Render fixed Markdown reports and shared completion counts

2026-09-24 17:34:03 +00:00
parent 61af29ecb8
commit 989b28e8c5
4 changed files with 311 additions and 0 deletions
--- a/internal/core/reporting/markdown.go
+++ b/internal/core/reporting/markdown.go
@@ -0,0 +1,180 @@
+package reporting
+
+import (
+	"fmt"
+	"sort"
+	"strings"
+
+	"gitea.maximumdirect.net/eric/audita/internal/core/redaction"
+)
+
+// CompletionLine is the brief successful CLI summary. It uses the same counts
+// as both YAML documents and remains independent of file-output selection.
+func CompletionLine(snapshot Snapshot) string {
+	return fmt.Sprintf("Completed: %d corrections applied; %d work units with errors; %d recoveries.",
+		count(snapshot.Summary.CorrectionsApplied), count(snapshot.Summary.WorkUnitsWithErrors), count(snapshot.Summary.Recoveries))
+}
+
+func count(value *int) int {
+	if value == nil {
+		return 0
+	}
+	return *value
+}
+
+// Markdown renders a fixed-layout human report from one processing snapshot.
+func Markdown(snapshot Snapshot, secrets ...string) []byte {
+	r := redaction.New(secrets)
+	safe := func(value string) string { return escapeMarkdown(r.String(value)) }
+	var b strings.Builder
+	fmt.Fprintf(&b, "# Audita report: %s\n\n", safe(snapshot.RunID))
+	fmt.Fprintf(&b, "Status: %s\n\n", safe(snapshot.Status))
+	fmt.Fprintf(&b, "Corrections applied: %d  \n", count(snapshot.Summary.CorrectionsApplied))
+	fmt.Fprintf(&b, "Work units: %s total; %s completed; %s interrupted; %s not started  \n",
+		showCount(snapshot.Summary.WorkUnitsTotal), showCount(snapshot.Summary.WorkUnitsCompleted),
+		showCount(snapshot.Summary.WorkUnitsInterrupted), showCount(snapshot.Summary.WorkUnitsNotStarted))
+	fmt.Fprintf(&b, "Errors: %s work units; %s run errors  \n", showCount(snapshot.Summary.WorkUnitsWithErrors), showCount(snapshot.Summary.RunErrors))
+	fmt.Fprintf(&b, "Recoveries: %s of %s model operations  \n", showCount(snapshot.Summary.Recoveries), showCount(snapshot.Summary.LLMOperations))
+	fmt.Fprintf(&b, "Proposal responses: %s; with filtering: %s\n\n",
+		showCount(snapshot.Summary.ProposalResponses), showCount(snapshot.Summary.ProposalResponsesWithFiltering))
+	fmt.Fprintf(&b, "Transcript input: %s  \n", safe(snapshot.TranscriptPath))
+	fmt.Fprintf(&b, "Glossary input: %s  \n", safe(snapshot.GlossaryPath))
+	output := snapshot.OutputPath
+	if output == "" {
+		output = "stdout"
+	}
+	fmt.Fprintf(&b, "Transcript output: %s\n\n", safe(output))
+
+	fmt.Fprintln(&b, "## Work-unit errors")
+	fmt.Fprintln(&b)
+	writeMatrix(&b, snapshot, safe, false)
+	fmt.Fprintln(&b)
+	fmt.Fprintln(&b, "## Recoveries")
+	fmt.Fprintln(&b)
+	writeMatrix(&b, snapshot, safe, true)
+	fmt.Fprintln(&b)
+	fmt.Fprintln(&b, "Legend: `—` = not started; `n*` = interrupted with n recorded errors or recoveries. Completed error cells are 0 or 1.")
+	fmt.Fprintln(&b)
+	writeReasons(&b, "Error reasons", errorReasons(snapshot.Errors), safe)
+	writeReasons(&b, "Recovery reasons", recoveryReasons(snapshot.Recoveries), safe)
+	fmt.Fprintln(&b, "A recovery can have multiple reasons; reason counts need not sum to the recovery total.")
+	fmt.Fprintln(&b)
+	fmt.Fprintln(&b, "Details: `errors.yml` and `recoveries.yml`.")
+	return []byte(b.String())
+}
+
+func showCount(value *int) string {
+	if value == nil {
+		return "unknown"
+	}
+	return fmt.Sprint(*value)
+}
+
+func writeMatrix(b *strings.Builder, snapshot Snapshot, safe func(string) string, recovery bool) {
+	if len(snapshot.StageNames) == 0 || len(snapshot.WorkUnits) == 0 {
+		fmt.Fprintln(b, "No work units.")
