diff --git a/changelog.d/maintenance/pending-video-bridge-fu07-structural-sampling.md b/changelog.d/maintenance/pending-video-bridge-fu07-structural-sampling.md new file mode 100644 index 0000000000..d1a9b761e8 --- /dev/null +++ b/changelog.d/maintenance/pending-video-bridge-fu07-structural-sampling.md @@ -0,0 +1 @@ +- **fix(video-bridge):** make opt-in segment-aware sampling use one bounded structural FFmpeg pass (scene, freeze, blur, exposure, and SI/TI), preserve long trailing segments, fail open to uniform sampling, and add real-media structural-oracle, overhead, post-dedup caption-call, and false-positive evidence while holding unconfigured model quality and gain-versus-cost claims. diff --git a/docs/security/GUARDRAILS.md b/docs/security/GUARDRAILS.md index f20cb80527..8eb4c8aa94 100644 --- a/docs/security/GUARDRAILS.md +++ b/docs/security/GUARDRAILS.md @@ -323,18 +323,56 @@ fallback. Videos are limited to 600 seconds, 8,192 pixels per dimension, and the long edge to at most 1,024 pixels without upscaling smaller inputs, and never receives a URL. Sampling is `uniform` by default. The optional `scene_aware` and experimental `segment_aware` policies perform one additional -fixed FFmpeg pass over the already validated local stream, select bounded -`showinfo` scene timestamps, and fall back deterministically to the same -uniform midpoints on detector failure, timeout, malformed output, or an empty -candidate set. Segment-aware mode allocates midpoint samples proportionally to -the validated scene intervals. The hard 16-frame cap is -applied after selection in every policy. A caller may optionally provide a +fixed FFmpeg pass over the already validated local stream. Scene-aware mode +selects bounded `showinfo` timestamps and falls back deterministically to the +same uniform midpoints on detector failure, timeout, malformed output, or an +empty candidate set. Segment-aware evidence and fallback behavior are detailed +below. The hard 16-frame cap is applied after selection in every policy. A caller may +optionally provide a finite focus window (`start`/`end` seconds); bounds are clamped to the media duration, reversed or non-finite windows are rejected, and all sampling policies are performed only inside the normalized interval. The resulting window is included in sampling metadata and in the untrusted description prefix so downstream models can distinguish a focused excerpt from the full timeline. + +#### FU-07 structural segment evidence + +`segment_aware` uses one bounded pre-analysis pass over the already validated +local video stream. The fixed filter chain first scales to at most 320 pixels +wide, detects scene changes and frozen intervals, then samples at 1 frame per +second for blur, average luma, and spatial/temporal information. The pass is +limited to 600 structural samples, one FFmpeg/filter thread, the same +`file`-only protocol and container allowlists, a 1 MiB process-output bound, +and at most 30 seconds inside the broker's shared abort/deadline. It never +accepts a command, filter, path, or URL from the request. + +The structural values are deterministic sampling evidence, not semantic video +understanding. They do not infer subjects, actions, captions, speech, or user +intent. Scene and freeze boundaries form segments; freeze coverage, blur, +exposure, spatial detail, and temporal change only influence how the existing +1–16 frame budget is allocated. A fully frozen segment is capped at one frame, +while non-frozen segments compete for the remaining budget. When boundaries +outnumber frames, uniform timeline coverage is retained so rapid early cuts +cannot hide a long trailing segment. Scene boundaries within the 1-second +analysis resolution of a freeze boundary are coalesced. + +Missing filters, malformed/empty evidence, a detector error, or the bounded +pre-analysis timeout fail open to the exact uniform midpoint policy. A caller +abort or broker deadline does not fail open: it terminates the in-flight +subprocess, prevents later frame extraction, and the private temporary tree is +removed in `finally`. + +`scripts/perf/video-bridge-fu07-eval.ts` generates deterministic real FFmpeg +fixtures for post-dedup caption-call savings, dense-motion budget allocation, +blur/exposure/SI-TI evidence, rapid cuts with a long tail, and gradual-fade +false positives. It records pre-analysis wall time and, where `/usr/bin/time` +is available, child CPU and peak RSS. Its quality checks are structural oracles +only. Real caption-model quality remains `HOLD` because this harness has no +authorized endpoint or frozen judge. Monetary savings also remain `HOLD` +unless `--caption-cost-per-call-usd` supplies an explicit positive per-call +estimate; the script never fabricates either result. + Each frame is limited to 4 MiB, all raw frames together to 23 MiB, and the serialized broker response to 32 MiB. A private temporary directory is removed in `finally`. OmniRoute does not bundle FFmpeg and does not accept a custom diff --git a/scripts/perf/video-bridge-fu07-eval.ts b/scripts/perf/video-bridge-fu07-eval.ts new file mode 100644 index 0000000000..33cd67665a --- /dev/null +++ b/scripts/perf/video-bridge-fu07-eval.ts @@ -0,0 +1,493 @@ +/** + * Real-media FU-07 structural-sampling evaluation. + * + * Run: node --import tsx/esm scripts/perf/video-bridge-fu07-eval.ts + * Optional estimate: append --caption-cost-per-call-usd . + * + * This evaluates deterministic structural oracles, not semantic model quality. + * Model quality and monetary savings remain HOLD without an external receipt. + */ +import { execFile } from "node:child_process"; +import { access, mkdir, mkdtemp, rm } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import { performance } from "node:perf_hooks"; +import { promisify } from "node:util"; + +import { deduplicateVideoFrames } from "../../src/lib/guardrails/videoBridgeHelpers"; +import { + analyzeVideoStructure, + calculateSamplingDecision, + extractFramesFromLocalVideo, + readBoundedExtractedFrames, + type VideoCommandRunner, + type VideoStructuralAnalysis, + type VideoStructuralSample, +} from "../../src/lib/guardrails/videoBridgeRuntime"; + +const execFileAsync = promisify(execFile); +const REQUIRED_FILTERS = ["scdet", "freezedetect", "blurdetect", "signalstats", "siti"]; +const TIME_MARKER = "__FU07_TIME__"; + +interface ChildCost { + maxRssKiB: number | null; + systemSeconds: number | null; + userSeconds: number | null; + wallMs: number; +} + +interface FixtureResult { + captionCallsAvoided: number; + childCost: ChildCost; + freezeIntervals: number; + name: string; + oracle: Record; + passed: boolean; + sceneCandidates: number; + structuralFrames: number; + uniformFrames: number; +} + +function average(values: Array): number | null { + const finite = values.filter( + (value): value is number => value !