FinTransFlow

Production-style monorepo for FinTransFlow, a dual-core platform for bilingual annual-report translation (Chinese ↔ English financial disclosure).

Repository layout

Path	Role
backend/	FastAPI — HTTP API and orchestration. Split into Operations Core and Translation Intelligence Core packages (see below).
frontend/	Next.js — operator dashboard and review UX (to be wired to the API).
data_pipeline/	Batch / CLI pipelines — parsing, alignment, TM build, glossary prep, preprocessing.
ml/	QE / routing — feature extraction, model training, and inference entrypoints.
docs/	Product and technical specifications.

Dual-core architecture

1. Operations Core (backend/app/operations_core/) — client intake through delivery: quotes, orders, projects, payments, tracking, archival concerns (API surface under backend/app/api/v1/operations.py).
2. Translation Intelligence Core (backend/app/translation_intel_core/) — document intelligence: parsing, segmentation, TM/glossary, MT/revision hooks, QA, QE-based routing, review queue, assembly (API surface under backend/app/api/v1/translation_intel.py).

Cross-core contracts (events, DTOs, job IDs) should live in shared modules as the implementation grows.

Environment configuration

Copy the example env files per package:

• backend/.env.example → backend/.env
• frontend/.env.example → frontend/.env.local
• data_pipeline/.env.example → data_pipeline/.env
• ml/.env.example → ml/.env

Local development (scaffold)

Backend and frontend are intentionally minimal until modules are implemented.

# Backend (from repo root)
cd backend && python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
uvicorn app.main:app --reload --port 8000

# Frontend
cd frontend && npm install && npm run dev

Pipeline and ML packages are installable in editable mode; see each directory’s pyproject.toml and TODOs.

Documentation

• docs/PROJECT_SCOPE.md — scope, constraints, and deliverables
• docs/ARCHITECTURE.md — dual-core boundaries and package map (scaffold)