+		return
+	}
+	columns := snapshot.StageNames
+	fmt.Fprint(b, "| Chunk |")
+	for _, name := range columns {
+		fmt.Fprintf(b, " %s |", safe(name))
+	}
+	fmt.Fprint(b, "\n| --- |")
+	for range columns {
+		fmt.Fprint(b, " ---: |")
+	}
+	fmt.Fprintln(b)
+	byChunk := make(map[int]map[int]WorkUnit)
+	chunkMeta := make(map[int]WorkUnit)
+	keys := make([]int, 0)
+	for _, unit := range snapshot.WorkUnits {
+		if byChunk[unit.ChunkIndex] == nil {
+			byChunk[unit.ChunkIndex] = make(map[int]WorkUnit)
+			keys = append(keys, unit.ChunkIndex)
+		}
+		byChunk[unit.ChunkIndex][unit.StageIndex] = unit
+		chunkMeta[unit.ChunkIndex] = unit
+	}
+	sort.Ints(keys)
+	for _, chunk := range keys {
+		meta := chunkMeta[chunk]
+		fmt.Fprintf(b, "| Chunk %d (segments %d–%d) |", chunk+1, meta.StartSegmentID, meta.EndSegmentID)
+		for stage := range columns {
+			unit, exists := byChunk[chunk][stage]
+			cell := "—"
+			if exists && unit.State != "not_started" {
+				n := unit.ErrorCount
+				if recovery {
+					n = unit.RecoveryCount
+				}
+				cell = fmt.Sprint(n)
+				if unit.State == "interrupted" {
+					cell += "*"
+				}
+			}
+			fmt.Fprintf(b, " %s |", cell)
+		}
+		fmt.Fprintln(b)
+	}
+}
+
+func writeReasons(b *strings.Builder, title string, counts map[string]int, safe func(string) string) {
+	fmt.Fprintf(b, "### %s\n\n", title)
+	if len(counts) == 0 {
+		fmt.Fprintln(b, "None.")
+		fmt.Fprintln(b)
+		return
+	}
+	keys := make([]string, 0, len(counts))
+	for key := range counts {
+		keys = append(keys, key)
+	}
+	sort.Strings(keys)
+	for _, key := range keys {
+		fmt.Fprintf(b, "- `%s`: %d\n", safe(key), counts[key])
+	}
+	fmt.Fprintln(b)
+}
+
+func errorReasons(records []ErrorRecord) map[string]int {
+	counts := make(map[string]int)
+	for _, record := range records {
+		seen := make(map[string]bool)
+		if record.Scope == "run" {
+			seen[record.Reason] = true
+		}
+		for _, failure := range record.Failures {
+			seen[failure.Reason] = true
+		}
+		for reason := range seen {
+			if reason != "" {
+				counts[reason]++
+			}
+		}
+	}
+	return counts
+}
+
+func recoveryReasons(records []RecoveryRecord) map[string]int {
+	counts := make(map[string]int)
+	for _, record := range records {
+		seen := make(map[string]bool)
+		for _, reason := range record.Reasons {
+			seen[reason] = true
+		}
+		for reason := range seen {
+			counts[reason]++
+		}
+	}
+	return counts
+}
+
+func escapeMarkdown(value string) string {
+	value = strings.ReplaceAll(value, "\\", "\\\\")
+	value = strings.ReplaceAll(value, "|", "\\|")
+	value = strings.ReplaceAll(value, "`", "\\`")
+	value = strings.ReplaceAll(value, "\r\n", "\n")
+	value = strings.ReplaceAll(value, "\r", "\n")
+	return strings.ReplaceAll(value, "\n", "<br>")
+}
--- a/internal/core/reporting/markdown_test.go
+++ b/internal/core/reporting/markdown_test.go
@@ -0,0 +1,79 @@
+package reporting
+
+import (
+	"bytes"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"time"
+
+	"gopkg.in/yaml.v3"
+)
+
+func TestMarkdownFixedLayoutAndCounts(t *testing.T) {
+	one, two, three, four, zero := 1, 2, 3, 4, 0
+	snapshot := Snapshot{
+		RunID: "run-123", Status: StatusCompleted, StartedAt: time.Date(2026, 9, 24, 10, 0, 0, 0, time.UTC),
+		TranscriptPath: "in|put.json", GlossaryPath: "glossary`secret`.yaml", OutputPath: "out\nfile.json",
+		StageNames: []string{"grammar_1", "grammar_2"},
+		WorkUnits: []WorkUnit{
+			{StageIndex: 0, ModuleInstance: "grammar_1", ChunkIndex: 0, StartSegmentID: 1, EndSegmentID: 2, State: "completed", ErrorCount: 1, RecoveryCount: 1, CorrectionsApplied: 1},
+			{StageIndex: 1, ModuleInstance: "grammar_2", ChunkIndex: 0, StartSegmentID: 1, EndSegmentID: 2, State: "interrupted", RecoveryCount: 2},
+			{StageIndex: 0, ModuleInstance: "grammar_1", ChunkIndex: 1, StartSegmentID: 3, EndSegmentID: 4, State: "completed"},
