Add deterministic analysis input identities

This commit is contained in:
2026-08-29 19:02:16 +00:00
parent 62de6abdbf
commit ab5751459a
12 changed files with 815 additions and 150 deletions

View File

@@ -8,6 +8,7 @@ import (
"path/filepath"
"strings"
"gitea.maximumdirect.net/eric/narratio/internal/artifactmodel"
"gitea.maximumdirect.net/eric/narratio/internal/fileops"
"gitea.maximumdirect.net/eric/narratio/internal/manifest"
"gitea.maximumdirect.net/eric/narratio/internal/pathsafe"
@@ -30,6 +31,9 @@ type AnalyzeEvidence struct {
SourceID string
Path string
ProducerRunID string
Contract *artifactmodel.ContractMetadata
Checksum string
Size int64
}
// InspectAnalyzeEvidence verifies that one configured artifact has supported,
@@ -107,6 +111,9 @@ func InspectAnalyzeEvidence(
SourceID: sourceID,
Path: filepath.Join(paths.Root, filepath.FromSlash(configuredPath)),
ProducerRunID: record.ProducerRunID,
Contract: cloneArtifactContract(record.Output.Contract),
Checksum: record.Output.Checksum,
Size: record.OutputSize,
}
}
@@ -129,6 +136,9 @@ func (c *ArtifactCatalog) HydrateAnalyzeArtifacts(
if evidence.State != AnalyzeEvidenceCurrent {
continue
}
_ = c.markAvailableFromAnalyzeManifest(evidence.SourceID, evidence.Path, evidence.ProducerRunID)
_ = c.markAvailableFromAnalyzeManifest(
evidence.SourceID, evidence.Path, evidence.ProducerRunID,
evidence.Checksum, evidence.Size, evidence.Contract,
)
}
}

View File

@@ -157,6 +157,18 @@ func TestHydrateAnalyzeArtifactsUsesOnlyCurrentConfiguredKeys(t *testing.T) {
if !entry.Available || entry.Provenance != ArtifactProvenanceCurrentAnalyzeManifest || entry.ProducerRunID != "run-1" {
t.Fatalf("entry = %#v", entry)
}
record := m.Stages["analyze"].AnalyzeArtifacts["session_recap"]
if entry.Checksum != record.Output.Checksum || entry.Size != record.OutputSize {
t.Fatalf("entry content identity = (%q, %d), want (%q, %d)", entry.Checksum, entry.Size, record.Output.Checksum, record.OutputSize)
}
if entry.Contract == nil || *entry.Contract != *record.Output.Contract {
t.Fatalf("entry contract = %#v, want %#v", entry.Contract, record.Output.Contract)
}
entry.Contract.SchemaVersion = "mutated"
again, _ := catalog.Lookup(ConfiguredArtifactSourceID("session_recap"))
if again.Contract == nil || again.Contract.SchemaVersion != "1" {
t.Fatalf("catalog contract was mutated through lookup: %#v", again.Contract)
}
if _, ok := catalog.Lookup(ConfiguredArtifactSourceID("removed")); ok {
t.Fatal("removed manifest record was advertised in current catalog")
}

View File

@@ -107,6 +107,9 @@ type ResolvedSessionArtifact struct {
OutputKind string
ProducerRunID string
Provenance string
Contract *artifactmodel.ContractMetadata
Checksum string
Size int64
}
// SessionArtifactNotFoundError includes context when a known artifact cannot be read.
@@ -259,6 +262,9 @@ func ResolveSessionArtifactWithCatalog(paths SessionPaths, m *manifest.Manifest,
OutputKind: entry.OutputKind,
ProducerRunID: entry.ProducerRunID,
Provenance: entry.Provenance,
Contract: cloneArtifactContract(entry.Contract),
Checksum: entry.Checksum,
Size: entry.Size,
}, nil
}

