PDF Statement Batch v0.2.0 — Save and reopen review batches

Continue a PDF review batch after closing the tab

The offline cache keeps the app and parser assets. v0.2.0 adds a separate, manual Batch → Save batch (.zip) action that includes original PDFs, corrected rows, review flags, file statuses and preferences. Use Open saved batch to continue, including any pending original files.

English app · 中文版 · Save/open change

• No document upload or automatic financial-file storage. Saved batch ZIPs contain readable financial data and remain under your control.
• Validates batch metadata and original-file hashes before replacing current work; an invalid import leaves it intact. Saved batches support up to 100 MB of original files and 1,000 files. Processing retains its 50 MB per-file boundary, including retries.
• Actual browser tests stopped a 500-input synthetic queue, saved it, reloaded disconnected, reopened the batch and resumed to 2,000 rows. Independent checks covered 18,000 fields and all 500 bundled source files. Inputs use two original layouts, not 500 banks.
• An additional actual-tab-close test restored selected file, edited rows, review flags, language, theme and numeric/deduplication settings; already parsed rows reopen without initializing the PDF worker.

The software ZIP contains a ready-to-serve app/, start guide, attribution and source receipt. It is separate from your saved .psb.zip batch. First parser initialization still requires a connection; a cached app alone is not saved work.

Original MIT parser credit remains with Priyanshu Dangi. New batch/review/export workflow by LydiaTools. This is an early preview with synthetic regression coverage.

ZIP SHA-256: 34eeac94514f15f430f7dca89cdfcd7f86b8865e68e693941c34251d467fd619.

PDF Statement Batch v0.1.0 — 对账单合表

PDF Statement Batch / 对账单合表 — v0.1.0 preview

Queue text-based PDF statements, review editable rows and download one Excel workbook with a complete file ledger. English and Chinese interfaces; no account, AI key or document upload.

• Open the browser app · 中文版.
• The ZIP contains a ready-to-serve app/ folder, source commit receipt, licenses and a short start guide. Python 3 can serve it locally; this is not a native installer. Pyodide is downloaded on first connected use, so this bundle is not cold offline.
• Source filenames, extracted row order, editable cells, review state, same-schema/currency grouping, literal text/date handling and explicit failed/canceled/duplicate statuses.
• Real browser checks processed 500 byte-distinct PDFs from two original synthetic layouts, producing 2,000 rows. Independent workbook readback checked 18,000 data fields and all 500 file statuses. Stop/resume, encrypted/image-only/broken inputs, incompatible columns/currencies and warm-cache disconnected use were exercised. Dated scope.

MIT. PDF extraction/parser core credited to Priyanshu Dangi; batch/review/export workflow by LydiaTools. See NOTICE.md. Synthetic tests do not establish real-bank coverage or overall extraction accuracy. Review source PDFs before bookkeeping use; no OCR or exact per-transaction PDF page coordinates.

ZIP SHA-256: 476d21669ce4240b2ea03f3750b1406eecb96e20bece776f7fb37a4b7f45dd4f.