== null && value !== undefined && Number.isFinite(value) + ); + return finite.length > 0 ? finite.reduce((sum, value) => sum + value, 0) / finite.length : null; +} + +function samplesIn( + analysis: VideoStructuralAnalysis, + startSeconds: number, + endSeconds: number +): VideoStructuralSample[] { + return analysis.samples.filter( + (sample) => sample.timestampSeconds >= startSeconds && sample.timestampSeconds < endSeconds + ); +} + +async function generateFixture(outputPath: string, args: readonly string[]): Promise { + await execFileAsync( + "ffmpeg", + ["-hide_banner", "-loglevel", "error", ...args, "-threads", "1", "-y", outputPath], + { maxBuffer: 1024 * 1024, timeout: 30_000 } + ); +} + +async function generateStaticFixture(outputPath: string): Promise { + await generateFixture(outputPath, [ + "-f", + "lavfi", + "-i", + "color=c=blue:s=320x180:d=8:r=12", + "-c:v", + "libx264", + "-preset", + "ultrafast", + "-pix_fmt", + "yuv420p", + ]); +} + +async function generateMixedFixture(outputPath: string): Promise { + await generateFixture(outputPath, [ + "-f", + "lavfi", + "-i", + "color=c=black:s=320x180:d=6:r=12", + "-f", + "lavfi", + "-i", + "testsrc2=s=320x180:d=4:r=12", + "-filter_complex", + "[0:v][1:v]concat=n=2:v=1:a=0,format=yuv420p[v]", + "-map", + "[v]", + "-c:v", + "libx264", + "-preset", + "ultrafast", + ]); +} + +async function generateBlurExposureFixture(outputPath: string): Promise { + await generateFixture(outputPath, [ + "-f", + "lavfi", + "-i", + "testsrc2=s=320x180:d=3:r=12", + "-f", + "lavfi", + "-i", + "color=c=black:s=320x180:d=3:r=12", + "-f", + "lavfi", + "-i", + "testsrc2=s=320x180:d=4:r=12", + "-filter_complex", + "[0:v]gblur=sigma=12[blur];[blur][1:v][2:v]concat=n=3:v=1:a=0,format=yuv420p[v]", + "-map", + "[v]", + "-c:v", + "libx264", + "-preset", + "ultrafast", + ]); +} + +async function generateDenseTailFixture(outputPath: string): Promise { + const args: string[] = []; + for (const source of [ + "color=c=black:s=160x90:d=0.5:r=10", + "color=c=white:s=160x90:d=0.5:r=10", + "color=c=black:s=160x90:d=0.5:r=10", + "color=c=white:s=160x90:d=0.5:r=10", + "testsrc2=s=160x90:d=8:r=10", + ]) { + args.push("-f", "lavfi", "-i", source); + } + args.push( + "-filter_complex", + "[0:v][1:v][2:v][3:v][4:v]concat=n=5:v=1:a=0,format=yuv420p[v]", + "-map", + "[v]", + "-c:v", + "libx264", + "-preset", + "ultrafast" + ); + await generateFixture(outputPath, args); +} + +async function generateGradualFadeFixture(outputPath: string): Promise { + await generateFixture(outputPath, [ + "-f", + "lavfi", + "-i", + "color=c=white:s=320x180:d=8:r=12", + "-vf", + "fade=t=out:st=0:d=8,format=yuv420p", + "-c:v", + "libx264", + "-preset", + "ultrafast", + ]); +} + +async function supportsTimeBinary(): Promise { + try { + await access("/usr/bin/time"); + return true; + } catch { + return false; + } +} + +function parseTimeCost(stderr: string, wallMs: number): ChildCost { + const match = new RegExp(`${TIME_MARKER} ([\\d.]+) ([\\d.]+) ([\\d.]+)`).exec(stderr); + return { + maxRssKiB: match ? Number(match[3]) : null, + systemSeconds: match ? Number(match[2]) : null, + userSeconds: match ? Number(match[1]) : null, + wallMs, + }; +} + +async function timedAnalysis( + inputPath: string, + durationSeconds: number, + useTimeBinary: boolean +): Promise<{ analysis: VideoStructuralAnalysis; cost: ChildCost }> { + let cost: ChildCost = { + maxRssKiB: null, + systemSeconds: null, + userSeconds: null, + wallMs: 0, + }; + const runner: VideoCommandRunner = async (executable, args, options) => { + const startedAt = performance.now(); + const command = useTimeBinary ? "/usr/bin/time" : executable; + const commandArgs = useTimeBinary + ? ["-f", `${TIME_MARKER} %U %S %M`, executable, ...args] + : [...args]; + const result = await execFileAsync(command, commandArgs, { + encoding: "utf8", + maxBuffer: 1024 * 1024, + signal: options.signal, + timeout: options.timeoutMs, + }); + cost = parseTimeCost(String(result.stderr), performance.now() - startedAt); + return { stderr: String(result.stderr), stdout: String(result.stdout) }; + }; + const analysis = await analyzeVideoStructure(inputPath, { + durationSeconds, + runner, + streamIndex: 0, + timeoutMs: 30_000, + }); + return { analysis, cost }; +} + +function sampling( + durationSeconds: number, + frameCount: number, + analysis: VideoStructuralAnalysis +): { structural: number[]; uniform: number[] } { + const uniform = calculateSamplingDecision(durationSeconds, frameCount, "uniform").timestamps; + const structural = calculateSamplingDecision( + durationSeconds, + frameCount, + "segment_aware", + analysis.sceneCandidates, + null, + analysis + ).timestamps; + return { structural, uniform }; +} + +async function captionCallsAfterDedup( + inputPath: string, + outputDirectory: string, + samplingPolicy: "segment_aware" | "uniform" +): Promise { + await mkdir(outputDirectory, { mode: 0o700 }); + const frames = await extractFramesFromLocalVideo(inputPath, outputDirectory, { + durationSeconds: 8, + frameCount: 8, + samplingPolicy, + streamIndex: 0, + timeoutMs: 30_000, + }); + const bytes = await readBoundedExtractedFrames(frames); + const deduplicated = await deduplicateVideoFrames( + frames.map((frame, index) => ({ + dataUri: `data:image/jpeg;base64,${bytes[index].toString("base64")}`, + timestampSeconds: frame.timestampSeconds, + })) + ); + return deduplicated.frames.length; +} + +function result( + name: string, + cost: ChildCost, + analysis: VideoStructuralAnalysis, + uniform: number[], + structural: number[], + oracle: Record, + captionCallsAvoided = 0 +): FixtureResult { + const booleans = Object.values(oracle).filter( + (value): value is boolean => typeof value === "boolean" + ); + return { + captionCallsAvoided, + childCost: cost, + freezeIntervals: analysis.freezeIntervals.length, + name, + oracle, + passed: booleans.every(Boolean), + sceneCandidates: analysis.sceneCandidates.length, + structuralFrames: structural.length, + uniformFrames: uniform.length, + }; +} + +async function main(): Promise { + const version = await execFileAsync("ffmpeg", ["-version"], { timeout: 5_000 }); + const filters = await execFileAsync("ffmpeg", ["-hide_banner", "-filters"], { + maxBuffer: 2 * 1024 * 1024, + timeout: 5_000, + }); + const missingFilters = REQUIRED_FILTERS.filter( + (filter) => !new RegExp(`\\b${filter}\\b`).test(String(filters.stdout)) + ); + if (missingFilters.length > 0) + throw new Error(`Missing required FFmpeg filters: ${missingFilters.join(", ")}`); + + const directory = await mkdtemp(join(tmpdir(), "video-fu07-eval-")); + const useTimeBinary = await supportsTimeBinary(); + const results: FixtureResult[] = []; + try { + const staticPath = join(directory, "static.mp4"); + await generateStaticFixture(staticPath); + const staticRun = await timedAnalysis(staticPath, 8, useTimeBinary); + const staticSampling = sampling(8, 8, staticRun.analysis); + const uniformCaptionCalls = await captionCallsAfterDedup( + staticPath, + join(directory, "static-uniform"), + "uniform" + ); + const structuralCaptionCalls = await captionCallsAfterDedup( + staticPath, + join(directory, "static-structural"), + "segment_aware" + ); + const staticCaptionCallsAvoided = Math.max(0, uniformCaptionCalls - structuralCaptionCalls); + results.push( + result( + "static-caption-savings", + staticRun.cost, + staticRun.analysis, + staticSampling.uniform, + staticSampling.structural, + { + fullFreezeDetected: staticRun.analysis.freezeIntervals.some( + (interval) => interval.startSeconds <= 1 && interval.endSeconds >= 7 + ), + oneIncrementalCaptionCallAvoided: staticCaptionCallsAvoided === 1, + structuralCaptionCalls, + uniformCaptionCalls, + }, + staticCaptionCallsAvoided + ) + ); + + const mixedPath = join(directory, "mixed.mp4"); + await generateMixedFixture(mixedPath); + const mixedRun = await timedAnalysis(mixedPath, 10, useTimeBinary); + const mixedSampling = sampling(10, 4, mixedRun.analysis); + const uniformDense = mixedSampling.uniform.filter((timestamp) => timestamp > 6).length; + const structuralDense = mixedSampling.structural.filter((timestamp) => timestamp > 6).length; + results.push( + result( + "dense-budget-quality-oracle", + mixedRun.cost, + mixedRun.analysis, + mixedSampling.uniform, + mixedSampling.structural, + { + denseFramesStructural: structuralDense, + denseFramesUniform: uniformDense, + denseRegionGetsMoreBudget: structuralDense > uniformDense, + frozenRegionRetainsCoverage: mixedSampling.structural.some((timestamp) => timestamp < 6), + } + ) + ); + + const qualityPath = join(directory, "blur-exposure.mp4"); + await generateBlurExposureFixture(qualityPath); + const qualityRun = await timedAnalysis(qualityPath, 10, useTimeBinary); + const qualitySampling = sampling(10, 6, qualityRun.analysis); + const blurred = samplesIn(qualityRun.analysis, 0, 3); + const dark = samplesIn(qualityRun.analysis, 3, 6); + const sharp = samplesIn(qualityRun.analysis, 6, 10); + const blurredBlur = average(blurred.map((sample) => sample.blur)); + const blurredSpatial = average(blurred.map((sample) => sample.spatialInformation)); + const darkLuma = average(dark.map((sample) => sample.brightness)); + const sharpBlur = average(sharp.map((sample) => sample.blur)); + const sharpSpatial = average(sharp.map((sample) => sample.spatialInformation)); + const sharpTemporal = average(sharp.map((sample) => sample.temporalInformation)); + const sharpLuma = average(sharp.map((sample) => sample.brightness)); + results.push( + result( + "blur-exposure-spatial-temporal-evidence", + qualityRun.cost, + qualityRun.analysis, + qualitySampling.uniform, + qualitySampling.structural, + { + blurMetricSeparated: + blurredBlur !== null && sharpBlur !== null && Math.abs(blurredBlur - sharpBlur) >= 0.05, + blurredBlur: blurredBlur ?? "missing", + darkLuma: darkLuma ?? "missing", + exposureSeparated: darkLuma !== null && sharpLuma !== null && sharpLuma - darkLuma >= 50, + sharpBlur: sharpBlur ?? "missing", + sharpSpatial: sharpSpatial ?? "missing", + sharpTemporal: sharpTemporal ?? "missing", + spatialDetailSeparated: + blurredSpatial !== null && sharpSpatial !== null && sharpSpatial - blurredSpatial >= 20, + structuralKeepsSharpRegion: + qualitySampling.structural.filter((timestamp) => timestamp >= 6).length >= 2, + temporalChangeDetected: sharpTemporal !== null && sharpTemporal >= 5, + } + ) + ); + + const tailPath = join(directory, "dense-tail.mp4"); + await generateDenseTailFixture(tailPath); + const tailRun = await timedAnalysis(tailPath, 10, useTimeBinary); + const tailSampling = sampling(10, 4, tailRun.analysis); + results.push( + result( + "dense-cuts-long-tail-regression", + tailRun.cost, + tailRun.analysis, + tailSampling.uniform, + tailSampling.structural, + { + multipleEarlyCuts: tailRun.analysis.sceneCandidates.length >= 3, + trailingEightSecondsRepresented: tailSampling.structural.some( + (timestamp) => timestamp > 2 + ), + } + ) + ); + + const fadePath = join(directory, "gradual-fade.mp4"); + await generateGradualFadeFixture(fadePath); + const fadeRun = await timedAnalysis(fadePath, 8, useTimeBinary); + const fadeSampling = sampling(8, 4, fadeRun.analysis); + results.push( + result( + "gradual-fade-false-positive", + fadeRun.cost, + fadeRun.analysis, + fadeSampling.uniform, + fadeSampling.structural, + { + hardCutFalsePositives: fadeRun.analysis.sceneCandidates.length, + noHardCutBurst: fadeRun.analysis.sceneCandidates.length <= 1, + noCaptionBudgetPruning: fadeSampling.structural.length === fadeSampling.uniform.length, + } + ) + ); + } finally { + await rm(directory, { force: true, recursive: true }); + } + + const callsAvoided = results.reduce((sum, fixture) => sum + fixture.captionCallsAvoided, 0); + const costFlag = process.argv.indexOf("--caption-cost-per-call-usd"); + const explicitCost = Number(costFlag >= 0 ? process.argv[costFlag + 1] : Number.NaN); + const report = { + captionCost: + Number.isFinite(explicitCost) && explicitCost > 0 + ? { + estimatedUsdAvoided: callsAvoided * explicitCost, + source: "explicit environment input", + status: "ESTIMATED_FROM_INPUT", + } + : { + reason: "--caption-cost-per-call-usd was not supplied with a positive number", + status: "HOLD", + }, + ffmpegVersion: String(version.stdout).split("\n")[0], + fixtures: results, + modelQuality: { + reason: + "No authorized real caption-model endpoint, credentials, or frozen judge rubric were configured; deterministic structural oracles are not semantic quality.", + status: "HOLD", + }, + gainCostComparison: { + reason: + "The real post-dedup caption-call delta is measured, but no authorized caption latency/cost receipt or child CPU/RSS receipt is configured.", + status: "HOLD", + }, + resourceCost: useTimeBinary + ? { source: "/usr/bin/time", status: "MEASURED" } + : { + reason: "/usr/bin/time is unavailable; wall time is measured but child CPU/RSS are not", + status: "HOLD", + }, + summary: { + captionCallsAvoided: callsAvoided, + failed: results.filter((fixture) => !fixture.passed).map((fixture) => fixture.name), + passed: results.filter((fixture) => fixture.passed).length, + total: results.length, + }, + timeBinary: useTimeBinary ? "/usr/bin/time" : null, + }; + console.log(JSON.stringify(report, null, 2)); + if (report.summary.failed.length > 0) process.exitCode = 1; +} + +await main(); diff --git a/src/lib/guardrails/videoBridgeRuntime.ts b/src/lib/guardrails/videoBridgeRuntime.ts index fea9769381..c64eb22b0d 100644 --- a/src/lib/guardrails/videoBridgeRuntime.ts +++ b/src/lib/guardrails/videoBridgeRuntime.ts @@ -69,6 +69,24 @@ export interface VideoSamplingDecision extends VideoSamplingMetadata { timestamps: number[]; } +export interface VideoStructuralInterval { + endSeconds: number; + startSeconds: number; +} +export interface VideoStructuralSample { + blur?: number | null; + brightness?: number | null; + sceneScore?: number | null; + spatialInformation?: number | null; + temporalInformation?: number | null; + timestampSeconds: number; +} +export interface VideoStructuralAnalysis { + freezeIntervals: VideoStructuralInterval[]; + samples: VideoStructuralSample[]; + sceneCandidates: number[]; +} + export function resolveVideoFocusWindow( durationSeconds: number, bounds: VideoFocusBounds @@ -95,6 +113,10 @@ export const VIDEO_FRAME_MAX_BYTES = 4 * 1024 * 1024; export const VIDEO_FRAMES_TOTAL_MAX_BYTES = 23 * 1024 * 1024; export const VIDEO_MAX_DIMENSION = 8_192; export const VIDEO_MAX_PIXELS = 33_554_432; +const VIDEO_STRUCTURAL_ANALYSIS_FPS = 1; +const VIDEO_STRUCTURAL_ANALYSIS_MAX_SAMPLES = 600; +const VIDEO_STRUCTURAL_ANALYSIS_MAX_WIDTH = 320; +const VIDEO_STRUCTURAL_SCENE_THRESHOLD = 10; const SAFE_FORMATS = new Set([ "3g2", @@ -111,7 +133,6 @@ const SAFE_FORMATS = new Set([ "webm", ]); const SAFE_FORMAT_WHITELIST = [...SAFE_FORMATS].join(","); - const defaultRunner: VideoCommandRunner = async (executable, args, options) => { const result = await execFileAsync(executable, [...args], { encoding: "utf8", @@ -122,7 +143,6 @@ const defaultRunner: VideoCommandRunner = async (executable, args, options) => { }); return { stdout: String(result.stdout), stderr: String(result.stderr) }; }; - function assertLocalPath(filePath: string): void { if (!isAbsolute(filePath) || filePath.includes("\0") || filePath.includes("://")) { throw new Error("Video runtime requires a local path"); @@ -223,7 +243,6 @@ function normalizeSceneCandidates( } return [...unique].sort((left, right) => left - right); } - export function parseSceneChangeTimestamps(output: string, durationSeconds: number): number[] { const candidates: number[] = []; const timestampPattern = /\bpts_time:([+-]?(?:\d+(?:\.\d*)?|\.\d+))\b/g; @@ -233,67 +252,256 @@ export function parseSceneChangeTimestamps(output: string, durationSeconds: numb } return normalizeSceneCandidates(durationSeconds, candidates); } - -/** Allocate midpoint samples proportionally across validated scene segments. */ +const STRUCTURAL_METRIC_FIELDS = { + "lavfi.blur": "blur", + "lavfi.scd.score": "sceneScore", + "lavfi.signalstats.YAVG": "brightness", + "lavfi.siti.si": "spatialInformation", + "lavfi.siti.ti": "temporalInformation", +} as const; +function parseStructuralSamples(output: string, durationSeconds: number): VideoStructuralSample[] { + const samples = new Map(); + const pattern = /\bpts_time:([+-]?(?:\d+(?:\.\d*)?|\.\d+))[^\n]*\r?\n([A-Za-z0-9_.]+)=([^\s]+)/g; + for (const match of output.matchAll(pattern)) { + const timestamp = Number(Number(match[1]).toFixed(3)); + const field = STRUCTURAL_METRIC_FIELDS[match[2] as keyof typeof STRUCTURAL_METRIC_FIELDS]; + const metric = Number(match[3]); + const unusable = !field || timestamp < 0 || timestamp >= durationSeconds; + if (unusable || (!Number.isFinite(metric) && !samples.has(timestamp))) continue; + if (!samples.has(timestamp)) { + if (samples.size >= VIDEO_STRUCTURAL_ANALYSIS_MAX_SAMPLES) continue; + samples.set(timestamp, { + timestampSeconds: timestamp, + }); + } + const sample = samples.get(timestamp); + if (sample) sample[field] = Number.isFinite(metric) ? metric : null; + } + return [...samples.values()].sort( + (left, right) => left.timestampSeconds - right.timestampSeconds + ); +} +function parseStructuralMetricEvents(output: string, metric: string): number[] { + const pattern = new RegExp(`${metric}:\\s*([+-]?(?:\\d+(?:\\.\\d*)?|\\.\\d+))`, "g"); + return [...output.matchAll(pattern)].map((match) => Number(match[1])).filter(Number.isFinite); +} +function parseFreezeIntervals(output: string, durationSeconds: number): VideoStructuralInterval[] { + const starts = parseStructuralMetricEvents(output, "freeze_start"); + const ends = parseStructuralMetricEvents(output, "freeze_end"); + const durations = parseStructuralMetricEvents(output, "freeze_duration"); + return starts + .map((start, index) => { + const startSeconds = Math.max(0, Math.min(durationSeconds, start)); + const inferredEnd = start + (durations[index] ?? durationSeconds - start); + const endSeconds = Math.max( + startSeconds, + Math.min(durationSeconds, ends[index] ?? inferredEnd) + ); + return { endSeconds, startSeconds }; + }) + .filter((interval) => interval.endSeconds - interval.startSeconds >= 1); +} +export function parseVideoStructuralAnalysis( + metadataOutput: string, + diagnosticOutput: string, + durationSeconds: number +): VideoStructuralAnalysis { + if (!Number.isFinite(durationSeconds) || durationSeconds <= 0) { + throw new Error("Video structural analysis requires a positive duration"); + } + const samples = parseStructuralSamples(metadataOutput, durationSeconds); + const diagnosticScenes = [ + ...diagnosticOutput.matchAll(/lavfi\.scd\.score:\s*[\d.]+,\s*lavfi\.scd\.time:\s*([\d.]+)/g), + ].map((match) => Number(match[1])); + return { + freezeIntervals: parseFreezeIntervals(diagnosticOutput, durationSeconds), + samples, + sceneCandidates: normalizeSceneCandidates(durationSeconds, [ + ...samples + .filter((sample) => (sample.sceneScore ?? 0) >= VIDEO_STRUCTURAL_SCENE_THRESHOLD) + .map((sample) => sample.timestampSeconds), + ...diagnosticScenes, + ]), + }; +} +interface StructuralSamplingSegment { + endSeconds: number; + frozen: boolean; + priority: number; + startSeconds: number; +} +function averageStructuralMetric(values: Array): number | null { + const finite = values.filter( + (value): value is number => value !== null && value !== undefined && Number.isFinite(value) + ); + return finite.length > 0 ? finite.reduce((sum, value) => sum + value, 0) / finite.length : null; +} +function normalizedStructuralMetric( + samples: readonly VideoStructuralSample[], + field: Exclude, + fallback: number, + scale: number +): number { + return Math.min( + 1, + Math.max( + 0, + (averageStructuralMetric(samples.map((sample) => sample[field])) ?? fallback) / scale + ) + ); +} +function structuralSegmentPriority( + startSeconds: number, + endSeconds: number, + analysis: VideoStructuralAnalysis +): StructuralSamplingSegment { + const length = endSeconds - startSeconds; + const samples = analysis.samples.filter( + (sample) => sample.timestampSeconds >= startSeconds && sample.timestampSeconds < endSeconds + ); + const freezeCoverage = Math.min( + 1, + analysis.freezeIntervals.reduce( + (sum, interval) => + sum + + Math.max( + 0, + Math.min(endSeconds, interval.endSeconds) - Math.max(startSeconds, interval.startSeconds) + ), + 0 + ) / length + ); + const spatial = normalizedStructuralMetric(samples, "spatialInformation", 40, 100); + const temporal = normalizedStructuralMetric(samples, "temporalInformation", 10, 30); + const sharpness = 1 - normalizedStructuralMetric(samples, "blur", 10, 20); + const brightness = averageStructuralMetric(samples.map((sample) => sample.brightness)); + const exposure = brightness === null || (brightness >= 24 && brightness <= 232) ? 