AI-powered payments and payroll planner that ingests payroll/bill files, validates accounts via RazorpayX sandbox, runs risk checks, and generates structured payment plans with LLM reasoning.
All payments remain in sandbox mode (no real money movement). Human approvers review, approve, and export manifests with full audit trails and 2FA.

• Payroll ingestion — Upload CSV/XLS files or API feeds.
• RazorpayX sandbox integration — Beneficiary/account verification, test payouts, and webhook event handling.
• AI Planner — LLM generates structured payment plans with explanations.
• Risk engine — Deterministic rules flag anomalies (amount > limit, duplicate entries, unverified bank).
• Approvals workflow — Dashboard for batch review, item-level evidence, 2FA for high-risk approvals.
• Audit logging — Immutable logs of all uploads, plans, approvals, and connector calls.
• Export manifests — Downloadable CSV/ZIP for manual execution if required.
• Secure by design — RBAC, masked sensitive fields, sandbox-only in V1.

• Backend: Python 3.11, FastAPI, Alembic, Redis (RQ workers)
• Frontend: React + TypeScript, TailwindCSS (shadcn/ui)
• Database: PostgreSQL + S3 (uploads storage)
• LLM: External LLM API (planner, JSON schema validation)
• Payments Sandbox: RazorpayX Sandbox (test payouts, verification)
• Infra: Docker, GitHub Actions CI/CD

git clone https://github.com/149189/Cashora.git
cd Cashora

.env

DATABASE_URL=postgresql://user:pass@db:5432/payments
REDIS_URL=redis://redis:6379/0
S3_BUCKET=uploads-bucket
RAZORPAYX_API_KEY=rzp_test_xxxxx
RAZORPAYX_API_SECRET=xxxxx
RAZORPAYX_WEBHOOK_SECRET=whsec_xxxxx
LLM_API_KEY=sk-xxxxx

alembic upgrade head

uvicorn app.main:app --reload