Produce the requested section count when feasible
This commit is contained in:
@@ -1192,6 +1192,20 @@ func TestSubprocessTranscriptPreflightWithoutReporting(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
func TestSubprocessInfeasibleSectionTargetIsAnInputFailure(t *testing.T) {
|
||||
f := newProcessFixture(t, "normal", "[grammar]")
|
||||
code, stdout, stderr := f.run(t, true, "--target-sections", "10")
|
||||
if code != 2 || stdout != "" || f.provider.callCount() != 0 || !strings.Contains(stderr, "target_sections") {
|
||||
t.Fatalf("infeasible target: code=%d stdout=%q stderr=%q calls=%d", code, stdout, stderr, f.provider.callCount())
|
||||
}
|
||||
if _, err := os.Stat(f.output); !os.IsNotExist(err) {
|
||||
t.Fatalf("unexpected transcript output: %v", err)
|
||||
}
|
||||
if document := f.errors(t); document.Status != reporting.StatusFailed {
|
||||
t.Fatalf("report status: %+v", document)
|
||||
}
|
||||
}
|
||||
|
||||
func TestSubprocessOutputPermissionsHonorUmask(t *testing.T) {
|
||||
for _, mask := range []string{"0077", "0022"} {
|
||||
t.Run(mask, func(t *testing.T) {
|
||||
|
||||
@@ -78,8 +78,8 @@ specified by the [editorial acceptance boundary](policy/architecture.md#correcti
|
||||
| `llm.operation_timeout` | `15m` | Positive duration per logical operation. |
|
||||
| `validation.max_prompt_tokens` | `2048` | Positive validation batching budget. |
|
||||
| `context.description` | empty | Prompt context; at most 500 bytes after trimming. |
|
||||
| `chunking.target_sections` | unset | Optional positive target. |
|
||||
| `chunking.max_section_tokens`, `chunking.min_section_tokens` | `8192`, `2048` | Positive section bounds; minimum must not exceed maximum. |
|
||||
| `chunking.target_sections` | unset | Optional positive count. Processing requires exactly this many nonempty sections when specified. An infeasible count is an input error. |
|
||||
| `chunking.max_section_tokens`, `chunking.min_section_tokens` | `8192`, `2048` | The maximum bounds a section's estimated tokens, except for an oversized segment in a section of its own. The minimum is validated against the maximum and retained for compatibility. |
|
||||
| `normalization.max_segment_gap`, `normalization.ellipsis_gap` | `4`, `3.5` | Nonnegative seconds. |
|
||||
| `normalization.max_segment_duration`, `normalization.max_segment_tokens` | `60`, `2048` | Positive seconds and token estimate. |
|
||||
| `output.schema` | `bare-segments` | Output format; see [output contracts](integrations/output-files.md). |
|
||||
|
||||
@@ -277,6 +277,12 @@ Verify the audited weights `[2,5,1]` with target 2 and maximum 6, impossible
|
||||
targets, and the oversized singleton case. Check contiguity, order, coverage,
|
||||
and feasibility independently of the chosen cut. Keep a single chunking pass.
|
||||
|
||||
**Complete.** Explicit targets use rolling legal ends and suffix feasibility to
|
||||
admit exactly the requested count when possible. An independent enumeration
|
||||
checked 1,024 weight sequences for every possible count, including zero-token
|
||||
and oversized segments. The real CLI returned status 2 for infeasibility
|
||||
without dispatching model work. The CLI build and full offline suite passed.
|
||||
|
||||
## Stage 8 — Share edit location and protection evidence
|
||||
|
||||
Audit #12, #18 (punctuation portion), #19, #38, #47.