1 : 0.25; + const interest = exposure * (0.2 + spatial * 0.3 + temporal * 0.4 + sharpness * 0.1); + const maxTemporal = Math.max(0, ...samples.map((sample) => sample.temporalInformation ?? 0)); + return { + endSeconds, + frozen: freezeCoverage >= 0.8 && maxTemporal <= 1, + priority: length * Math.max(0.05, interest) * (1 - freezeCoverage * 0.75), + startSeconds, + }; +} +function allocateStructuralFrames( + segments: readonly StructuralSamplingSegment[], + frameCount: number +): number[] { + if (segments.length > frameCount) return segments.map(() => 0); + const allocation = segments.map(() => 1); + let remaining = frameCount - segments.length; + const totalPriority = segments.reduce( + (sum, segment) => sum + (segment.frozen ? 0 : segment.priority), + 0 + ); + if (totalPriority <= 0) return allocation; + const idealExtras = segments.map((segment) => + segment.frozen ? 0 : (segment.priority / totalPriority) * remaining + ); + const extras = idealExtras.map((value) => Math.floor(value)); + remaining -= extras.reduce((sum, value) => sum + value, 0); + const remainderOrder = idealExtras + .map((value, index) => ({ index, remainder: value - Math.floor(value) })) + .sort((left, right) => right.remainder - left.remainder || left.index - right.index); + for (let index = 0; index < remaining; index++) extras[remainderOrder[index].index] += 1; + return allocation.map((value, index) => value + extras[index]); +} +function timestampsFromSegmentAllocation( + segments: readonly Pick[], + allocation: readonly number[] +): number[] { + return segments.flatMap((segment, segmentIndex) => + Array.from( + { length: allocation[segmentIndex] }, + (_unused, index) => + segment.startSeconds + + ((index + 0.5) * (segment.endSeconds - segment.startSeconds)) / allocation[segmentIndex] + ) + ); +} +function calculateLengthWeightedSegmentTimestamps( + startSeconds: number, + endSeconds: number, + frameCount: number, + boundaries: readonly number[] +): number[] { + const uniform = calculateFrameTimestamps(endSeconds - startSeconds, frameCount).map( + (timestamp) => timestamp + startSeconds + ); + const starts = [startSeconds, ...boundaries]; + const ends = [...boundaries, endSeconds]; + const segments = starts.map((start, index) => ({ + endSeconds: ends[index], + frozen: false, + priority: ends[index] - start, + startSeconds: start, + })); + return segments.length > frameCount + ? uniform + : timestampsFromSegmentAllocation(segments, allocateStructuralFrames(segments, frameCount)); +} +/** Allocate a bounded caption budget across validated structural segments. */ export function calculateSegmentAwareTimestamps( durationSeconds: number, requestedFrameCount: number, sceneCandidates: readonly number[], - focusWindow: VideoFocusWindow | null = null + focusWindow: VideoFocusWindow | null = null, + structuralAnalysis: VideoStructuralAnalysis | null = null ): number[] { const startSeconds = focusWindow?.startSeconds ?? 0; const endSeconds = focusWindow?.endSeconds ?? durationSeconds; const uniform = calculateFrameTimestamps(endSeconds - startSeconds, requestedFrameCount).map( (timestamp) => timestamp + startSeconds ); - const boundaries = normalizeSceneCandidates(durationSeconds, sceneCandidates).filter( - (timestamp) => timestamp > startSeconds && timestamp < endSeconds + const structuralBoundaries = structuralAnalysis?.freezeIntervals.flatMap((interval) => [ + interval.startSeconds, + interval.endSeconds, + ]); + const sceneBoundaries = sceneCandidates.filter( + (candidate) => + !structuralBoundaries?.some( + (boundary) => Math.abs(candidate - boundary) <= 1 / VIDEO_STRUCTURAL_ANALYSIS_FPS + ) ); - if (boundaries.length === 0) return uniform; - const segmentStarts = [startSeconds, ...boundaries]; - const segmentEnds = [...boundaries, endSeconds]; - const lengths = segmentStarts.map((segmentStart, index) => segmentEnds[index] - segmentStart); - const segmentCount = lengths.length; - const frameCount = uniform.length; - if (segmentCount > frameCount) { - return [...uniform].map((timestamp, index) => { - const segmentIndex = Math.min( - segmentCount - 1, - Math.floor((index * segmentCount) / frameCount) - ); - const segmentStart = segmentStarts[segmentIndex]; - const segmentEnd = segmentEnds[segmentIndex]; - return segmentStart + (segmentEnd - segmentStart) / 2; - }); + const boundaries = normalizeSceneCandidates(durationSeconds, [ + ...sceneBoundaries, + ...(structuralBoundaries ?? []), + ]).filter((timestamp) => timestamp > startSeconds && timestamp < endSeconds); + if (!structuralAnalysis) { + return boundaries.length === 0 + ? uniform + : calculateLengthWeightedSegmentTimestamps( + startSeconds, + endSeconds, + uniform.length, + boundaries + ); } - const allocation = lengths.map(() => 1); - let remaining = frameCount - segmentCount; - const idealExtra = lengths.map((length) => (length / (endSeconds - startSeconds)) * remaining); - const extras = idealExtra.map((value) => Math.floor(value)); - remaining -= extras.reduce((sum, value) => sum + value, 0); - const remainderOrder = idealExtra - .map((value, index) => ({ index, remainder: value - Math.floor(value) })) - .sort((left, right) => right.remainder - left.remainder || left.index - right.index); - for (let index = 0; index < remaining; index++) extras[remainderOrder[index].index] += 1; - for (let index = 0; index < allocation.length; index++) allocation[index] += extras[index]; - const timestamps: number[] = []; - for (let segmentIndex = 0; segmentIndex < segmentCount; segmentIndex++) { - const count = allocation[segmentIndex]; - const segmentStart = segmentStarts[segmentIndex]; - const segmentLength = lengths[segmentIndex]; - for (let index = 0; index < count; index++) { - timestamps.push(segmentStart + ((index + 0.5) * segmentLength) / count); - } + const starts = [startSeconds, ...boundaries]; + const ends = [...boundaries, endSeconds]; + const segments = starts.map((start, index) => + structuralSegmentPriority(start, ends[index], structuralAnalysis) + ); + const allocation = allocateStructuralFrames(segments, uniform.length); + if (segments.length > uniform.length) { + return calculateLengthWeightedSegmentTimestamps( + startSeconds, + endSeconds, + uniform.length, + boundaries + ); } - return timestamps; + return timestampsFromSegmentAllocation(segments, allocation); } - export function calculateSamplingDecision( durationSeconds: number, requestedFrameCount: number, policy: VideoSamplingPolicy, sceneCandidates: readonly number[] = [], - focusWindow: VideoFocusWindow | null = null + focusWindow: VideoFocusWindow | null = null, + structuralAnalysis: VideoStructuralAnalysis | null = null ): VideoSamplingDecision { const startSeconds = focusWindow?.startSeconds ?? 