruncated, err := loadDiagnostics(index.DiagnosticsPath)
if err != nil {
return baseResult, err
}
if receipt.NormalizedOutputCount != len(index.Lanes) || receipt.RejectedOutputCount != len(rejections) ||
receipt.WarningGroupCount != len(warnings) || receipt.DiagnosticGroupCount != len(diagnostics) {
return baseResult, fmt.Errorf("notarius receipt counts do not match published bundle")
}
warningOccurrences, err := sumWarningOccurrences(warnings)
if err != nil {
return baseResult, err
}
if receipt.WarningOccurrenceCount != warningOccurrences ||
receipt.DiagnosticOccurrenceCount != diagnosticOccurrences ||
receipt.DiagnosticsTruncated != diagnosticsTruncated {
return baseResult, fmt.Errorf("notarius receipt occurrence counts do not match published bundle")
}
baseResult.Receipt = receipt
baseResult.Index = index
baseResult.BundleRoot = bundleRoot
baseResult.Rejections = rejections
baseResult.Warnings = warnings
baseResult.Diagnostics = diagnostics
return baseResult, nil
}
@@ -154,16 +178,74 @@ func validateRunRequest(req RunRequest) error {
}
type receiptDocument struct {
SchemaVersion string `json:"schema_version"`
RunID string `json:"run_id"`
PipelineID string `json:"pipeline_id"`
OutputDirectory string `json:"output_directory"`
IndexFile string `json:"index_file"`
NormalizedOutputCount *int `json:"normalized_output_count"`
RejectedOutputCount *int `json:"rejected_output_count"`
WarningCount *int `json:"warning_count"`
ValidationStatus string `json:"validation_status"`
DebugDirectory string `json:"debug_directory"`
SchemaVersion string `json:"schema_version"`
RunID string `json:"run_id"`
PipelineID string `json:"pipeline_id"`
OutputDirectory string `json:"output_directory"`
IndexFile string `json:"index_file"`
NormalizedOutputCount *int `json:"normalized_output_count"`
RejectedOutputCount *int `json:"rejected_output_count"`
WarningGroupCount *int `json:"warning_group_count"`
WarningOccurrenceCount *int `json:"warning_occurrence_count"`
DiagnosticGroupCount *int `json:"diagnostic_group_count"`
DiagnosticOccurrenceCount *int `json:"diagnostic_occurrence_count"`
DiagnosticsTruncated *bool `json:"diagnostics_truncated"`
ValidationStatus string `json:"validation_status"`
ValidationSummaries []validationSummaryDocument `json:"validation_summaries"`
DebugDirectory string `json:"debug_directory"`
}
type validationSummaryDocument struct {
Stage string `json:"stage"`
StepID string `json:"step_id"`
LaneID string `json:"lane_id"`
ModuleKey string `json:"module_key"`
ChunkID string `json:"chunk_id"`
Status string `json:"status"`
RejectingValidators []string `json:"rejecting_validators"`
ReasonCodes []string `json:"reason_codes"`
IncompleteValidators []string `json:"incomplete_validators"`
ProducerAttemptCount *int `json:"producer_attempt_count"`
TerminalAction string `json:"terminal_action"`
}
func validValidationStatus(value string) bool {
switch value {
case "approved", "rejected", "incomplete":
return true
default:
return false
}
}
func validateValidationSummaries(documents []validationSummaryDocument) ([]ValidationSummary, error) {
summaries := make([]ValidationSummary, 0, len(documents))
for _, document := range documents {
if document.Status != "complete" && document.Status != "rejected" && document.Status != "incomplete" {
return nil, fmt.Errorf("notarius validation summary status %q is invalid", document.Status)
}
if document.ProducerAttemptCount == nil || *document.ProducerAttemptCount <= 0 || !validTerminalAction(document.TerminalAction) {
