Align video catalogue limits and OCR coverage with native evidence

src/memory/document-evidence.ts

@@ -20,6 +20,9 @@ function record(value:unknown):Record<string,unknown>{if(!value||typeof value!==
 function text(value:unknown,max=4096,empty=false):string {
  if(typeof value!=='string'||value.length>max||(!empty&&!value.length)||decoder.decode(encoder.encode(value))!==value)throw Error('Invalid document evidence text');return value;
 }
+function filename(value:unknown):string{
+ const result=text(value,2048);if(Array.from(result).length>1024)throw Error('Document filename exceeds its character limit');return result;
+}
 function count(value:unknown,max:number,min=0):number{if(typeof value!=='number'||!Number.isSafeInteger(value)||value<min||value>max)throw Error('Invalid document evidence count');return value;}
 function list(value:unknown,max:number):unknown[]{if(!Array.isArray(value)||value.length>max)throw Error('Document evidence exceeds its list limit');return value;}
 function attachment(value:unknown):DocumentAttachment {
@@ -38,7 +41,7 @@ function references(value:unknown,allowed:string[]):CellReference[]{return list(
 export function parseDocumentEvidence(value:unknown,source:DocumentSource):DocumentEvidence {
  const v=record(value);sameAttachment(v.original,source.binding.original);sameAttachment(v.manifest,source.binding.manifest);
  const format=text(v.format,64);if(format!==source.binding.format)throw Error('Document format does not match this source');
- const filename=text(v.filename,1024),parser=text(v.parser,128),segments:EvidenceSegment[]=[],cells=new Map<string,DocumentCell>();
+ const documentFilename=filename(v.filename),parser=text(v.parser,128),segments:EvidenceSegment[]=[],cells=new Map<string,DocumentCell>();
  const metadata=v.metadata===undefined?{}:record(v.metadata),pdfOcr=metadata.pdf_ocr===undefined?undefined:parsePdfOcrEvidence(metadata.pdf_ocr);
  if(pdfOcr&&format!=='pdf')throw Error('PDF OCR settings do not match the document format');
  const occupied=new Set<string>(),identities=new Set<string>(),tables=new Map<string,DocumentCell[]>();
@@ -81,7 +84,7 @@ export function parseDocumentEvidence(value:unknown,source:DocumentSource):Docum
    }
   }}
  }
- return {filename,format,parser,segments,cells,tables:Array.from(tables,([locator,cells])=>({locator,cells:cells.sort((a,b)=>a.row-b.row||a.column-b.column)})),notes,pdfOcr,media:parseMediaTranscript(format,parser,metadata,list(v.segments,20000))};
+ return {filename:documentFilename,format,parser,segments,cells,tables:Array.from(tables,([locator,cells])=>({locator,cells:cells.sort((a,b)=>a.row-b.row||a.column-b.column)})),notes,pdfOcr,media:parseMediaTranscript(format,parser,metadata,list(v.segments,20000))};
 }
 export function cellExcerpt(cell:DocumentCell,evidence:DocumentEvidence){
  const segment=evidence.segments[cell.segment],bytes=encoder.encode(segment.text),start=cell.start-segment.start,end=cell.end-segment.start;

src/memory/document-video.ts

@@ -19,6 +19,9 @@ function text(value:unknown,max:number):string{
  if(typeof value!=='string'||!value.length||value.length>max||decoder.decode(encoder.encode(value))!==value)throw Error('Invalid video evidence text.');
  return value;
 }
+function characters(value:unknown,max:number):string{
+ const result=text(value,max*2);if(Array.from(result).length>max)throw Error('Video text exceeds its character limit.');return result;
+}
 function integer(value:unknown,max:number,min=0):number{
  if(typeof value!=='number'||!Number.isSafeInteger(value)||value<min||value>max)throw Error('Invalid video evidence count.');
  return value;
@@ -41,19 +44,21 @@ function timeBase(value:unknown):[string,bigint,bigint]{
 }
 function regions(value:unknown,content:string):VideoRegion[]{
  const encoded=encoder.encode(content);let previous=0;
- return list(value,10000).map(item=>{
-  const v=record(item),start=integer(v.start,encoded.length),end=integer(v.end,encoded.length),label=text(v.text,100000);
+ const result=list(value,20000).map(item=>{
+  const v=record(item),start=integer(v.start,encoded.length),end=integer(v.end,encoded.length),label=characters(v.text,100000);
   if(v.coordinate_space!=='normalized_displayed_frame_top_left'||start<previous||end<=start
-   ||decoder.decode(encoded.subarray(start,end))!==label)throw Error('Video region does not match its UTF-8 source span.');
+   ||decoder.decode(encoded.subarray(start,end))!==label||decoder.decode(encoded.subarray(previous,start)).trim())throw Error('Video region does not match its UTF-8 source span.');
   previous=end;
   const box=list(v.box,4);
   if(box.length!==4||!box.every(n=>typeof n==='number'&&Number.isFinite(n)&&n>=0&&n<=1))throw Error('Invalid video OCR box.');
   const rectangle=box as [number,number,number,number];
   if(rectangle[0]>=rectangle[2]||rectangle[1]>=rectangle[3])throw Error('Video OCR box is empty.');
   const score=v.score??null;
   if(score!==null&&(typeof score!=='number'||!Number.isFinite(score)||score<0||score>1))throw Error('Invalid OCR recognizer score.');
-  return {text:label,start,end,box:[...rectangle],score};
+  return {text:label,start,end,box:[...rectangle] as [number,number,number,number],score};
  });
+ if(decoder.decode(encoded.subarray(previous)).trim())throw Error('Video OCR regions omit recognized text.');
+ return result;
 }
 