0; const endSeconds = focusWindow?.endSeconds ?? durationSeconds; @@ -309,21 +517,17 @@ export function calculateSamplingDecision( timestamps: uniform, }; } - const candidates = normalizeSceneCandidates(durationSeconds, sceneCandidates).filter( - (timestamp) => timestamp >= startSeconds && timestamp < endSeconds + (timestamp) => timestamp > startSeconds && timestamp < endSeconds ); - if (candidates.length === 0) { - return { - candidateCount: 0, - ...(focusWindow ? { focusWindow } : {}), - policyEffective: "uniform", - policyRequested: policy, - timestamps: uniform, - }; - } - - if (policy === "segment_aware") { + const focusHasSample = structuralAnalysis?.samples.some( + (sample) => sample.timestampSeconds >= startSeconds && sample.timestampSeconds < endSeconds + ); + const focusHasFreeze = structuralAnalysis?.freezeIntervals.some( + (interval) => interval.startSeconds < endSeconds && interval.endSeconds > startSeconds + ); + const hasStructuralEvidence = Boolean(focusHasSample || focusHasFreeze); + if (policy === "segment_aware" && (candidates.length > 0 || hasStructuralEvidence)) { return { candidateCount: candidates.length, ...(focusWindow ? { focusWindow } : {}), @@ -333,11 +537,20 @@ export function calculateSamplingDecision( durationSeconds, requestedFrameCount, candidates, - focusWindow + focusWindow, + structuralAnalysis ), }; } - + if (candidates.length === 0) { + return { + candidateCount: 0, + ...(focusWindow ? { focusWindow } : {}), + policyEffective: "uniform", + policyRequested: policy, + timestamps: uniform, + }; + } const frameCount = uniform.length; const selected = candidates.length <= frameCount @@ -369,7 +582,6 @@ export function calculateSamplingDecision( timestamps: selected, }; } - export async function detectSceneChangeTimestamps( inputPath: string, options: { @@ -412,7 +624,72 @@ export async function detectSceneChangeTimestamps( ); return parseSceneChangeTimestamps(`${result.stdout}\n${result.stderr}`, options.durationSeconds); } - +const STRUCTURAL_ANALYSIS_FILTER = [ + `scale=w='min(${VIDEO_STRUCTURAL_ANALYSIS_MAX_WIDTH},iw)':h=-2:flags=fast_bilinear`, + `scdet=threshold=${VIDEO_STRUCTURAL_SCENE_THRESHOLD}`, + "freezedetect=n=-60dB:d=1", + `fps=${VIDEO_STRUCTURAL_ANALYSIS_FPS}`, + "siti", + "blurdetect=radius=10:block_width=32:block_height=32", + "signalstats", + ...[ + "lavfi.scd.score", + "lavfi.siti.si", + "lavfi.siti.ti", + "lavfi.blur", + "lavfi.signalstats.YAVG", + ].map((key) => `metadata=mode=print:key=${key}:file=-`), +].join(","); +export async function analyzeVideoStructure( + inputPath: string, + options: { + durationSeconds: number; + runner?: VideoCommandRunner; + signal?: AbortSignal; + streamIndex: number; + timeoutMs?: number; + } +): Promise { + assertLocalPath(inputPath); + if (!Number.isFinite(options.durationSeconds) || options.durationSeconds <= 0) { + throw new Error("Video structural analysis requires a positive duration"); + } + if (!Number.isInteger(options.streamIndex) || options.streamIndex < 0) { + throw new Error("Video stream index is invalid"); + } + const result = await (options.runner ?? defaultRunner)( + "ffmpeg", + [ + "-nostdin", + "-hide_banner", + "-loglevel", + "info", + "-nostats", + "-protocol_whitelist", + "file", + "-format_whitelist", + SAFE_FORMAT_WHITELIST, + "-threads", + "1", + "-filter_threads", + "1", + "-i", + inputPath, + "-map", + `0:${options.streamIndex}`, + "-vf", + STRUCTURAL_ANALYSIS_FILTER, + "-an", + "-frames:v", + String(VIDEO_STRUCTURAL_ANALYSIS_MAX_SAMPLES), + "-f", + "null", + "-", + ], + { signal: options.signal, timeoutMs: Math.min(options.timeoutMs ?? 30_000, 30_000) } + ); + return parseVideoStructuralAnalysis(result.stdout, result.stderr, options.durationSeconds); +} export async function probeLocalVideo( inputPath: string, options: { @@ -547,18 +824,31 @@ export async function extractFramesFromLocalVideo( assertLocalPath(outputDirectory); const policy = options.samplingPolicy ?? "uniform"; let sceneCandidates: number[] = []; + let structuralAnalysis: VideoStructuralAnalysis | null = null; if (policy !== "uniform") { try { - sceneCandidates = await detectSceneChangeTimestamps(inputPath, { - durationSeconds: options.durationSeconds, - runner: options.runner, - signal: options.signal, - streamIndex: options.streamIndex, - timeoutMs: Math.min(options.timeoutMs ?? 30_000, 30_000), - }); + if (policy === "segment_aware") { + structuralAnalysis = await analyzeVideoStructure(inputPath, { + durationSeconds: options.durationSeconds, + runner: options.runner, + signal: options.signal, + streamIndex: options.streamIndex, + timeoutMs: Math.min(options.timeoutMs ?? 30_000, 30_000), + }); + sceneCandidates = structuralAnalysis.sceneCandidates; + } else { + sceneCandidates = await detectSceneChangeTimestamps(inputPath, { + durationSeconds: options.durationSeconds, + runner: options.runner, + signal: options.signal, + streamIndex: options.streamIndex, + timeoutMs: Math.min(options.timeoutMs ?? 30_000, 30_000), + }); + } } catch { if (options.signal?.aborted) throw new Error("Video extraction request aborted"); sceneCandidates = []; + structuralAnalysis = null; } } const focusWindow = options.focusWindow @@ -569,7 +859,8 @@ export async function extractFramesFromLocalVideo( options.frameCount, policy, sceneCandidates, - focusWindow + focusWindow, + structuralAnalysis ); if (!Number.isInteger(options.streamIndex) || options.streamIndex < 0) { throw new Error("Video stream index is invalid"); diff --git a/tests/unit/guardrails/videoBridgeFu07StructuralSampling.test.ts b/tests/unit/guardrails/videoBridgeFu07StructuralSampling.test.ts new file mode 100644 index 0000000000..5c5695f018 --- /dev/null +++ b/tests/unit/guardrails/videoBridgeFu07StructuralSampling.test.ts @@ -0,0 +1,486 @@ +import assert from "node:assert/strict"; +import { execFile } from "node:child_process"; +import { access, mkdtemp, readFile, rm } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import test from "node:test"; +import { promisify } from "node:util"; + +import { + analyzeVideoStructure, + calculateSamplingDecision, + extractFramesFromLocalVideo, + extractVideoFramesFromBytes, + parseVideoStructuralAnalysis, + type VideoCommandRunner, + type VideoStructuralAnalysis, +} from "../../../src/lib/guardrails/videoBridgeRuntime.ts"; + +const execFileAsync = promisify(execFile); + +async function writeFrozenThenMotionFixture(fixturePath: string): Promise { + await execFileAsync( + "ffmpeg", + [ + "-hide_banner", + "-loglevel", + "error", + "-f", + "lavfi", + "-i", + "color=c=black:s=320x180:d=6:r=12", + "-f", + "lavfi", + "-i", + "testsrc2=s=320x180:d=4:r=12", + "-filter_complex", + "[0:v][1:v]concat=n=2:v=1:a=0,format=yuv420p[v]", + "-map", + "[v]", + "-c:v", + "libx264", + "-preset", + "ultrafast", + "-threads", + "1", + "-y", + fixturePath, + ], + { timeout: 30_000 } + ); +} + +const realRunner: VideoCommandRunner = async (executable, args, options) => { + const result = await execFileAsync(executable, [...args], { + encoding: "utf8", + maxBuffer: 1024 * 1024, + signal: options.signal, + timeout: options.timeoutMs, + }); + return { stderr: String(result.stderr), stdout: String(result.stdout) }; +}; + +function structuralAnalysis( + overrides: Partial = {} +): VideoStructuralAnalysis { + return { + freezeIntervals: [{ endSeconds: 6, startSeconds: 0 }], + samples: [ + { + blur: null, + brightness: 16, + sceneScore: 0, + spatialInformation: 0, + temporalInformation: 0, + timestampSeconds: 1, + }, + { + blur: 4.8, + brightness: 121, + sceneScore: 42, + spatialInformation: 120, + temporalInformation: 32, + timestampSeconds: 6, + }, + { + blur: 4.9, + brightness: 122, + sceneScore: 0, + spatialInformation: 118, + temporalInformation: 28, + timestampSeconds: 8, + }, + ], + sceneCandidates: [6], + ...overrides, + }; +} + +test("parses scene, freeze, blur, exposure, and spatial-temporal evidence", () => { + const metadata = [ + "frame:0 pts:0 pts_time:0", + "lavfi.scd.score=0.000", + "frame:0 pts:0 pts_time:0", + "lavfi.siti.si=0.00", + "frame:0 pts:0 pts_time:0", + "lavfi.siti.ti=0.00", + "frame:0 pts:0 pts_time:0", + "lavfi.blur=-nan", + "frame:0 pts:0 pts_time:0", + "lavfi.signalstats.YAVG=16", + "frame:6 pts:6 pts_time:6", + "lavfi.scd.score=41.013", + "frame:6 pts:6 pts_time:6", + "lavfi.siti.si=108.50", + "frame:6 pts:6 pts_time:6", + "lavfi.siti.ti=66.51", + "frame:6 pts:6 pts_time:6", + "lavfi.blur=4.75", + "frame:6 pts:6 pts_time:6", + "lavfi.signalstats.YAVG=121.5", + ].join("\n"); + const stderr = [ + "lavfi.freezedetect.freeze_start: 0", + "lavfi.freezedetect.freeze_duration: 6", + "lavfi.freezedetect.freeze_end: 6", + ].join("\n"); + + const analysis = parseVideoStructuralAnalysis(metadata, stderr, 10); + + assert.deepEqual(analysis.sceneCandidates, [6]); + assert.deepEqual(analysis.freezeIntervals, [{ endSeconds: 6, startSeconds: 0 }]); + assert.deepEqual(analysis.samples, [ + { + blur: null, + brightness: 16, + sceneScore: 0, + spatialInformation: 0, + temporalInformation: 0, + timestampSeconds: 0, + }, + { + blur: 4.75, + brightness: 121.5, + sceneScore: 41.013, + spatialInformation: 108.5, + temporalInformation: 66.51, + timestampSeconds: 6, + }, + ]); +}); + +test("runs all structural filters in one fixed, local-only, bounded FFmpeg pass", async () => { + const calls: Array<{ args: string[]; timeoutMs: number }> = []; + const runner: VideoCommandRunner = async (executable, args, options) => { + assert.equal(executable, "ffmpeg"); + calls.push({ args: [...args], timeoutMs: options.timeoutMs }); + return { + stderr: "lavfi.freezedetect.freeze_start: 0\nlavfi.freezedetect.freeze_end: 2", + stdout: "frame:0 pts:0 pts_time:0\nlavfi.scd.score=0", + }; + }; + + await analyzeVideoStructure("/tmp/input.mp4", { + durationSeconds: 8, + runner, + streamIndex: 2, + timeoutMs: 4_000, + }); + + assert.equal(calls.length, 1, "structural analysis must decode the video exactly once"); + assert.equal(calls[0].timeoutMs, 4_000); + assert.ok(calls[0].args.includes("-nostdin")); + assert.deepEqual(calls[0].args.slice(calls[0].args.indexOf("-map"), -1), [ + "-map", + "0:2", + "-vf", + calls[0].args[calls[0].args.indexOf("-vf") + 1], + "-an", + "-frames:v", + "600", + "-f", + "null", + ]); + const filter = calls[0].args[calls[0].args.indexOf("-vf") + 1]; + for (const expected of ["scdet", "freezedetect", "blurdetect", "signalstats", "siti"]) { + assert.match(filter, new RegExp(expected)); + } + assert.equal( + calls[0].args.some((argument) => argument.includes("://")), + false + ); +}); + +test("spends one frame on a frozen segment and reallocates the budget to dense motion", () => { + const analysis = structuralAnalysis(); + const decision = calculateSamplingDecision( + 10, + 4, + "segment_aware", + analysis.sceneCandidates, + null, + analysis + ); + + assert.equal(decision.policyEffective, "segment_aware"); + assert.equal(decision.timestamps.length, 4); + assert.equal(decision.timestamps.filter((timestamp) => timestamp < 6).length, 1); + assert.equal(decision.timestamps.filter((timestamp) => timestamp > 6).length, 3); +}); + +test("avoids redundant caption work for an entirely frozen video", () => { + const analysis = structuralAnalysis({ + freezeIntervals: [{ endSeconds: 8, startSeconds: 0 }], + samples: [ + { + blur: null, + brightness: 81, + sceneScore: 0, + spatialInformation: 0, + temporalInformation: 0, + timestampSeconds: 4, + }, + ], + sceneCandidates: [], + }); + const decision = calculateSamplingDecision(8, 8, "segment_aware", [], null, analysis); + + assert.equal(decision.policyEffective, "segment_aware"); + assert.equal(decision.timestamps.length, 1); + assert.deepEqual(decision.timestamps, [4]); +}); + +test("does not prune a moving clip when freeze evidence is absent", () => { + const analysis = structuralAnalysis({ + freezeIntervals: [], + samples: [ + { + blur: 4.8, + brightness: 120, + sceneScore: 0, + spatialInformation: 100, + temporalInformation: 30, + timestampSeconds: 1, + }, + { + blur: 4.9, + brightness: 122, + sceneScore: 0, + spatialInformation: 105, + temporalInformation: 32, + timestampSeconds: 7, + }, + ], + sceneCandidates: [], + }); + const decision = calculateSamplingDecision(8, 4, "segment_aware", [], null, analysis); + + assert.equal(decision.policyEffective, "segment_aware"); + assert.deepEqual(decision.timestamps, [1, 3, 5, 7]); +}); + +test("uses lower FFmpeg blur scores as sharper evidence for the extra frame", () => { + const common = { + brightness: 120, + sceneScore: 0, + spatialInformation: 50, + temporalInformation: 10, + }; + const analysis = structuralAnalysis({ + freezeIntervals: [], + samples: [ + { ...common, blur: 17, timestampSeconds: 1 }, + { ...common, blur: 4, timestampSeconds: 5 }, + ], + sceneCandidates: [4], + }); + const decision = calculateSamplingDecision(8, 3, "segment_aware", [4], null, analysis); + + assert.equal(decision.timestamps.filter((timestamp) => timestamp < 4).length, 1); + assert.equal(decision.timestamps.filter((timestamp) => timestamp > 4).length, 