+			{StageIndex: 1, ModuleInstance: "grammar_2", ChunkIndex: 1, StartSegmentID: 3, EndSegmentID: 4, State: "not_started"},
+		},
+		Summary: Summary{WorkUnitsTotal: &four, WorkUnitsCompleted: &two, WorkUnitsInterrupted: &one, WorkUnitsNotStarted: &one,
+			WorkUnitsWithErrors: &one, RunErrors: &zero, CorrectionsApplied: &one, LLMOperations: &four,
+			ProposalResponses: &one, ProposalResponsesWithFiltering: &one, Recoveries: &three},
+		Errors:     []ErrorRecord{{Scope: "work_unit", Failures: []Failure{{Reason: "transport_failed"}, {Reason: "transport_failed"}}}},
+		Recoveries: []RecoveryRecord{{Reasons: []string{"output_repaired", "proposals_filtered"}}, {Reasons: []string{"output_repaired"}}, {Reasons: []string{"service_recovered"}}},
+	}
+	got := Markdown(snapshot, "secret")
+	fixture := filepath.Join("testdata", "report.md")
+	if os.Getenv("UPDATE_REPORT_GOLDEN") == "1" {
+		if err := os.WriteFile(fixture, got, 0600); err != nil {
+			t.Fatal(err)
+		}
+	}
+	want, err := os.ReadFile(fixture)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !bytes.Equal(got, want) {
+		t.Fatalf("Markdown differs from %s:\n%s", fixture, got)
+	}
+	text := string(got)
+	for _, fragment := range []string{"| grammar_1 | grammar_2 |", "| Chunk 1 (segments 1–2) | 1 | 0* |", "| Chunk 1 (segments 1–2) | 1 | 2* |", "| Chunk 2 (segments 3–4) | 0 | — |", "in\\|put.json", "out<br>file.json", "glossary\\`[REDACTED]\\`.yaml", "- `transport_failed`: 1", "- `output_repaired`: 2"} {
+		if !strings.Contains(text, fragment) {
+			t.Fatalf("missing %q in:\n%s", fragment, text)
+		}
+	}
+	if strings.Contains(text, "secret") {
+		t.Fatal("secret leaked")
+	}
+	var decoded struct {
+		Summary Summary `yaml:"summary"`
+	}
+	yamlBytes, err := YAML(snapshot.ErrorsDocument())
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := yaml.Unmarshal(yamlBytes, &decoded); err != nil {
+		t.Fatal(err)
+	}
+	if CompletionLine(snapshot) != "Completed: 1 corrections applied; 1 work units with errors; 3 recoveries." || *decoded.Summary.WorkUnitsWithErrors != 1 || *decoded.Summary.Recoveries != 3 {
+		t.Fatalf("CLI/YAML count disagreement: %s %+v", CompletionLine(snapshot), decoded.Summary)
+	}
+}
+
+func TestMarkdownNoWorkAndEscapedStageLabel(t *testing.T) {
+	text := string(Markdown(Snapshot{RunID: "r", StageNames: []string{"bad|label`\nsecret"}}, "secret"))
+	if strings.Count(text, "No work units.") != 2 || strings.Contains(text, "secret") {
+		t.Fatal(text)
+	}
+	text = string(Markdown(Snapshot{RunID: "r", StageNames: []string{"bad|label`\nsecret"}, WorkUnits: []WorkUnit{{StageIndex: 0, ChunkIndex: 0, State: "completed"}}}, "secret"))
+	if !strings.Contains(text, "bad\\|label\\`<br>[REDACTED]") {
+		t.Fatal(text)
+	}
+}
--- a/internal/core/reporting/testdata/report.md
+++ b/internal/core/reporting/testdata/report.md
@@ -0,0 +1,43 @@
+# Audita report: run-123
+
+Status: completed
+
+Corrections applied: 1  
+Work units: 4 total; 2 completed; 1 interrupted; 1 not started  
+Errors: 1 work units; 0 run errors  
+Recoveries: 3 of 4 model operations  
+Proposal responses: 1; with filtering: 1
+
+Transcript input: in\|put.json  
+Glossary input: glossary\`[REDACTED]\`.yaml  
+Transcript output: out<br>file.json
+
+## Work-unit errors
+
+| Chunk | grammar_1 | grammar_2 |
+| --- | ---: | ---: |
+| Chunk 1 (segments 1–2) | 1 | 0* |
+| Chunk 2 (segments 3–4) | 0 | — |
+
+## Recoveries
+
+| Chunk | grammar_1 | grammar_2 |
+| --- | ---: | ---: |
+| Chunk 1 (segments 1–2) | 1 | 2* |
+| Chunk 2 (segments 3–4) | 0 | — |
+
+Legend: `—` = not started; `n*` = interrupted with n recorded errors or recoveries. Completed error cells are 0 or 1.
+
+### Error reasons
+
+- `transport_failed`: 1
+
+### Recovery reasons
+
+- `output_repaired`: 2
+- `proposals_filtered`: 1
+- `service_recovered`: 1
+
+A recovery can have multiple reasons; reason counts need not sum to the recovery total.
+
+Details: `errors.yml` and `recoveries.yml`.