Changelog

Tẹle idagbasoke Parser Gbólóhùn Gbólóhùn Bank. Alabapin nipasẹ [RSS] (/changelog/rss.xml) tabi wo ibi ipamọ GitHub fun awọn iwifunni idasilẹ.

v0.0.8 — 2026-04-11 (Latest) — "Full Platform"

• Multi-currency balance verification — verify_balance_multi_currency() groups by currency, runs Golden Rule per group.
• hledger + beancount export — to_hledger() and to_beancount() in bankstatementparser.export.
• Bulk directory scanner — scan_and_ingest() scans folder trees, deduplicates across batch.
• Account mapping rules — AccountMapper with ordered regex rules from JSON config.
• REST API — FastAPI wrapper with /ingest and /health endpoints ([api] extra).

v0.0.7 — 2026-04-08 — "Universal Vision"

• Direct Ollama bridge (ollama_direct_completion) — bypasses LiteLLM long-prompt hang.
• Strip mode (VisionExtractor.strip_rows=True) — splits dense pages into overlapping bands for small local models.
• Recommended vision model changed from llava to minicpm-v.

v0.0.6 — 2026-04-08 — "Intelligence Layer"

• Dropped Python 3.9 support (now 3.10-3.14).
• Enrichment module (Categorizer, EnrichedTransaction, DEFAULT_CATEGORY_SCHEMA).
• Interactive review mode with --type review CLI command.
• Per-row bounding box extraction (Transaction.source_bbox).

v0.0.5 — 2026-04-08 — "Universal Extraction"

• Hybrid PDF pipeline (smart_ingest()) with deterministic/text-LLM/vision-LLM routing.
• LLMExtractor for digital PDFs via LiteLLM.
• VisionExtractor for scanned PDFs via multimodal vision models.
• Golden Rule balance verification (opening + credits - debits == closing).
• Idempotent deduplication via transaction_hash (MD5 fingerprint).

v0.0.4 - 2026-03-15 (Titun)

• Ṣafikun faili afiwera pẹluparse_files_parallel()lilo ProcessPoolExecutor. Fikun ṣiṣan otitọ fun awọn faili PAIN.001 nla (50 MB+) pẹlu iranti didi.
• Awọn iṣapeye iṣẹ: Iwọn CAMT kọja 27,000 tx/s, PAIN.001 kọja 52,000 tx/s.
• Fi kunDeduplicatorkilasi fun wiwa awọn ẹda-ẹda deede ati awọn ifura ti a fura si pẹlu awọn ikun igbekele.
• Fi kunfrom_string()atifrom_bytes()awọn ọna fun ni-iranti parsing lai disk I/O.
• Fi kuniter_secure_xml_entries()fun ni aabo ZIP pamosi sisẹ.
• CI ti o gbooro sii pẹlu imuṣiṣẹ ala iṣẹ.

v0.0.3 - 2025-11-20

• Ṣafikun CSV, OFX, QFX, ati atilẹyin parser MT940.
• Fikun kika laifọwọyi erin pẹludetect_statement_format()aticreate_parser().
• Atunṣe PII ti a ṣafikun (ni aiyipada ni CLI ati ipo ṣiṣanwọle).
• Awọn oluranlọwọ okeere ti a ṣafikun fun CSV, JSON, ati Excel.
• Ṣe afikun atilẹyin Polars DataFrame iyan.
• Suite idanwo gbooro si awọn idanwo 718 pẹlu agbegbe agbegbe 100%.

v0.0.2 - 2025-06-10

-Afikun PAIN.001 parser (Pain001Parser) fun ISO 20022 awọn faili ibẹrẹ gbigbe kirẹditi. Ni wiwo CLI ti a ṣafikun (python -m bankstatementparser.cli).

• Fi kun ipo sisanwọle pẹluparse_streaming().
• Fikun afọwọsi igbewọle ati awọn opin iwọn faili.

v0.0.1 - 2025-01-15

• Itusilẹ akọkọ.
• CAMT.053 atọka (CamtParser) fun ISO 20022 awọn alaye banki-si-onibara.
• pandas DataFrame o wu.
• Ipilẹ aabo XML líle (Aabo XXE, no_nẹtiwọọki).

Wo itan ifaramo ni kikun lori GitHub.