2); +}); + +test("malformed-only structural metadata fails open to uniform sampling", () => { + const analysis = parseVideoStructuralAnalysis("frame:0 pts:0 pts_time:0\nlavfi.blur=-nan", "", 8); + const decision = calculateSamplingDecision(8, 4, "segment_aware", [], null, analysis); + + assert.deepEqual(analysis.samples, []); + assert.equal(decision.policyEffective, "uniform"); + assert.deepEqual(decision.timestamps, [1, 3, 5, 7]); +}); + +test("keeps the long trailing segment when scene boundaries outnumber the frame budget", () => { + const decision = calculateSamplingDecision(20, 4, "segment_aware", [1, 2, 3, 4]); + + assert.equal(decision.timestamps.length, 4); + assert.ok( + decision.timestamps.some((timestamp) => timestamp > 4), + "the 16-second tail must not be dropped by early short cuts" + ); +}); + +test("preserves the legacy length-weighted allocation without structural evidence", () => { + const decision = calculateSamplingDecision(10, 8, "segment_aware", [2]); + + assert.equal(decision.policyEffective, "segment_aware"); + assert.deepEqual( + decision.timestamps.map((timestamp) => Number(timestamp.toFixed(3))), + [0.5, 1.5, 2.667, 4, 5.333, 6.667, 8, 9.333] + ); +}); + +test("does not report a focus-window boundary as usable segment evidence", () => { + const decision = calculateSamplingDecision(10, 4, "segment_aware", [2], { + endSeconds: 8, + startSeconds: 2, + }); + + assert.equal(decision.policyEffective, "uniform"); + assert.equal(decision.candidateCount, 0); + assert.deepEqual(decision.timestamps, [2.75, 4.25, 5.75, 7.25]); +}); + +test("does not claim segment-aware evidence that falls outside the focus window", () => { + const analysis = structuralAnalysis({ + freezeIntervals: [{ endSeconds: 10, startSeconds: 8 }], + samples: [{ timestampSeconds: 9, temporalInformation: 0 }], + sceneCandidates: [], + }); + const decision = calculateSamplingDecision( + 10, + 4, + "segment_aware", + [], + { endSeconds: 8, startSeconds: 2 }, + analysis + ); + + assert.equal(decision.policyEffective, "uniform"); + assert.deepEqual(decision.timestamps, [2.75, 4.25, 5.75, 7.25]); +}); + +test("structural timeout fails open to uniform while an abort stops extraction", async () => { + let analysisCalls = 0; + const timeoutRunner: VideoCommandRunner = async (_executable, args) => { + if (args.some((argument) => argument.includes("freezedetect"))) { + analysisCalls += 1; + throw new Error("structural deadline exceeded"); + } + return { stderr: "", stdout: "" }; + }; + + const frames = await extractFramesFromLocalVideo("/tmp/input.mp4", "/tmp/frames", { + durationSeconds: 8, + frameCount: 4, + runner: timeoutRunner, + samplingPolicy: "segment_aware", + streamIndex: 0, + timeoutMs: 250, + }); + assert.equal(analysisCalls, 1); + assert.equal(frames.sampling.policyEffective, "uniform"); + assert.deepEqual( + frames.map((frame) => frame.timestampSeconds), + [1, 3, 5, 7] + ); + + const controller = new AbortController(); + let frameExtractionCalls = 0; + const abortRunner: VideoCommandRunner = async (_executable, args, options) => { + if (args.some((argument) => argument.includes("freezedetect"))) { + assert.equal(options.signal, controller.signal); + controller.abort(); + throw new Error("aborted inside structural analysis"); + } + frameExtractionCalls += 1; + return { stderr: "", stdout: "" }; + }; + await assert.rejects( + () => + extractFramesFromLocalVideo("/tmp/input.mp4", "/tmp/frames", { + durationSeconds: 8, + frameCount: 4, + runner: abortRunner, + samplingPolicy: "segment_aware", + signal: controller.signal, + streamIndex: 0, + timeoutMs: 250, + }), + /aborted/ + ); + assert.equal(frameExtractionCalls, 0); +}); + +test("real FFmpeg evidence distinguishes a frozen dark segment from dense motion", async (t) => { + try { + await execFileAsync("ffmpeg", ["-version"], { timeout: 5_000 }); + } catch { + t.skip("FFmpeg is an optional runtime dependency"); + return; + } + + const directory = await mkdtemp(join(tmpdir(), "video-fu07-real-")); + const fixturePath = join(directory, "frozen-then-motion.mp4"); + try { + await writeFrozenThenMotionFixture(fixturePath); + + const analysis = await analyzeVideoStructure(fixturePath, { + durationSeconds: 10, + streamIndex: 0, + timeoutMs: 30_000, + }); + const decision = calculateSamplingDecision( + 10, + 4, + "segment_aware", + analysis.sceneCandidates, + null, + analysis + ); + + assert.ok(analysis.samples.length >= 8); + assert.ok(analysis.sceneCandidates.some((timestamp) => Math.abs(timestamp - 6) <= 1)); + assert.ok( + analysis.freezeIntervals.some( + (interval) => interval.startSeconds <= 1 && interval.endSeconds >= 5 + ) + ); + assert.ok(analysis.samples.some((sample) => (sample.spatialInformation ?? 0) > 20)); + assert.ok(analysis.samples.some((sample) => (sample.temporalInformation ?? 0) > 5)); + assert.ok(analysis.samples.some((sample) => (sample.blur ?? 0) > 0)); + assert.ok(analysis.samples.some((sample) => (sample.brightness ?? 255) < 24)); + assert.equal(decision.timestamps.filter((timestamp) => timestamp < 6).length, 1); + assert.equal(decision.timestamps.filter((timestamp) => timestamp > 6).length, 3); + } finally { + await rm(directory, { force: true, recursive: true }); + } +}); + +test("real FFmpeg abort stops preanalysis, skips frame extraction, and cleans the private tree", async (t) => { + try { + await execFileAsync("ffmpeg", ["-version"], { timeout: 5_000 }); + } catch { + t.skip("FFmpeg is an optional runtime dependency"); + return; + } + + const directory = await mkdtemp(join(tmpdir(), "video-fu07-abort-")); + const fixturePath = join(directory, "abort.mp4"); + const controller = new AbortController(); + let privateInputPath = ""; + let analysisStarted = false; + let frameExtractionCalls = 0; + try { + await writeFrozenThenMotionFixture(fixturePath); + const bytes = await readFile(fixturePath); + const runner: VideoCommandRunner = async (executable, args, options) => { + if (executable === "ffprobe") privateInputPath = args.at(-1) ?? ""; + if (args.some((argument) => argument.includes("freezedetect"))) { + analysisStarted = true; + setTimeout(() => controller.abort(), 25); + } else if (executable === "ffmpeg") { + frameExtractionCalls += 1; + } + return realRunner(executable, args, options); + }; + + await assert.rejects( + () => + extractVideoFramesFromBytes(bytes, { + frameCount: 4, + maxDurationSeconds: 600, + runner, + samplingPolicy: "segment_aware", + signal: controller.signal, + timeoutMs: 30_000, + }), + /aborted/ + ); + assert.equal(analysisStarted, true); + assert.equal(frameExtractionCalls, 0); + assert.notEqual(privateInputPath, ""); + await assert.rejects(() => access(privateInputPath)); + } finally { + await rm(directory, { force: true, recursive: true }); + } +});