Add bank statement analyzer with automatic work storage

2026-07-13 10:12:19 +05:30
parent d391e9b443
commit bbc5afe1c0
23 changed files with 1047 additions and 0 deletions
@@ -0,0 +1,54 @@
+from __future__ import annotations
+import re, pandas as pd
+from pathlib import Path
+from .base import *
+from .common import find
+
+class KotakParser(BaseParser):
+    bank_name='Kotak Mahindra Bank'; parser_name='KotakParser'
+    @classmethod
+    def detect(cls,text):
+        u=text.upper(); return 0.98 if 'KOTAK' in u and 'CHEQUE/REFERENCE#' in u and 'TRANSACTION DETAILS' in u else 0
+    def parse(self,path,text=None):
+        text=text or extract_text(path); meta=StatementMeta(bank_name=self.bank_name,source_file=Path(path).name,parser_name=self.parser_name,confidence='High')
+        meta.account_number=find(r'Account\s*#\s*Variant\s*KOTAK\s*\n.*?([0-9X*]{6,})',text,flags=re.I|re.S)
+        meta.ifsc=find(r'IFSC\s+([A-Z0-9]+)',text)
+        m=re.search(r'(\d{2}\s+[A-Za-z]{3},\s*\d{4})\s*-\s*(\d{2}\s+[A-Za-z]{3},\s*\d{4})',text)
+        if m:
+            meta.period_from=pd.to_datetime(m.group(1),dayfirst=True).strftime('%Y-%m-%d'); meta.period_to=pd.to_datetime(m.group(2),dayfirst=True).strftime('%Y-%m-%d')
+        date_re=re.compile(r'^\s*(\d{2}\s+[A-Za-z]{3},\s*\d{4})\s+(.*)$')
+        rows=[]; cur=None; page=1
+        for line in text.splitlines():
+            if '\f' in line: page+=line.count('\f')
+            m=date_re.match(line)
+            if m:
+                if ' - ' in m.group(2) and re.match(r'\d{2}\s+[A-Za-z]{3},',m.group(2).strip()): continue
+                if cur: rows.append(cur)
+                cur={'transaction_date':m.group(1),'value_date':m.group(1),'lines':[m.group(2)],'source_page':page}
+            elif cur and line.strip() and not re.match(r'^(DATE\s+TRANSACTION|Need help|Page \d+)',line.strip(),re.I): cur['lines'].append(line)
+        if cur: rows.append(cur)
+        out=[]
+        for r in rows:
+            first=r['lines'][0]; alltxt=norm(' '.join(r['lines']))
+            # amounts are signed in debit/credit columns, final unsigned balance
+            nums=list(re.finditer(r'([+-]?\d{1,3}(?:,\d{3})*\.\d{2})',first))
+            if not nums: continue
+            bal=amount(nums[-1].group(1)); debit=credit=None
+            if len(nums)>=2:
+                raw=nums[-2].group(1); txn=abs(amount(raw) or 0)
+                if raw.strip().startswith('-'): debit=txn
+                elif raw.strip().startswith('+'): credit=txn
+                else:
+                    # position fallback: debit column before credit
+                    credit=txn if nums[-2].start()>95 else None; debit=txn if nums[-2].start()<=95 else None
+            elif 'OPENING BALANCE' in first.upper():
+                meta.opening_balance=bal; continue
+            narr=first[:nums[-2].start() if len(nums)>=2 else nums[-1].start()].strip()+' '+' '.join(x.strip() for x in r['lines'][1:])
+            ref=''; z=re.search(r'\b(?:UPI|NEFT|IMPS|RTGS)-[A-Z0-9]+\b',alltxt,re.I); ref=z.group(0) if z else ''
+            if debit is None and credit is None: continue
+            out.append({**r,'narration':norm(narr),'reference_no':ref,'debit':debit,'credit':credit,'balance':bal})
+        if out:
+            if meta.opening_balance is None:
+                f=out[0]; meta.opening_balance=round((f['balance'] or 0)+(f.get('debit') or 0)-(f.get('credit') or 0),2)
+            meta.closing_balance=out[-1]['balance']
+        return meta,finalize(pd.DataFrame(out),meta)