View File

@@ -5,6 +5,7 @@ import (
"sort"
"strings"
"gitea.maximumdirect.net/eric/narratio/internal/artifactmodel"
"gitea.maximumdirect.net/eric/narratio/internal/artifactpolicy"
"gitea.maximumdirect.net/eric/narratio/internal/config"
)
@@ -64,6 +65,9 @@ type CatalogEntry struct {
Path string
Provenance string
ProducerRunID string
Contract *artifactmodel.ContractMetadata
Checksum string
Size int64
}
// ArtifactCatalog tracks built-in, configured, and extraction artifact definitions and runtime state.
@@ -98,6 +102,7 @@ func (c *ArtifactCatalog) RegisterExtractionArtifacts(configured map[string]Extr
if _, exists := c.extractionIndex[trimmed]; exists {
return fmt.Errorf("duplicate extraction artifact key %q", trimmed)
}
def := configured[key]
sourceID := ExtractionArtifactSourceID(trimmed)
if err := c.addEntry(CatalogEntry{
SourceID: sourceID,
@@ -105,6 +110,10 @@ func (c *ArtifactCatalog) RegisterExtractionArtifacts(configured map[string]Extr
ProducerStage: "extract",
OutputKind: "notarius_lane",
Planned: true,
Contract: &artifactmodel.ContractMetadata{
MediaType: def.MediaType, SchemaID: def.SchemaID,
SchemaVersion: def.SchemaVersion, ModuleKey: def.ModuleKey,
},
}); err != nil {
return fmt.Errorf("register extraction artifact %q: %w", trimmed, err)
}
@@ -221,7 +230,7 @@ func (c *ArtifactCatalog) Lookup(sourceID string) (CatalogEntry, bool) {
return CatalogEntry{}, false
}
entry, ok := c.entries[strings.TrimSpace(sourceID)]
return entry, ok
return cloneCatalogEntry(entry), ok
}
// SourceIDForConfiguredKey returns canonical source ID for one configured key.
@@ -255,7 +264,7 @@ func (c *ArtifactCatalog) ListConfigured() []CatalogEntry {
out := make([]CatalogEntry, 0, len(keys))
for _, key := range keys {
sourceID := c.configuredIndex[key]
out = append(out, c.entries[sourceID])
out = append(out, cloneCatalogEntry(c.entries[sourceID]))
}
return out
}
@@ -272,7 +281,7 @@ func (c *ArtifactCatalog) ListExtraction() []CatalogEntry {
sort.Strings(keys)
out := make([]CatalogEntry, 0, len(keys))
for _, key := range keys {
out = append(out, c.entries[c.extractionIndex[key]])
out = append(out, cloneCatalogEntry(c.entries[c.extractionIndex[key]]))
}
return out
}
@@ -282,22 +291,36 @@ func (c *ArtifactCatalog) MarkAvailableGenerated(sourceID, path string) error {
return c.markAvailable(sourceID, path, ArtifactProvenanceGeneratedCurrentAnalyzeRun)
}
func (c *ArtifactCatalog) markAvailableFromExtractManifest(sourceID, path, producerRunID string) error {
func (c *ArtifactCatalog) markAvailableFromExtractManifest(
sourceID, path, producerRunID, checksum string,
size int64,
contract *artifactmodel.ContractMetadata,
) error {
if err := c.markAvailable(sourceID, path, ArtifactProvenanceCurrentExtractManifest); err != nil {
return err
}
entry := c.entries[strings.TrimSpace(sourceID)]
entry.ProducerRunID = strings.TrimSpace(producerRunID)
entry.Checksum = strings.TrimSpace(checksum)
entry.Size = size
entry.Contract = cloneArtifactContract(contract)
c.entries[entry.SourceID] = entry
return nil
}
func (c *ArtifactCatalog) markAvailableFromAnalyzeManifest(sourceID, path, producerRunID string) error {
func (c *ArtifactCatalog) markAvailableFromAnalyzeManifest(
sourceID, path, producerRunID, checksum string,
size int64,
contract *artifactmodel.ContractMetadata,
) error {
if err := c.markAvailable(sourceID, path, ArtifactProvenanceCurrentAnalyzeManifest); err != nil {
return err
}
entry := c.entries[strings.TrimSpace(sourceID)]
entry.ProducerRunID = strings.TrimSpace(producerRunID)
entry.Checksum = strings.TrimSpace(checksum)
entry.Size = size
entry.Contract = cloneArtifactContract(contract)
c.entries[entry.SourceID] = entry
return nil
}
@@ -318,10 +341,29 @@ func (c *ArtifactCatalog) markAvailable(sourceID, path, provenance string) error
entry.Available = true
entry.Path = trimmedPath
entry.Provenance = provenance
entry.ProducerRunID = ""
entry.Checksum = ""
entry.Size = 0
if provenance == ArtifactProvenanceGeneratedCurrentAnalyzeRun {
entry.Contract = nil
}
c.entries[normalizedID] = entry
return nil
}
func cloneCatalogEntry(entry CatalogEntry) CatalogEntry {
entry.Contract = cloneArtifactContract(entry.Contract)
return entry
}
func cloneArtifactContract(contract *artifactmodel.ContractMetadata) *artifactmodel.ContractMetadata {
if contract == nil {
return nil
}
clone := *contract
return &clone
}
func (c *ArtifactCatalog) addEntry(entry CatalogEntry) error {
if c == nil {
return fmt.Errorf("artifact catalog is nil")