return nil, fmt.Errorf("notarius validation summary is missing required fields")
}
summaries = append(summaries, ValidationSummary{
Stage: document.Stage, StepID: document.StepID, LaneID: document.LaneID,
ModuleKey: document.ModuleKey, ChunkID: document.ChunkID, Status: document.Status,
RejectingValidators: append([]string(nil), document.RejectingValidators...),
ReasonCodes: append([]string(nil), document.ReasonCodes...),
IncompleteValidators: append([]string(nil), document.IncompleteValidators...),
ProducerAttemptCount: *document.ProducerAttemptCount, TerminalAction: document.TerminalAction,
})
}
return summaries, nil
}
func validTerminalAction(value string) bool {
switch value {
case "accepted", "reject_output", "warn_continue", "fail_run":
return true
default:
return false
}
}
func loadReceipt(path, pipelineID string) (Receipt, error) {
@@ -177,7 +259,9 @@ func loadReceipt(path, pipelineID string) (Receipt, error) {
if strings.TrimSpace(document.RunID) == "" || strings.TrimSpace(document.PipelineID) == "" ||
strings.TrimSpace(document.OutputDirectory) == "" || strings.TrimSpace(document.ValidationStatus) == "" ||
document.NormalizedOutputCount == nil ||
document.RejectedOutputCount == nil || document.WarningCount == nil {
document.RejectedOutputCount == nil || document.WarningGroupCount == nil ||
document.WarningOccurrenceCount == nil || document.DiagnosticGroupCount == nil ||
document.DiagnosticOccurrenceCount == nil || document.DiagnosticsTruncated == nil {
return Receipt{}, fmt.Errorf("notarius receipt is missing required fields")
}
if document.IndexFile != canonicalIndexFile {
@@ -186,9 +270,18 @@ func loadReceipt(path, pipelineID string) (Receipt, error) {
if document.PipelineID != pipelineID {
return Receipt{}, fmt.Errorf("notarius receipt pipeline id %q does not match requested pipeline %q", document.PipelineID, pipelineID)
}
if *document.NormalizedOutputCount < 0 || *document.RejectedOutputCount < 0 || *document.WarningCount < 0 {
if *document.NormalizedOutputCount < 0 || *document.RejectedOutputCount < 0 ||
*document.WarningGroupCount < 0 || *document.WarningOccurrenceCount < 0 ||
*document.DiagnosticGroupCount < 0 || *document.DiagnosticOccurrenceCount < 0 {
return Receipt{}, fmt.Errorf("notarius receipt counts must be non-negative")
}
if !validValidationStatus(document.ValidationStatus) {
return Receipt{}, fmt.Errorf("notarius receipt validation_status %q is invalid", document.ValidationStatus)
}
validationSummaries, err := validateValidationSummaries(document.ValidationSummaries)
if err != nil {
return Receipt{}, err
}
if !filepath.IsAbs(document.OutputDirectory) {
return Receipt{}, fmt.Errorf("notarius receipt output directory must be absolute")
}
@@ -196,16 +289,21 @@ func loadReceipt(path, pipelineID string) (Receipt, error) {
return Receipt{}, fmt.Errorf("notarius receipt debug directory must be absolute when present")
}
return Receipt{
SchemaVersion: document.SchemaVersion,
RunID: document.RunID,
PipelineID: document.PipelineID,
OutputDirectory: filepath.Clean(document.OutputDirectory),
IndexFile: document.IndexFile,
NormalizedOutputCount: *document.NormalizedOutputCount,
RejectedOutputCount: *document.RejectedOutputCount,
WarningCount: *document.WarningCount,
ValidationStatus: document.ValidationStatus,
DebugDirectory: document.DebugDirectory,
SchemaVersion: document.SchemaVersion,
RunID: document.RunID,
PipelineID: document.PipelineID,
OutputDirectory: filepath.Clean(document.OutputDirectory),
IndexFile: document.IndexFile,
