FAQ

Aṣiri Data Ati Ibamu

Ṣe eyikeyi data fi awọn amayederun mi silẹ?

Rara — paapaa fun isediwon PDF. Bank Statement Parser ṣiṣẹ bi ile-ikawe ti ko ni ipinlẹ. Gbogbo sisẹ -- itupalẹ, isọdọtun PII, isediwon iwe-ipamọ -- waye laarin iranti asiko agbegbe rẹ. Opo gigun PDF aladapọ lo Ollama fun iṣawari LLM agbegbe — ko si awọn API awọsanma. Awọn parser XML ti wa ni lile pẹlu no_network=True, ti o dina gbogbo wiwọle ti njade ni ipele parser. Data inawo rẹ ko fi agbegbe rẹ silẹ rara.

Bawo ni isọdọtun PII ṣe n ṣiṣẹ?

Awọn aaye ifura ni a boju ṣaaju ki wọn de ọgbọn ohun elo rẹ. Parser naa ṣe idanimọ awọn orukọ onigbese, awọn orukọ onigbese, IBANs, ati awọn adirẹsi ifiweranṣẹ, ti o rọpo wọn pẹlu ***REDACTED*** ninu iṣelọpọ console ati ipo sisanwọle.

• Isọdọtun wa ni titan nipasẹ aiyipada ninu iṣelọpọ CLI ati ipo sisanwọle.
• Awọn okeere faili (CSV, JSON, Excel) ṣe idaduro data ti ko ni isọdọtun fun sisẹ isalẹ.
• Yan wọle si data ni kikun pẹlu --show-pii lori CLI tabi redact_pii=False ninu API.

Njẹ ilana isediwon jẹ ipinnu bi?

Bẹẹni fun awọn ọna kika ti a ṣeto — iṣelọpọ baiti-kan-naa ni gbogbo ṣiṣe. Fifun faili igbewọle kanna, awọn parser ipinnu (CAMT, PAIN.001, CSV, OFX, QFX, MT940) ṣe abajade esi kanna ni gbogbo igba. Ko si aileto, ko si itọkasi awoṣe, ko si iṣapẹẹrẹ heuristic.

Fun opo gigun PDF aladapọ, awọn ọna isediwon ti o da lori LLM le ṣe awọn iyatọ kekere laarin awọn ṣiṣe. Eyi ni idi ti a fi ṣe iṣayẹwo gbogbo isediwon PDF pẹlu Ofin Goolu (opening + credits − debits == closing) ati pe awọn aiyede ti a samisi le ṣe atunyẹwo ni ifọrọwanilẹnuwo.

CI fi agbara mu ipinnu pẹlu awọn idanwo 718 ni agbegbe 100%, pẹlu iruju ti o da lori ohun-ini nipasẹ Hypothesis.

Awọn iṣedede ibamu wo ni iṣẹ akanṣe naa tẹle?

Iṣẹ akanṣe naa ṣetọju awọn iwe ibamu ti o tẹle ISO 13485 pẹlu itọpa kikun:

• Iforukọsilẹ Ewu ti o ni iwọn pẹlu igbelewọn iwuwo/iṣeeṣe ati igbelewọn eewu iyokù.
• Eto Ijẹrisi ati Ifọwọsi pẹlu awọn igbesẹ 19 ti a ṣe ẹnu-ọna kọja awọn ipele 5.
• Ilana Iṣakoso Ayipada pẹlu iṣiro ipa ati awọn ilana yipo pada.
• Iforukọsilẹ SOUP ti o bo gbogbo awọn igbẹkẹle pẹlu awọn ipele eewu ati ipasẹ EOL.
• Maapu Itọpa ti n ṣe ìtọ́ka awọn igbewọle apẹrẹ si imuse ati ijẹrisi.

Gbogbo itusilẹ pẹlu CycloneDX SBOM, SHA-256 checksums, ati ẹri ijẹrisi ikole GitHub.

Iṣẹ Ṣiṣe Ati Iwọn

Bawo ni Bank Statement Parser ṣe yara to?

Awọn opin iṣẹ ṣiṣe jẹ ifọwọsi ni CI lori gbogbo igbimọ:

Iyara isediwon PDF da lori ọna ti a darí si: ipinnu (ti ko to iṣẹju-aaya kan), ọrọ-LLM (awọn iṣẹju-aaya), iran-LLM (awọn iṣẹju-aaya fun oju-iwe kọọkan).

Bawo ni a ṣe n ṣakoso awọn faili nla?