View File

@@ -39,8 +39,11 @@ func (c *ArtifactCatalog) HydrateExtractionArtifacts(
if proof.State != ExtractionEvidenceValid {
return
}
for sourceID, path := range proof.Outputs {
_ = c.markAvailableFromExtractManifest(sourceID, path, proof.ProducerRunID)
for sourceID, output := range proof.Outputs {
_ = c.markAvailableFromExtractManifest(
sourceID, output.Path, proof.ProducerRunID,
output.Checksum, output.Size, output.Contract,
)
}
}

View File

@@ -68,6 +68,17 @@ func TestHydrateExtractionArtifactsAcceptsOnlyCompleteCurrentBundle(t *testing.T
if entry.Provenance != ArtifactProvenanceCurrentExtractManifest || entry.ProducerRunID != "extract-run-1" {
t.Fatalf("hydrated provenance = %#v", entry)
}
wantOutput := currentManifest.Stages["extract"].Outputs[0]
info, err := os.Stat(wantOutput.LocalPath)
if err != nil {
t.Fatal(err)
}
if entry.Checksum != wantOutput.Checksum || entry.Size != info.Size() {
t.Fatalf("hydrated content identity = (%q, %d), want (%q, %d)", entry.Checksum, entry.Size, wantOutput.Checksum, info.Size())
}
if entry.Contract == nil || *entry.Contract != *wantOutput.Contract {
t.Fatalf("hydrated contract = %#v, want %#v", entry.Contract, wantOutput.Contract)
}
resolved, err := ResolveSessionArtifactWithCatalog(paths, currentManifest, entry.SourceID, catalog)
if err != nil {
t.Fatalf("ResolveSessionArtifactWithCatalog() error = %v", err)