NormalizedOutputCount: *document.NormalizedOutputCount,
RejectedOutputCount: *document.RejectedOutputCount,
WarningGroupCount: *document.WarningGroupCount,
WarningOccurrenceCount: *document.WarningOccurrenceCount,
DiagnosticGroupCount: *document.DiagnosticGroupCount,
DiagnosticOccurrenceCount: *document.DiagnosticOccurrenceCount,
DiagnosticsTruncated: *document.DiagnosticsTruncated,
ValidationStatus: document.ValidationStatus,
ValidationSummaries: validationSummaries,
DebugDirectory: document.DebugDirectory,
}, nil
}
@@ -214,6 +312,7 @@ type indexDocument struct {
OutputFiles *[]laneDocument `json:"output_files"`
RejectedFile string `json:"rejected_file"`
WarningsFile string `json:"warnings_file"`
DiagnosticsFile string `json:"diagnostics_file"`
ChunkMap *pipelineDocument `json:"chunk_map"`
EvidenceContext *pipelineDocument `json:"evidence_context"`
}
@@ -250,6 +349,7 @@ func loadIndex(bundleRoot, indexPath string) (Index, error) {
{name: "manifest_file", got: document.ManifestFile, want: canonicalManifestFile},
{name: "rejected_file", got: document.RejectedFile, want: canonicalRejectedFile},
{name: "warnings_file", got: document.WarningsFile, want: canonicalWarningsFile},
{name: "diagnostics_file", got: document.DiagnosticsFile, want: canonicalDiagnosticsFile},
} {
if field.got != field.want {
return Index{}, fmt.Errorf("notarius index %s %q is incompatible; want %q", field.name, field.got, field.want)
@@ -260,10 +360,11 @@ func loadIndex(bundleRoot, indexPath string) (Index, error) {
}
index := Index{
Path: indexPath,
ManifestFile: document.ManifestFile,
RejectedFile: document.RejectedFile,
WarningsFile: document.WarningsFile,
Path: indexPath,
ManifestFile: document.ManifestFile,
RejectedFile: document.RejectedFile,
WarningsFile: document.WarningsFile,
DiagnosticsFile: document.DiagnosticsFile,
}
var err error
if index.ManifestPath, err = resolveRegularFile(bundleRoot, index.ManifestFile); err != nil {
@@ -275,6 +376,9 @@ func loadIndex(bundleRoot, indexPath string) (Index, error) {
if index.WarningsPath, err = resolveRegularFile(bundleRoot, index.WarningsFile); err != nil {
return Index{}, fmt.Errorf("resolve notarius warning file: %w", err)
}
if index.DiagnosticsPath, err = resolveRegularFile(bundleRoot, index.DiagnosticsFile); err != nil {
return Index{}, fmt.Errorf("resolve notarius diagnostics file: %w", err)
}
seenLanes := make(map[string]struct{}, len(*document.OutputFiles))
for _, lane := range *document.OutputFiles {
@@ -362,12 +466,34 @@ func loadRejections(path string) ([]RejectionSummary, error) {
return summaries, nil
}
type warningDocument struct {
Warnings *[]struct {
type findingGroupDocument struct {
Disposition string `json:"disposition"`
Category string `json:"category"`
ReasonCode string `json:"reason_code"`
Origin diagnosticOriginDocument `json:"origin"`
OccurrenceCount *int `json:"occurrence_count"`
Samples *[]struct {
Scope string `json:"scope"`
ReasonCode string `json:"reason_code"`
Message string `json:"message"`
} `json:"warnings"`
ChunkID string `json:"chunk_id"`
ChunkIndex *int `json:"chunk_index"`
} `json:"samples"`
OmittedSampleCount *int `json:"omitted_sample_count"`
}
type diagnosticOriginDocument struct {
Stage string `json:"stage"`
StepID string `json:"step_id"`
LaneID string `json:"lane_id"`
ModuleKey string `json:"module_key"`
ValidatorKey string `json:"validator_key"`
}
type warningDocument struct {
SchemaVersion string `json:"schema_version"`
GroupCount *int `json:"group_count"`