Sisanwọle pẹlu iranti ti o ni opin — a ṣe idanwo ni awọn iṣowo 50,000 fun faili kan. Lo parse_streaming() lati ṣe ilana awọn faili XML ni afikun. Iṣowo kọọkan ni a fi jade bi iwe-itumọ; awọn eroja ni a parẹ lẹhin sisẹ lati ṣe idiwọ idagbasoke iranti. Iranti ko dagba pẹlu iwọn faili — idanwo iṣowo-50K (25+ MB) lo kere ju iranti 2x idanwo iṣowo-10K.

Fun awọn faili ti o kọja 50 MB (fun apẹẹrẹ, awọn ipele host-to-host PAIN.001 pẹlu awọn sisanwo 100K+), parser naa ṣiṣan nipasẹ faili igba diẹ pẹlu yiyọ aaye orukọ ti o da lori chunk — iwe kikun ko ni kojọpọ sinu iranti rara.

Bawo ni a ṣe n ṣe ilana awọn ile-ipamọ ZIP ni aabo?

iter_secure_xml_entries() ṣe ifọwọsi ọmọ ẹgbẹ kọọkan ṣaaju isediwon:

• Fila iwọn titẹ sii (aiyipada 10 MB fun titẹ sii)
• Lapapọ fila iwọn ti a ko fi sinu (aiyipada 50 MB)
• Opin ipin funmorawon (aiyipada 100:1) lati ṣe idiwọ awọn bombu ZIP
• Ijusile titẹsi ti paroko

Ko si faili ti a kọ si disk. Awọn baiti XML kọja taara si parser nipasẹ from_bytes().

Ṣe mo le ṣe itupalẹ awọn faili lọpọlọpọ ni afiwe?

Bẹẹni. Lo parse_files_parallel() ti o pin iṣẹ kọja ProcessPoolExecutor:

from bankstatementparser import parse_files_parallel

results = parse_files_parallel([
    "statements/jan.xml",
    "statements/feb.xml",
    "statements/mar.xml",
])
for r in results:
    print(r.path, r.status, len(r.transactions), "rows")

Fun gbigbe PDF pupọ, lo scan_and_ingest() ti o ṣe ilana awọn igi folda odidi pẹlu yiyọ ẹda laifọwọyi.

Awọn Ọna Kika Ti A Ṣe Atilẹyin

Awọn ọna kika alaye banki wo ni a ṣe atilẹyin?

Ọna kika	Boṣewa	Awọn Oriṣi Faili	Parser/Ọna
CAMT.053	ISO 20022 Banki-si-Onibara Gbólóhùn	.xml	CamtParser
PAIN.001	ISO 20022 Ibẹrẹ Gbigbe Kirẹditi	.xml	Pain001Parser
CSV	Okeere banki gbogbogbo	.csv	CsvStatementParser
OFX	Open Financial Exchange	.ofx	OfxParser
QFX	Quicken Financial Exchange	.qfx	QfxParser
MT940	Boṣewa SWIFT	.mt940, .sta	Mt940Parser
PDF	Awọn alaye oni-nọmba ati ti a ṣe scan	.pdf	smart_ingest()

Bawo ni opo gigun PDF aladapọ ṣe n ṣiṣẹ?

Opo gigun aladapọ (v0.0.5+) darí awọn PDF nipasẹ awọn ọna isediwon mẹta ni ọgbọn:

• Ọna A (Ipinnu): Awọn tabili PDF ti a ṣeto ti a ṣe itupalẹ taara — ọfẹ, yara julọ, ko nilo LLM.
• Ọna B (Ọrọ-LLM): Awọn PDF oni-nọmba pẹlu awọn eto eka ti a yọ jade nipasẹ LLM agbegbe (LiteLLM/Ollama).
• Ọna C (Iran-LLM): Awọn alaye ti a ṣe scan tabi fọtokopi ti a ṣe ilana pẹlu awọn awoṣe iran olopọ-iru.

A ṣe iṣayẹwo gbogbo isediwon pẹlu Ofin Goolu (opening + credits − debits == closing). Awọn aiyede le ṣe atunyẹwo ni ifọrọwanilẹnuwo pẹlu --type review.

Ṣe parser naa ṣe itọju awọn ede-ede banki pato ti CAMT.053?

Bẹẹni — aibikita aaye orukọ nipasẹ apẹrẹ. Parser naa yọ awọn aaye orukọ XML kuro ṣaaju sisẹ, ti o mu eyikeyi iyatọ CAMT.053 (camt.053.001.02, camt.053.001.04, tabi awọn iwe-ipamọ banki ohun-ini) laisi iṣeto pato aaye orukọ. Awọn ibeere XPath fojusi eto eroja, kii ṣe awọn URI aaye orukọ.