View File

@@ -6,6 +6,7 @@ import (
"path/filepath"
"strings"
"gitea.maximumdirect.net/eric/narratio/internal/artifactmodel"
"gitea.maximumdirect.net/eric/narratio/internal/fileops"
"gitea.maximumdirect.net/eric/narratio/internal/manifest"
)
@@ -23,7 +24,15 @@ const (
type ExtractionEvidence struct {
State ExtractionEvidenceState
Reason, ProducerRunID string
Outputs map[string]string
Outputs map[string]ExtractionEvidenceOutput
}
// ExtractionEvidenceOutput is the verified semantic identity of one lane.
type ExtractionEvidenceOutput struct {
Path string
Checksum string
Size int64
Contract *artifactmodel.ContractMetadata
}
// InspectExtractionEvidence verifies structure, confinement, identities, contracts, and payload bytes.
@@ -72,7 +81,7 @@ func InspectExtractionEvidence(
for key, d := range configured {
expected[ExtractionArtifactSourceID(key)] = d
}
seen, outputs := map[string]struct{}{}, map[string]string{}
seen, outputs := map[string]struct{}{}, map[string]ExtractionEvidenceOutput{}
indexSeen := false
for _, out := range r.Outputs {
if strings.TrimSpace(out.ProducerRunID) != runID {
@@ -82,7 +91,7 @@ func InspectExtractionEvidence(
if indexSeen || out.Kind != extractionIndexKind || filepath.Clean(out.LocalPath) != filepath.Join(root, "index.json") {
return ExtractionEvidence{State: ExtractionEvidenceObsolete, Reason: "extract index path is not canonical"}
}
if state, reason := inspectExtractionPayload(root, out.LocalPath, out.Checksum); state != ExtractionEvidenceValid {
if state, reason, _ := inspectExtractionPayload(root, out.LocalPath, out.Checksum); state != ExtractionEvidenceValid {
return ExtractionEvidence{State: state, Reason: reason}
}
indexSeen = true
@@ -98,11 +107,15 @@ func InspectExtractionEvidence(
if !compatibleCatalogExtractionContract(out.Contract, d) || !compatibleCatalogExtractionProvenance(out.ExternalProvenance, receiptRunID, receiptPipelineID, d) {
return ExtractionEvidence{State: ExtractionEvidenceObsolete, Reason: "extract output contract or provenance is incompatible"}
}
if state, reason := inspectExtractionPayload(root, out.LocalPath, out.Checksum); state != ExtractionEvidenceValid {
state, reason, size := inspectExtractionPayload(root, out.LocalPath, out.Checksum)
if state != ExtractionEvidenceValid {
return ExtractionEvidence{State: state, Reason: reason}
}
seen[out.SourceID] = struct{}{}
outputs[out.SourceID] = out.LocalPath
outputs[out.SourceID] = ExtractionEvidenceOutput{
Path: out.LocalPath, Checksum: out.Checksum, Size: size,
Contract: cloneArtifactContract(out.Contract),
}
}
if !indexSeen || len(seen) != len(expected) || len(r.Outputs) != len(expected)+1 {
return ExtractionEvidence{State: ExtractionEvidenceObsolete, Reason: "extract result is incomplete"}
@@ -110,30 +123,30 @@ func InspectExtractionEvidence(
return ExtractionEvidence{State: ExtractionEvidenceValid, ProducerRunID: runID, Outputs: outputs}
}
func inspectExtractionPayload(root, path, checksum string) (ExtractionEvidenceState, string) {
func inspectExtractionPayload(root, path, checksum string) (ExtractionEvidenceState, string, int64) {
if !filepath.IsAbs(path) || !pathWithinExtractionRoot(root, path) || strings.TrimSpace(checksum) == "" {
return ExtractionEvidenceUnsafe, "extract output path or checksum is unsafe"
return ExtractionEvidenceUnsafe, "extract output path or checksum is unsafe", 0
}
info, err := os.Lstat(path)
if os.IsNotExist(err) {
return ExtractionEvidenceObsolete, "extract output is missing"
return ExtractionEvidenceObsolete, "extract output is missing", 0
}
if !safeExtractionComponents(root, path) {
return ExtractionEvidenceUnsafe, "extract output path contains unsafe components"
return ExtractionEvidenceUnsafe, "extract output path contains unsafe components", 0
}
if err != nil || !info.Mode().IsRegular() || info.Mode()&os.ModeSymlink != 0 {
return ExtractionEvidenceUnsafe, "extract output is not a regular file"
return ExtractionEvidenceUnsafe, "extract output is not a regular file", 0
}
actual, err := SHA256File(path)
if err != nil {
return ExtractionEvidenceUnsafe, "extract output checksum cannot be read"
return ExtractionEvidenceUnsafe, "extract output checksum cannot be read", 0
}
if actual != checksum {
return ExtractionEvidenceObsolete, "extract output checksum does not match durable bytes"
return ExtractionEvidenceObsolete, "extract output checksum does not match durable bytes", 0
}
body, err := fileops.ReadRegularFile(path, MaxExtractionPayloadBytes)
if err != nil || !json.Valid(body) {
return ExtractionEvidenceObsolete, "extract output is not valid JSON"
return ExtractionEvidenceObsolete, "extract output is not valid JSON", 0
}
return ExtractionEvidenceValid, ""
return ExtractionEvidenceValid, "", info.Size()
}