OccurrenceCount *int `json:"occurrence_count"`
Groups *[]findingGroupDocument `json:"groups"`
}
func loadWarnings(path string) ([]WarningSummary, error) {
@@ -375,19 +501,151 @@ func loadWarnings(path string) ([]WarningSummary, error) {
if err := decodeBoundedJSON(path, maxSummaryBytes, &document); err != nil {
return nil, fmt.Errorf("decode notarius warnings: %w", err)
}
if document.Warnings == nil {
return nil, fmt.Errorf("notarius warning document is missing warnings array")
if document.SchemaVersion != warningsSchemaVersion || document.GroupCount == nil ||
document.OccurrenceCount == nil || document.Groups == nil {
return nil, fmt.Errorf("notarius warning document is missing or incompatible required fields")
}
summaries := make([]WarningSummary, 0, len(*document.Warnings))
for _, item := range *document.Warnings {
if strings.TrimSpace(item.ReasonCode) == "" || strings.TrimSpace(item.Message) == "" {
return nil, fmt.Errorf("notarius warning entries require reason_code and message")
if *document.GroupCount < 0 || *document.GroupCount > maxWarningGroups || *document.OccurrenceCount < 0 ||
*document.GroupCount != len(*document.Groups) {
return nil, fmt.Errorf("notarius warning document counts are inconsistent")
}
summaries := make([]WarningSummary, 0, len(*document.Groups))
occurrences := 0
for _, group := range *document.Groups {
if err := validateFindingGroup(group); err != nil {
return nil, fmt.Errorf("notarius warning group: %w", err)
}
summaries = append(summaries, WarningSummary{Scope: item.Scope, ReasonCode: item.ReasonCode})
if group.Disposition != "warning" {
return nil, fmt.Errorf("notarius warning group disposition %q is invalid", group.Disposition)
}
if *group.OccurrenceCount > int(^uint(0)>>1)-occurrences {
return nil, fmt.Errorf("notarius warning occurrence count overflows")
}
occurrences += *group.OccurrenceCount
summaries = append(summaries, WarningSummary{
Disposition: group.Disposition, Category: group.Category, ReasonCode: group.ReasonCode,
Origin: diagnosticOrigin(group.Origin), OccurrenceCount: *group.OccurrenceCount,
})
}
if occurrences != *document.OccurrenceCount {
return nil, fmt.Errorf("notarius warning document occurrence count is inconsistent")
}
return summaries, nil
}
type diagnosticDocument struct {
SchemaVersion string `json:"schema_version"`
GroupCount *int `json:"group_count"`
OccurrenceCount *int `json:"occurrence_count"`
Truncated *bool `json:"truncated"`
UnrepresentedOccurrenceCount *int `json:"unrepresented_occurrence_count"`
Groups *[]findingGroupDocument `json:"groups"`
}
func loadDiagnostics(path string) ([]DiagnosticSummary, int, bool, error) {
var document diagnosticDocument
if err := decodeBoundedJSON(path, maxSummaryBytes, &document); err != nil {
return nil, 0, false, fmt.Errorf("decode notarius diagnostics: %w", err)
}
if document.SchemaVersion != diagnosticsSchemaVersion || document.GroupCount == nil ||
document.OccurrenceCount == nil || document.Truncated == nil ||
document.UnrepresentedOccurrenceCount == nil || document.Groups == nil {
return nil, 0, false, fmt.Errorf("notarius diagnostics document is missing or incompatible required fields")
}
if *document.GroupCount < 0 || *document.GroupCount > maxDiagnosticGroups || *document.OccurrenceCount < 0 ||
*document.UnrepresentedOccurrenceCount < 0 || *document.GroupCount != len(*document.Groups) {
return nil, 0, false, fmt.Errorf("notarius diagnostics document counts are inconsistent")
}
if !*document.Truncated && *document.UnrepresentedOccurrenceCount != 0 {