 export function parseVideoCatalogue(value:unknown,source:DocumentSource,episodeId:number,space:string):VideoCatalogue{
@@ -84,7 +89,7 @@ export function parseVideoCatalogue(value:unknown,source:DocumentSource,episodeI
   const relative=(position-start)*numerator;
   if(relative<BigInt(requestedSeconds.at(-1)!)*denominator||relative>=duration)throw Error('Video frame falls outside its sampling interval.');
   totalBytes+=bytes;lastOrdinal=ordinal;lastTimestamp=position;
-  return {ordinal,presentationTimestamp:pts,requestedSeconds,width,height,bytes,sha256:hash(f.png_sha256),ocrEngine:text(f.ocr_engine,96),
+  return {ordinal,presentationTimestamp:pts,requestedSeconds,width,height,bytes,sha256:hash(f.png_sha256),ocrEngine:characters(f.ocr_engine,96),
    empty:f.empty,text:'',regions:[] as VideoRegion[]};
  });
  if(!frames.length||totalBytes>maxTotalBytes||covered+unavailableRequests!==requests.length)throw Error('Video sampling coverage is inconsistent.');
@@ -100,7 +105,7 @@ export function parseVideoCatalogue(value:unknown,source:DocumentSource,episodeI
  if(frames.some(frame=>!frame.empty&&!seen.has(String(frame.ordinal))))throw Error('Video frame text is missing.');
  const modelRevision=text(v.model_revision,128);
  if(!/^[A-Za-z0-9][A-Za-z0-9._-]*$/.test(modelRevision))throw Error('Invalid video OCR model revision.');
- return {document,sourceSha256,decoderRevision:hash(v.decoder_revision),policyRevision:text(v.policy_revision,96),modelRevision,
+ return {document,sourceSha256,decoderRevision:hash(v.decoder_revision),policyRevision:characters(v.policy_revision,96),modelRevision,
   streamIndex,timeBase:base,startTimestamp,durationTicks,intervalSeconds,decodedFrames,unavailableRequests,frames};
 }
 

tests/document-video.test.ts

@@ -122,3 +122,32 @@ test('all-empty sampled frames retain a catalogue with no fabricated source text
  f.catalogue.evidence.video.frames[0].empty=false;
  assert.throws(()=>parseVideoCatalogue(f.catalogue,f.source,7,'alpha'));
 });
+
+test('one sampled frame may carry the native 20000-region allowance',()=>{
+ const f=videoFixture(),segment=f.catalogue.evidence.segments[0];
+ segment.text=Array.from({length:10001},()=> 'x').join('\n');f.source.content=segment.text;
+ segment.regions=Array.from({length:10001},(_,i)=>({...segment.regions[0],text:'x',start:i*2,end:i*2+1}));
+ assert.equal(parseVideoCatalogue(f.catalogue,f.source,7,'alpha').frames[0].regions.length,10001);
+});
+test('OCR regions must account for nonwhitespace text between and after boxes',()=>{
+ for(const [content,start,end] of [['prefix Café',7,12],['Café suffix',0,5]] as const){
+  const f=videoFixture(),segment=f.catalogue.evidence.segments[0];
+  segment.text=content;f.source.content=content;segment.regions[0].start=start;segment.regions[0].end=end;
+  assert.throws(()=>parseVideoCatalogue(f.catalogue,f.source,7,'alpha'));
+ }
+});
+test('video OCR metadata uses native Unicode character limits',()=>{
+ const f=videoFixture(),v=f.catalogue.evidence.video;
+ v.frames[0].ocr_engine='😀'.repeat(96);f.catalogue.evidence.segments[0].metadata.engine=v.frames[0].ocr_engine;
+ v.policy_revision='😀'.repeat(96);
+ assert.equal(parseVideoCatalogue(f.catalogue,f.source,7,'alpha').frames[0].ocrEngine,v.frames[0].ocr_engine);
+});
+test('native supplementary-character filenames fit the catalogue character limit',()=>{
+ const f=videoFixture();f.catalogue.evidence.filename='😀'.repeat(512)+'.mp4';
+ assert.equal(parseVideoCatalogue(f.catalogue,f.source,7,'alpha').document.filename,f.catalogue.evidence.filename);
+});
+test('a native-valid OCR region counts Unicode characters rather than UTF16 units',()=>{
+ const f=videoFixture(),label='😀'.repeat(50001),segment=f.catalogue.evidence.segments[0];
+ segment.text=label;f.source.content=label;segment.regions[0].text=label;segment.regions[0].end=new TextEncoder().encode(label).length;
+ assert.equal(parseVideoCatalogue(f.catalogue,f.source,7,'alpha').frames[0].regions[0].text,label);
+});