Rowshni

Shedding light on your ledger

AI-powered reconciliation for month-end close. Upload files, map fields, review transformed data, and run agents.

1. Upload2. Map3. Preview4. Reconcile
User Guide
Loading...
?
A

Anonymous mode

Use all reconciliation features without signing in. Data persists in this browser.

Workflow Progress

Follow the sequence below to move from raw files to final reconciliation output.

1

Upload Files

Upload GL, subledger, and transaction data

Pending
2

Map Columns

Map your CSV columns to canonical fields

Pending
3

Preview Data

Review transformed data before running

Pending
4

Run Agents

Execute reconciliation with AI agents

Pending

Step 1

Upload Files

Upload your GL balance, subledger balance, and transaction files (CSV/TSV/TXT format).

Max 20MB

GL Trial Balance*

CSV/TSV/TXT - Period auto-detected from filename

Subledger Balance (AP/AR Aging)*

CSV/TSV/TXT - Period auto-detected from filename

Transaction Detail

Optional: Transaction-level detail for variance investigation (CSV/TSV/TXT)

Supporting Files

Optional: Upload supporting CSV/TSV/TXT files (multiple files allowed).

Column Mapping

Map your columns

Map source columns to canonical fields. You can auto-suggest and then adjust only the mismatches.

No file uploaded for gl balance

Upload a file in the Upload Workspace above

Complete required mappings to continue

Mappings are saved locally in your browser

Data Preview

Upload and map your files to see a preview

AI Reconciliation

Multi-agent console

Run validation, analysis, investigation, and report generation using schema v0.1.0.

4 agents

How threshold works

Variances above this amount are flagged as material and need follow-up. Lower values increase sensitivity.

No data ready

Upload files and apply mappings before running agents.

Sample data

Download test scenario files

Realistic reconciliation scenarios for balanced cases, variances, cutoff timing differences, and multi-period analysis.

About Rowshni

Rowshni means "light". We illuminate your reconciliations with AI-powered insights, bringing clarity to complex financial data.

Tech Stack
AI Agents: 4Version: 0.1.0Fields: 11
  • Next.js 16 web application
  • Fastify orchestrator service (OpenAI, Claude, Gemini modes)
  • Spec-Kit contracts for data validation
  • Gemini 2.0 Flash (default free-tier 4-agent pipeline)
  • Claude skills subagents for mapping and variance investigation
  • Anonymous mode with browser localStorage
What are Spec-Kit contracts?

Spec-Kit contracts are structured JSON schemas that define the "source of truth" for data validation and system behavior.

They define:

  • Data Models: Canonical schemas likecanonical_balance(account_code, period, amount, currency)
  • Interfaces: Exact inputs/outputs for the reconciliation tool that AI agents consume
  • Workflows: Expected process flows for upload and reconciliation

Your uploaded CSVs are transformed to match these canonical schemas. This ensures consistency across different accounting systems and allows AI agents to reliably process your data.

Contract version 0.1.0 | 11 fields total

AI Runtime

Default: Gemini 2.0 Flash free tier. Optional OpenAI agents and Claude skills run when configured.

View README ->View Spec ->
Data Model

Agentic reconciliation workspace that ingests GL/subledger data, maps heterogeneous headers, and coordinates AI subagents. Reports include Organization (if set), Reporting Period, and Report Generated On.

Balance Fields

  • account coderequired
  • periodrequired
  • currencyoptional
  • amountrequired

Transaction Fields

  • account coderequired
  • booked atrequired
  • debitoptional
  • creditoptional
  • amountoptional
  • narrativeoptional
  • source periodoptional
AI Agent Pipeline

1. Data Validation Agent

Validates data quality and detects issues

2. Reconciliation Analyst

Analyzes variances and identifies patterns

3. Variance Investigator

Investigates material variances and root causes

4. Report Generator

Creates audit-ready documentation