|
||||
|
||||
@@ -6,6 +6,7 @@ import (
|
||||
"fmt"
|
||||
"time"
|
||||
|
||||
"gitea.maximumdirect.net/eric/audita/v3/internal/core/chunking"
|
||||
"gitea.maximumdirect.net/eric/audita/v3/internal/core/diagnostics"
|
||||
"gitea.maximumdirect.net/eric/audita/v3/internal/core/reporting"
|
||||
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/outcomes"
|
||||
@@ -38,6 +39,11 @@ func finishProcess(inv commandkit.Invocation, input processInput, prepared *prep
|
||||
}
|
||||
result := prepared.Prepared.Output.Finish(inv.Context, primary, transcript, render)
|
||||
if result != nil {
|
||||
var infeasible *chunking.InfeasibleTargetError
|
||||
var outputFailure diagnostics.FileFailure
|
||||
if inv.Context.Err() == nil && errors.As(result, &infeasible) && !errors.As(result, &outputFailure) {
|
||||
return commandkit.Usage(infeasible.Error()).WithCause(result)
|
||||
}
|
||||
return commandkit.Failure(result.Error()).WithCause(result)
|
||||
}
|
||||
_ = reporter.AtEnd().Info(reporting.CompletionLine(snapshot))
|
||||
|
||||
@@ -2,6 +2,7 @@ package chunking
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"slices"
|
||||
|
||||
"gitea.maximumdirect.net/eric/audita/v3/internal/core/schema"
|
||||
)
|
||||
@@ -100,66 +101,35 @@ func (c *Chunker) ChunkTranscript(transcript *schema.Transcript) ([]Section, err
|
||||
totalTokens += segmentTokens[i]
|
||||
}
|
||||
|
||||
minPossibleSections := c.calculateMinPossibleSections(segmentTokens)
|
||||
|
||||
var desiredSections int
|
||||
useExplicitTarget := false
|
||||
if c.config.TargetSections != nil {
|
||||
desiredSections = *c.config.TargetSections
|
||||
useExplicitTarget = true
|
||||
if desiredSections <= 0 {
|
||||
return nil, fmt.Errorf("target_sections must be positive, got %d", desiredSections)
|
||||
}
|
||||
if err := c.validateTargetSections(desiredSections, segmentTokens); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
} else {
|
||||
desiredSections = ceilDiv(totalTokens, c.config.MaxSectionTokens)
|
||||
if desiredSections < minPossibleSections {
|
||||
desiredSections = minPossibleSections
|
||||
}
|
||||
if desiredSections < 1 {
|
||||
desiredSections = 1
|
||||
}
|
||||
if desiredSections > len(transcript.Segments) {
|
||||
desiredSections = len(transcript.Segments)
|
||||
target := *c.config.TargetSections
|
||||
if target <= 0 {
|
||||
return nil, fmt.Errorf("target_sections must be positive, got %d", target)
|
||||
}
|
||||
return c.buildSectionsWithExplicitTarget(transcript.Segments, segmentTokens, target, ceilDiv(totalTokens, target))
|
||||
}
|
||||
|
||||
targetTokensPerSection := ceilDiv(totalTokens, desiredSections)
|
||||
if useExplicitTarget {
|
||||
return c.buildSectionsWithExplicitTarget(
|
||||
transcript.Segments,
|
||||
segmentTokens,
|
||||
desiredSections,
|
||||
targetTokensPerSection,
|
||||
)
|
||||
}
|
||||
|
||||
return c.buildSectionsBalanced(transcript.Segments, segmentTokens, targetTokensPerSection), nil
|
||||
desiredSections := ceilDiv(totalTokens, c.config.MaxSectionTokens)
|
||||
desiredSections = max(desiredSections, c.calculateMinPossibleSections(segmentTokens), 1)
|
||||
desiredSections = min(desiredSections, len(transcript.Segments))
|
||||
return c.buildSectionsBalanced(transcript.Segments, segmentTokens, ceilDiv(totalTokens, desiredSections)), nil
|
||||
}
|
||||
|
||||
// validateTargetSections checks if the target section count is achievable.
|
||||
func (c *Chunker) validateTargetSections(target int, segmentTokens []int) error {
|
||||
// Maximum possible sections: limited by segment count
|
||||
maxPossibleSections := len(segmentTokens)
|
||||
// InfeasibleTargetError indicates that a requested count cannot partition the
|
||||
// normalized segments into nonempty, token-bounded sections.