Fun awọn banki ti o fi CAMT sinu apoowe aṣa, lo from_string() tabi from_bytes() lati fi iwe inu funni taara.

Ṣe mo le ṣe ìtọ́ka awọn akọle ọwọn CSV aṣa si ero boṣewa?

Bẹẹni — isọdọtun aifọwọyi, iṣeto odo. CsvStatementParser mọ awọn iyatọ akọsori ti o wọpọ: "Date", "Transaction Date", "Booking Date" gbogbo ṣe ìtọ́ka si aaye date. "Amount", "Value", "Sum" ṣe ìtọ́ka si amount. Awọn ọwọn kirẹditi/debiti ti a pin (fun apẹẹrẹ, "Credit" ati "Debit") ni a ṣe iwari ti a si dapọ sinu iye ti a fowo si kan ṣoṣo laifọwọyi.

Kini ọna kika iṣelọpọ?

Gbogbo awọn parser ṣe agbejade pandas DataFrames deede pẹlu awọn iru ọwọn ti o ni ibamu:

Ọna kika	Awọn Ọwọn Pataki
CAMT	Amount, Currency, DrCr, Debtor, Creditor, Reference, ValDt, BookgDt, AccountId
PAIN.001	PmtInfId, PmtMtd, InstdAmt, Currency, CdtrNm, EndToEndId, MsgId, CreDtTm, NbOfTxs
CSV/OFX/QFX/MT940	date, description, amount (deede)

O tun le ṣe okeere si CSV, JSON, Excel, Polars DataFrames, hledger, tabi ọna kika iwe iroyin beancount.

Awọn Ẹya PDF Ati LLM

Awọn awoṣe LLM wo ni opo gigun aladapọ ṣe atilẹyin?

Opo gigun naa lo LiteLLM bi ipele iyọkuro awoṣe, pẹlu afara Ollama taara fun awọn ifiranṣẹ iran. Awọn awoṣe ti a ṣe igbaniyanju:

• Isediwon ọrọ: Eyikeyi awoṣe ti o ni ibamu pẹlu LiteLLM (agbegbe tabi latọna jijin).
• Isediwon iran: ollama/minicpm-v (a ṣe igbaniyanju) fun awọn PDF ti a ṣe scan.
• Isọri: Eyikeyi awoṣe ti o ni ibamu pẹlu LiteLLM.

Gbogbo awọn awoṣe le ṣiṣẹ 100% ni agbegbe nipasẹ Ollama — ko nilo awọn bọtini API.

Kini iṣayẹwo Ofin Goolu?

A ṣe iṣayẹwo gbogbo isediwon PDF pẹlu idogba: opening balance + credits − debits == closing balance. Awọn abajade ni a samisi bi:

• VERIFIED: Awọn iwọntunwọnsi baramu ni deede.
• DISCREPANCY: Awọn iwọntunwọnsi ko baramu — a ṣe igbaniyanju atunyẹwo.
• FAILED: A ko le ṣe iṣayẹwo (data iwọntunwọnsi ko si).

Ṣe mo le ṣe isọri awọn iṣowo laifọwọyi?

Bẹẹni. Apa imudara (v0.0.6+) pese isọri iṣowo ti LLM ṣe:

from bankstatementparser.enrichment import Categorizer

categorizer = Categorizer()
enriched = categorizer.categorize_batch(transactions)

Ero aiyipada lo awọn ẹka 13 ti o ni ibamu pẹlu Plaid. O le pese ero ẹka tirẹ.

Ṣe mo le ṣe okeere si hledger tabi beancount?

Bẹẹni (v0.0.8+). Ṣe okeere awọn iṣowo si awọn ọna kika iwe iroyin iṣiro ọrọ-lasan pẹlu ìtọ́ka àkáǹtì:

from bankstatementparser.export import to_hledger, to_beancount

journal = to_hledger(transactions, account="Assets:Bank:Checking")

Awọn Ṣiṣan Iṣẹ Iṣura

Bawo ni parser naa ṣe mu awọn alaye owo-pupọ?

Iṣowo kọọkan ṣe itọju owo atilẹba rẹ — ko si iyipada ti ko tọ. Aaye Currency ni a yọ jade lati ikalara XML Ccy fun iṣowo kọọkan. Awọn alaye owo-pupọ wa bi-ti-jẹ. Ọna get_account_balances() da awọn iwọntunwọnsi ṣiṣi ati pipade padà fun akọọlẹ kọọkan pẹlu awọn koodu owo atilẹba.

