FinLang

FinLang explained. Human-readable rules. 27K+ rows/sec. Deterministic output. Original data unchanged. FinLang doesn't guess. It applies logic. 🎬 👇 youtu.be/Xbc_jW2r18k #FinTech #FinLang

Your auditor can't read Japanese. Your bank export might. FinLang handles Unicode headers, Unicode wildcards, 7 countries, 5 currencies, 4 writing systems. Your auditor reads the rules. FinLang reads everything else. #FinLang #i18n

One known limitation: Indian lakhs notation (2,45,000). Non-standard grouping, not yet supported. On the roadmap. If you deal with lakhs, DM me a sample row — I'll design the flag around real data. #FinLang #FinTech

Nine amount formats. Debit/credit columns. Accounting parens. Trailing minus. CR/DR suffixes. Currency symbols. EU locale. All normalised automatically. No pre-processing needed. #FinLang #DataQuality

2,836 lines. 6 source files. No TODOs. No FIXMEs. No debug code. Solo founder, enterprise codebase. Written for the person who reads it next. #FinLang #CodeQuality

FinLang v0.7.7 — built-in --verify. SHA-256 fingerprints on every row. Zero mismatches = zero corruption. Plus CR/DR fix, fuzzy suggest improvements, 118 tests / 9 gates. pip install --upgrade finlang #FinLang #FinTech

13 years of notebooks. One product that finally works. 5,000+ downloads. Still building evenings and weekends. If you're on a similar path, I'd like to hear your story. #FounderLife #FinTech

The detection regex had \b. Its sibling stripping regex never did. They were inconsistent for six months. Bug fix → correctness improvement → measurable performance gain as a side effect. The kind of result you only find when you actually measure what you ship.

Finance teams: I can build your FinLang map file and starter rulepack with you. You bring the use case. I bring the engine. Real data, real rules. DM me. #FinLang #DesignPartner

Solo founder. 20+ docs. Architecture, CLI reference, i18n, field specs, test suite docs. No technical writer. Documentation isn't overhead — it's due diligence preparation. #FinLang github.com/FinLang-Ltd/fi…

Same input. Same rules. Windows and Linux. Identical output. Determinism isn't a marketing word. It's a test result. #FinLang #QualityEngineering

EU AI Act rolling into application. Explainability. Auditability. Human oversight. If your categorisation engine can't explain its decisions, that's a conversation worth having now. #EUAIAct #FinTech

FinLang v0.7.7 — built-in integrity verification. --verify computes SHA-256 fingerprints on every row, comparing input to output. Zero mismatches = zero corruption. Plus CR/DR fix, improved fuzzy suggest, 118 tests across 9 gates. pip install --upgrade finlang #FinLang #FinTech

Discover uncategorised patterns. Suggest draft rules. Review, merge, re-run. Coverage compounds over time. The engine learns nothing — your ruleset grows. #FinLang

7 countries. 5 currencies. 4 writing systems. £ $ € ¥ ₹. English, German, French, Italian, Japanese, Hindi. One engine. Same deterministic output. #FinLang #i18n

Probabilistic categorisation is probably right. Most of the time. FinLang is deterministic. Same inputs. Same rules. Same output. Every time. No confidence scores. No retraining. #FinLang #FinTec

Building solo while working full-time. 5,000+ PyPI downloads and counting. Hardest part: the gap between what the product can do and what the world knows. What's yours? #StartupLife #FounderLif

Most auditors can't read Python or regex. They shouldn't have to. counterparty ~ "SHELL" → category = "Energy & Commodities". Hand them the .fin file. #FinLang #Compliance