accounting-bridge 0.6.0

Deterministic multi-ERP financial consolidation, POS-to-journal, and GL audit engine for SMEs

accounting-bridge

Deterministic multi-ERP financial consolidation, POS-to-journal conversion, and GL audit — as a typed Python library.

Every finance team with two or more entities runs the same manual loop each month: export trial balances from each ERP, rearrange columns, VLOOKUP account codes, reconcile currencies, and hunt for the difference when it doesn't balance. accounting-bridge does that loop in seconds — and it is deterministic: no LLM sits in the data path, so every number is reproducible and auditable, with row-level lineage back to the source file.

It reads exports from any accounting system (CSV, Excel, PDF, OFX/QFX, QuickBooks IIF, scanned images) and produces:

• Consolidated trial balances across entities and currencies, with IAS 21 FX translation and a cumulative translation adjustment.
• POS-to-journal conversion with balanced double entries and Canadian GST/QST tax splitting.
• GL audit packages with anomaly flags, balance checks, and Benford's-law analysis.

Install

pip install accounting-bridge          # core library, zero UI/server deps
pip install "accounting-bridge[pdf]"   # + PDF table extraction
pip install "accounting-bridge[ocr]"   # + scanned-image OCR
pip install "accounting-bridge[all]"   # everything

Requires Python 3.11+.

Library quickstart

The library's front door is a small set of typed functions that take a validated config and return a named result.

import pandas as pd
from accounting_bridge import consolidate, ConsolidationConfig

hq = pd.DataFrame({"account_code": ["4100"], "amount_local": [-10_000.0], "currency": ["USD"]})
eu = pd.DataFrame({"account_code": ["4100"], "amount_local": [-8_000.0], "currency": ["EUR"]})

cols = {"account_code": "account_code", "amount_local": "amount_local", "currency": "currency"}
config = ConsolidationConfig(
    columns_per_entity={"HQ": cols, "EU": cols},
    base_currency="USD",
    rates={"EUR": {"closing": 1.08, "average": 1.07, "historical": 1.05}},
)

result = consolidate([("HQ", hq), ("EU", eu)], config)
print(result.summary)        # consolidated by account, in USD
print(result.fx_warnings)    # any FX/rate issues surfaced

Other entry points follow the same shape:

from accounting_bridge import convert_pos, audit_ledger, summarize_trial_balance
from accounting_bridge import PosConfig, AuditConfig

pos_result   = convert_pos(pos_df, PosConfig(columns=..., accounts=...))
audit_result = audit_ledger(gl_df, AuditConfig(columns=..., large_amount=100_000))
tb_result    = summarize_trial_balance(tb_df, {"account_code": "Acct", "amount": "Balance"})

Reading messy real-world files:

from accounting_bridge.ingest import read_smart

df, diagnostics = read_smart("messy_export.xlsx")  # encoding, delimiter, header auto-detected

Command line

# Consolidate several trial balances into one audit-ready workbook
accounting-bridge consolidate --files entity_a.csv entity_b.xlsx \
    --base-currency USD --rates EUR:closing=1.08,average=1.07 --output consolidated.xlsx

# Inspect how a file is parsed (encoding, delimiter, columns)
accounting-bridge read --file messy_export.csv

Supported formats

Format	Extensions	Extra
Delimited text	.csv .tsv .txt	core
Excel	.xlsx .xls .xlsm	core
PDF tables	.pdf	[pdf]
Open Financial Exchange	.ofx .qfx	[ofx]
QuickBooks Desktop	.iif	core
Scanned images	.png .jpg	[ocr]

Architecture

Library-first. The engine is the product; the app and server are thin layers over it.

accounting_bridge/        core library — pure, typed, no UI/server deps
  models/                 pydantic v2 config + result models (validated at the boundary)
  ingest/                 multi-format readers (encoding/delimiter/header auto-detection)
  engine/                 consolidation, POS, audit, trial balance, intercompany, FX
  processing.py           the typed public API (consolidate / convert_pos / audit_ledger / ...)
cli.py                    one-shot command-line consolidation

A FastAPI server and a dedicated web frontend are layered on the same library (see the redesign in progress in CHANGELOG.md).

Self-hosting

docker compose up           # single-user instance, SQLite, no secrets required

Authentication and billing ship disabled by default; set the documented environment variables to enable a multi-tenant hosted deployment.

Development

uv venv --python 3.12 && source .venv/bin/activate
uv pip install -e ".[all,dev]"
make test        # python -m pytest
make lint        # ruff check + ruff format --check
make typecheck   # mypy --strict

See CONTRIBUTING.md for conventions and SECURITY.md to report vulnerabilities.

License

MIT © Wietze Suijker