Lati v0.0.8, verify_balance_multi_currency() ṣe akojọpọ awọn iṣowo nipasẹ owo ti o si ṣe iṣayẹwo Ofin Goolu ni ominira fun ẹgbẹ kọọkan — o wulo fun awọn akọọlẹ ti o mu awọn owo pupọ.

Ṣe parser naa ṣe atilẹyin mejeeji awọn ọna kika ti njade ati ti nwọle?

Bẹẹni. Pain001Parser mu awọn faili ibẹrẹ gbigbe kirẹditi ISO 20022 PAIN.001 (awọn sisanwo ti njade). CamtParser mu awọn faili alaye banki-si-onibara CAMT.053 (ijabọ ti nwọle). Mejeeji ṣe atilẹyin sisanwọle, isọdọtun PII, ati okeere si CSV, JSON, Excel, hledger, ati beancount. Lo detect_statement_format() lati ṣe idanimọ ọna kika laifọwọyi.

Kini yoo ṣẹlẹ nigbati titẹ sii iṣowo kan bajẹ?

Ihuwasi da lori ipo itupalẹ:

• parse() (ipo ipele) -- Awọn titẹ sii ti ko dara ti o padanu awọn aaye ti a beere (Amount, Currency, tabi CdtDbtInd) ni a fo pẹlu ikilọ akọsilẹ. Iyokù alaye naa ṣe itupalẹ daradara.
• parse_streaming() (ipo sisanwọle) -- Awọn aṣiṣe itupalẹ tan lẹsẹkẹsẹ bi awọn imukuro. Ko si ipadanu data. Ihuwasi ikuna-iyara yii jẹ ipinnu fun awọn ṣiṣan iṣẹ inawo nibiti gbogbo iṣowo gbọdọ jẹ iṣiro fun.
• smart_ingest() (PDF aladapọ) -- A mu awọn aṣiṣe isediwon sinu IngestResult pẹlu ipo iṣayẹwo, ti o jẹ ki atunyẹwo ifọrọwanilẹnuwo ṣeeṣe.

Bawo ni yiyọ ẹda ṣe n ṣiṣẹ?

A fi transaction_hash aláìyípadà (àtẹ́ka MD5) fun iṣowo kọọkan ti o da lori awọn aaye pataki rẹ. Eyi jẹ ki gbigbe afikun to ni aabo ṣeeṣe — atunṣe faili kanna ṣe awọn hash kanna, nitorinaa a ṣe iwari awọn ẹda-iwe laifọwọyi.

from bankstatementparser import CamtParser, Deduplicator

parser = CamtParser("statement.xml")
dedup = Deduplicator()
result = dedup.deduplicate(dedup.from_dataframe(parser.parse()))

print(f"Unique: {len(result.unique_transactions)}")
print(f"Exact duplicates: {len(result.exact_duplicates)}")
print(f"Suspected matches: {len(result.suspected_matches)}")

Fifi Sori Ẹrọ Ati Ibamu

Bawo ni mo ṣe fi Bank Statement Parser sori ẹrọ?

# Fifi sori ẹrọ ipilẹ (awọn parser ipinnu nikan)
pip install bankstatementparser

# Opo gigun PDF aladapọ
pip install 'bankstatementparser[hybrid]'         # Ọna ọrọ-LLM
pip install 'bankstatementparser[hybrid-vision]'   # Ọna iran-LLM

# Awọn afikun
pip install 'bankstatementparser[enrichment]'      # Isọri iṣowo
pip install 'bankstatementparser[api]'             # Iṣẹ-kekere REST API
pip install 'bankstatementparser[polars]'          # Atilẹyin Polars DataFrame

Awọn ẹya Python wo ni a ṣe atilẹyin?

Python 3.10 nipasẹ 3.14. A ti yọ atilẹyin Python 3.9 kuro ni v0.0.6 (EOL 2025-10-31). Gbogbo awọn ẹya ni a ṣe idanwo ni CI pẹlu awọn idanwo 718 ni agbegbe 100%.

Kini awọn igbẹkẹle?

Ile-ikawe ipilẹ ni awọn igbẹkẹle taara 5:

• lxml -- Itupalẹ XML pẹlu lile aabo
• pandas -- DataFrames ati ifọwọyi data
• openpyxl -- Okeere Excel
• pydantic -- Ifọwọsi data ati awọn awoṣe
• defusedxml -- Idaabobo XXE

Awọn afikun iyan fi kun: litellm, pypdf, pdfplumber, pypdfium2, fastapi, uvicorn, polars.