|
||||
type InfeasibleTargetError struct {
|
||||
Target int
|
||||
SegmentCount int
|
||||
MinimumSections int
|
||||
MaxSectionTokens int
|
||||
}
|
||||
|
||||
if target > maxPossibleSections {
|
||||
return fmt.Errorf(
|
||||
"target_sections (%d) is impossible: cannot have more sections than segments (%d)",
|
||||
target, maxPossibleSections)
|
||||
func (e *InfeasibleTargetError) Error() string {
|
||||
if e.Target > e.SegmentCount {
|
||||
return fmt.Sprintf("target_sections (%d) is impossible: cannot have more sections than segments (%d)", e.Target, e.SegmentCount)
|
||||
}
|
||||
|
||||
// Minimum possible sections: each segment must fit within max bounds
|
||||
minPossibleSections := c.calculateMinPossibleSections(segmentTokens)
|
||||
|
||||
if target < minPossibleSections {
|
||||
return fmt.Errorf(
|
||||
"target_sections (%d) is impossible: need at least %d sections to respect max_section_tokens (%d)",
|
||||
target, minPossibleSections, c.config.MaxSectionTokens)
|
||||
}
|
||||
|
||||
return nil
|
||||
return fmt.Sprintf("target_sections (%d) is impossible: need at least %d sections to respect max_section_tokens (%d)",
|
||||
e.Target, e.MinimumSections, e.MaxSectionTokens)
|
||||
}
|
||||
|
||||
// buildSectionsBalanced creates sections with a deterministic single-pass policy.
|
||||
@@ -245,59 +215,63 @@ func (c *Chunker) calculateMinPossibleSections(segmentTokens []int) int {
|
||||
return sections
|
||||
}
|
||||
|
||||
// buildSectionsWithExplicitTarget builds exactly desiredSections when feasible.
|
||||
// farthestLegalEnds records the exclusive end of the longest legal section
|
||||
// starting at each segment, with oversized segments admitted only alone.
|
||||
func farthestLegalEnds(segmentTokens []int, maximum int) []int {
|
||||
ends := make([]int, len(segmentTokens))
|
||||
end, windowTokens := 0, 0
|
||||
for start, weight := range segmentTokens {
|
||||
if weight > maximum {
|
||||
end = start + 1
|
||||
ends[start] = end
|
||||
continue
|
||||
}
|
||||
for end < len(segmentTokens) && segmentTokens[end] <= maximum-windowTokens {
|
||||
windowTokens += segmentTokens[end]
|
||||
end++
|
||||
}
|
||||
ends[start] = end
|
||||
windowTokens -= weight
|
||||
}
|
||||
return ends
|
||||
}
|
||||
|
||||
// buildSectionsWithExplicitTarget builds exactly target nonempty sections when feasible.
|
||||
func (c *Chunker) buildSectionsWithExplicitTarget(
|
||||
segments []schema.Segment,
|
||||
segmentTokens []int,
|
||||
desiredSections int,
|
||||
targetTokensPerSection int,
|
||||
target, targetTokensPerSection int,
|
||||
) ([]Section, error) {
|
||||
n := len(segments)
|
||||
cursor := 0
|
||||
sections := make([]Section, 0, desiredSections)
|
||||
|
||||
for sectionIdx := 0; sectionIdx < desiredSections; sectionIdx++ {
|
||||
if cursor >= n {
|
||||
break
|
||||
}
|
||||
|
||||
remainingSectionsAfter := desiredSections - sectionIdx - 1
|
||||
currentSegments := []schema.Segment{segments[cursor]}
|
||||
currentTokens := segmentTokens[cursor]
|
||||
cursor++
|
||||
|
||||
for cursor < n {
|
||||
remainingSegments := n - cursor
|
||||
|
||||
// Reserve one segment per future section to avoid empty sections.