return nil, 0, false, fmt.Errorf("notarius diagnostics document has unrepresented occurrences without truncation")
}
summaries := make([]DiagnosticSummary, 0, len(*document.Groups))
representedOccurrences := 0
for _, group := range *document.Groups {
if err := validateFindingGroup(group); err != nil {
return nil, 0, false, fmt.Errorf("notarius diagnostic group: %w", err)
}
if group.Disposition != "advisory" && group.Disposition != "observation" {
return nil, 0, false, fmt.Errorf("notarius diagnostic group disposition %q is invalid", group.Disposition)
}
if *group.OccurrenceCount > int(^uint(0)>>1)-representedOccurrences {
return nil, 0, false, fmt.Errorf("notarius diagnostic occurrence count overflows")
}
representedOccurrences += *group.OccurrenceCount
summaries = append(summaries, DiagnosticSummary{
Disposition: group.Disposition, Category: group.Category, ReasonCode: group.ReasonCode,
Origin: diagnosticOrigin(group.Origin), OccurrenceCount: *group.OccurrenceCount,
})
}
if *document.UnrepresentedOccurrenceCount > int(^uint(0)>>1)-representedOccurrences ||
representedOccurrences+*document.UnrepresentedOccurrenceCount != *document.OccurrenceCount {
return nil, 0, false, fmt.Errorf("notarius diagnostics document occurrence count is inconsistent")
}
return summaries, *document.OccurrenceCount, *document.Truncated, nil
}
func validateFindingGroup(group findingGroupDocument) error {
if strings.TrimSpace(group.Disposition) == "" || strings.TrimSpace(group.Category) == "" ||
strings.TrimSpace(group.ReasonCode) == "" || !validDiagnosticOriginStage(group.Origin.Stage) ||
!validDiagnosticCategory(group.Disposition, group.Category) ||
group.OccurrenceCount == nil || *group.OccurrenceCount <= 0 || group.Samples == nil ||
group.OmittedSampleCount == nil || *group.OmittedSampleCount < 0 {
return fmt.Errorf("missing required fields")
}
if len(*group.Samples) == 0 || len(*group.Samples) > maxFindingSamples ||
*group.OmittedSampleCount != *group.OccurrenceCount-len(*group.Samples) {
return fmt.Errorf("sample counts are inconsistent")
}
for _, sample := range *group.Samples {
if strings.TrimSpace(sample.Scope) == "" || strings.TrimSpace(sample.Message) == "" ||
(sample.ChunkIndex != nil && *sample.ChunkIndex < 0) {
return fmt.Errorf("samples require scope and message")
}
}
return nil
}
func validDiagnosticCategory(disposition, category string) bool {
switch disposition {
case "warning":
return category == "configuration" || category == "degradation" ||
category == "validation_incomplete" || category == "fallback"
case "advisory":
return category == "data_quality"
case "observation":
return category == "normalization"
default:
return false
}
}
func validDiagnosticOriginStage(stage string) bool {
switch stage {
case "references", "chunk", "extract", "merge", "normalize":
return true
default:
return false
}
}
func diagnosticOrigin(document diagnosticOriginDocument) DiagnosticOrigin {
return DiagnosticOrigin{
Stage: document.Stage, StepID: document.StepID, LaneID: document.LaneID,
ModuleKey: document.ModuleKey, ValidatorKey: document.ValidatorKey,
}
}
func sumWarningOccurrences(values []WarningSummary) (int, error) {
total := 0
for _, value := range values {
if value.OccurrenceCount > int(^uint(0)>>1)-total {
return 0, fmt.Errorf("notarius warning occurrence count overflows")
}
total += value.OccurrenceCount
}
return total, nil
}
func decodeBoundedJSON(path string, limit int64, destination any) error {
data, err := fileops.ReadRegularFile(path, limit)
if err != nil {