Gbogbo awọn igbẹkẹle ni awọn ẹya ti SHA-256 hash tiipa. CycloneDX SBOM ṣe ìtọ́ka gbogbo paati asiko ṣiṣe.

Ṣe o ṣiṣẹ lori macOS, Linux, ati Windows?

Bẹẹni. Ile-ikawe naa ṣiṣẹ lori macOS, Linux, ati Windows (nipasẹ WSL). Ko ni igbẹkẹle pato si ẹrọ kan.

Ṣe REST API wa?

Bẹẹni (v0.0.8+). Fi sori ẹrọ pẹlu pip install 'bankstatementparser[api]' ki o ṣiṣẹ:

bankstatementparser-api --port 8000

Awọn endpoint: POST /ingest (ṣe itupalẹ alaye) ati GET /health (ayẹwo ilera).

Isọdọtun Ati Aabo

Bawo ni mo ṣe le rii daju isọdọtun?

python -m pytest                              # 718 tests, 100% branch coverage
python scripts/verify_locked_hashes.py        # SHA-256 hash verification
git log --show-signature -1                   # Verify commit signature

Awọn idaabobo aabo wo ni a kọ sinu?

• Idaabobo XXE: resolve_entities=False, no_network=True, load_dtd=False
• Idaabobo Bombu ZIP: Awọn opin ipin funmorawon, awọn bọtini iwọn iwọle, ijusile iwọle ti paroko
• Idena Ipa-ọna Rin-kakiri: Atokọ apẹrẹ eewu ati ipinnu ami asopọ
• Ifọwọsi Titẹ Sii: Awọn opin iwọn faili (aiyipada 100 MB), itẹsiwaju/afọwọsi ọna kika
• Ẹwọn Ipese: SHA-256 awọn igbẹkẹle ti hash tiipa, CycloneDX SBOM, ẹri ijẹrisi ikole
• Awọn Igbimọ Ti A Fọwọ Si: A fi agbara mu ni CI
• Awọn LLM Agbegbe: Opo gigun PDF aladapọ lo Ollama — ko si awọn ipe API awọsanma

Bawo ni Bank Statement Parser ṣe afiwe si pyiso20022?

pyiso20022 jẹ ohun elo ISO 20022 gbooro ti o ṣe agbekalẹ awọn kilasi data Python lati awọn ero ISO XML. O bo ọpọlọpọ awọn iru ifiranṣẹ ISO 20022 (PACS, PAIN, CAMT, ADMI) pẹlu afọwọsi ero. Bank Statement Parser ni a kọ ni pataki fun itupalẹ alaye banki pẹlu atilẹyin PDF aladapọ, iṣayẹwo iwọntunwọnsi, imudara, okeere iwe iroyin, ati API iṣọkan kọja awọn ọna kika meje pẹlu awọn ọna kika ti kii ṣe ISO (CSV, OFX, QFX, MT940, PDF). Ti o ba nilo lati ṣe itupalẹ awọn alaye banki sinu DataFrames pẹlu aabo ipele iṣelọpọ, lo Bank Statement Parser. Ti o ba nilo lati ṣiṣẹ pẹlu katalogi ifiranṣẹ ISO 20022 ni kikun, lo pyiso20022.

Kini awọn akoko ipari iṣilọ SWIFT ISO 20022?

SWIFT ti ṣe atẹjade akoko iṣilọ ipele kan:

• Oṣu kọkanla 2026: Awọn adirẹsi ti a ṣeto ati arabara di dandan. Awọn ifiranṣẹ itọnisọna-pupọ MT101 yoo jẹ kiko. Ipele 1 Isakoso Ọran bẹrẹ.
• Oṣu kọkanla 2027: Gbogbo awọn ile-iṣẹ inawo gbọdọ ni anfani lati gba awọn alaye CAMT.053 ni abinibi. SWIFT yoo da iyipada MT si ọna kika ISO duro.
• Oṣu kọkanla 2028: Ifẹhinti ni kikun ti MT940, MT942, MT950, MT900, ati MT910. Iwọnyi yoo rọpo nipasẹ awọn deede CAMT.052, CAMT.053, ati CAMT.054.

Bank Statement Parser ṣe atilẹyin ọna kika MT940 atijọ ati awọn ọna kika CAMT.053/PAIN.001 ode oni, ti o jẹ ki o dara julọ fun akoko iyipada.