|
||||
if remainingSegments == remainingSectionsAfter {
|
||||
break
|
||||
}
|
||||
|
||||
nextTokens := segmentTokens[cursor]
|
||||
if currentTokens+nextTokens > c.config.MaxSectionTokens {
|
||||
break
|
||||
}
|
||||
if targetTokensPerSection == 0 || currentTokens < targetTokensPerSection {
|
||||
currentSegments = append(currentSegments, segments[cursor])
|
||||
currentTokens += nextTokens
|
||||
cursor++
|
||||
continue
|
||||
}
|
||||
break
|
||||
}
|
||||
|
||||
sections = append(sections, c.buildSection(len(sections), currentSegments, currentTokens))
|
||||
if target > n {
|
||||
return nil, &InfeasibleTargetError{Target: target, SegmentCount: n}
|
||||
}
|
||||
farthest := farthestLegalEnds(segmentTokens, c.config.MaxSectionTokens)
|
||||
minimum := make([]int, n+1)
|
||||
for start := n - 1; start >= 0; start-- {
|
||||
minimum[start] = 1 + minimum[farthest[start]]
|
||||
}
|
||||
if target < minimum[0] {
|
||||
return nil, &InfeasibleTargetError{Target: target, SegmentCount: n, MinimumSections: minimum[0], MaxSectionTokens: c.config.MaxSectionTokens}
|
||||
}
|
||||
|
||||
if cursor != n || len(sections) != desiredSections {
|
||||
return nil, fmt.Errorf(
|
||||
"target_sections (%d) is impossible under current constraints (got %d sections)",
|
||||
desiredSections,
|
||||
len(sections),
|
||||
)
|
||||
sections := make([]Section, 0, target)
|
||||
start := 0
|
||||
for i := range target {
|
||||
remaining := target - i - 1
|
||||
limit := min(farthest[start], n-remaining)
|
||||
total := 0
|
||||
for end := start + 1; end <= limit; end++ {
|
||||
total += segmentTokens[end-1]
|
||||
// Aim for the average where it leaves a feasible suffix. At
|
||||
// the limit the suffix fits by construction.
|
||||
if end == limit || (targetTokensPerSection > 0 && total >= targetTokensPerSection && minimum[end] <= remaining) {
|
||||
sections = append(sections, c.buildSection(i, slices.Clone(segments[start:end]), total))
|
||||
start = end
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return sections, nil
|
||||
}
|
||||
|
||||
@@ -308,5 +282,5 @@ func ceilDiv(numerator int, denominator int) int {
|
||||
if numerator <= 0 {
|
||||
return 0
|
||||
}
|
||||
return (numerator + denominator - 1) / denominator
|
||||
return 1 + (numerator-1)/denominator
|
||||
}
|
||||
|
||||
@@ -1,6 +1,9 @@
|
||||
package chunking
|
||||
|
||||
import (
|
||||
"errors"
|
||||
"math/bits"
|
||||
"strconv"
|
||||
"strings"
|
||||
"testing"
|
||||
|
||||
@@ -301,8 +304,9 @@ func TestChunkTargetSectionsImpossibleTooMany(t *testing.T) {
|
||||
chunker := NewChunker(ChunkingConfig{MaxSectionTokens: 100, MinSectionTokens: 1, TargetSections: &target})
|
||||
|
||||
_, err := chunker.ChunkTranscript(makeTranscript(segments))
|
||||
if err == nil {
|
||||
t.Fatal("expected error for impossible target_sections")
|
||||
var infeasible *InfeasibleTargetError
|
||||
if !errors.As(err, &infeasible) || infeasible.SegmentCount != len(segments) {
|
||||
t.Fatalf("expected typed segment-count error, got %v", err)
|
||||
}
|
||||
if !strings.Contains(err.Error(), "cannot have more sections than segments") {
|
||||
t.Fatalf("unexpected error: %v", err)
|
||||
@@ -318,14 +322,94 @@ func TestChunkTargetSectionsImpossibleTooFew(t *testing.T) {
|
||||
)
|
||||
|
||||
_, err := chunker.ChunkTranscript(makeTranscript(segments))
|
||||
if err == nil {
|
||||
t.Fatal("expected error for impossible target_sections")
|
||||
var infeasible *InfeasibleTargetError
|
||||
if !errors.As(err, &infeasible) || infeasible.MinimumSections != 3 {
|
||||
t.Fatalf("expected typed minimum-section error, got %v", err)
|
||||
}
|
||||
if !strings.Contains(err.Error(), "need at least 3 sections") {
|
||||
t.Fatalf("unexpected error: %v", err)
|
||||
}
|
||||
}
|
||||
|
||||
func TestChunkExplicitTargetKeepsFeasibleSuffix(t *testing.T) {
|
||||
segments := makeSegments([]string{"first", "second", "third"})
|
||||
chunker := NewChunkerWithEstimator(
|
||||
ChunkingConfig{MaxSectionTokens: 6, MinSectionTokens: 4, TargetSections: intPtr(2)},
|
||||
&mapTokenEstimator{byText: map[string]int{"first": 2, "second": 5, "third": 1}},
|
||||
)
|
||||
sections, err := chunker.ChunkTranscript(makeTranscript(segments))
|
||||
if err != nil || len(sections) != 2 || sections[0].EstimatedTokens != 2 || sections[1].EstimatedTokens != 6 {
|
||||
t.Fatalf("wanted weights [2, 6], got %+v (%v)", sections, err)
|
||||
}
|
||||
assertSegmentCoverageAndOrder(t, segments, sections)
|
||||
assertSectionMetadataConsistent(t, sections)
|
||||
}
|
||||
|
||||
func TestChunkExplicitTargetsMatchEnumeratedPartitions(t *testing.T) {
|
||||
// Enumerating cut masks gives an independent oracle for feasible counts,
|
||||
// including zero-weight segments around oversized singletons.
|
||||
const size, maximum = 5, 2
|
||||
texts := make([]string, size)
|
||||
for i := range texts {
|
||||
texts[i] = strconv.Itoa(i)
|
||||
}
|
||||
segments := makeSegments(texts)
|
||||
for sample := range 1 << (2 * size) {
|
||||
var weights [size]int
|
||||
byText := make(map[string]int, size)
|
||||
for i, text := range texts {
|
||||
weights[i] = (sample >> (2 * i)) & 3
|
||||
byText[text] = weights[i]
|
||||
}
|
||||
var feasible [size + 1]bool
|
||||
for cuts := range 1 << (size - 1) {
|
||||
start, total := 0, 0
|
||||
valid := true
|
||||
for i, weight := range weights {
|
||||
total += weight
|
||||
if i == size-1 || cuts&(1<<i) != 0 {
|
||||
if i > start && total > maximum {
|
||||
valid = false
|
||||
break
|
||||
}
|
||||
start, total = i+1, 0
|
||||
}
|
||||
}
|
||||
if valid {
|
||||
feasible[1+bits.OnesCount(uint(cuts))] = true
|
||||
}
|
||||
}
|
||||
for target := 1; target <= size; target++ {
|
||||
chunker := NewChunkerWithEstimator(
|
||||
ChunkingConfig{MaxSectionTokens: maximum, TargetSections: &target},
|
||||
&mapTokenEstimator{byText: byText},
|
||||
)
|
||||
sections, err := chunker.ChunkTranscript(makeTranscript(segments))
|
||||
if !feasible[target] {
|
||||
var infeasible *InfeasibleTargetError
|
||||
if !errors.As(err, &infeasible) {
|
||||
t.Fatalf("weights %v target %d should be infeasible, got %+v (%v)", weights, target, sections, err)
|
||||
}
|
||||
continue
|
||||
}
|
||||
if err != nil || len(sections) != target {
|
||||
t.Fatalf("weights %v target %d should be feasible, got %+v (%v)", weights, target, sections, err)
|
||||
}
|
||||
assertSegmentCoverageAndOrder(t, segments, sections)
|
||||
assertSectionMetadataConsistent(t, sections)
|
||||
for _, section := range sections {
|
||||
total := 0
|
||||
for _, segment := range section.Segments {
|
||||
total += weights[segment.ID-1]
|
||||
}
|
||||
if total != section.EstimatedTokens || (total > maximum && len(section.Segments) != 1) {
|
||||
t.Fatalf("weights %v target %d yielded invalid section %+v", weights, target, section)
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func TestChunkManySmallSegmentsBalanced(t *testing.T) {
|
||||
texts := []string{"1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11"}
|
||||
segments := makeSegments(texts)
|
||||
|
||||
Reference in New Issue
Block a user