Uno

Business Glossary
Future Outlook
Quick View
Transaction ledger
Agent Warren

Inspiration

Small and medium-sized merchants running their businesses on PayPal face a common operational problem: PayPal provides world-class payment infrastructure – Subscriptions, Invoicing, Disputes, and Reporting yet non-technical business owners find themselves toggling between fragmented admin tabs, manually tracking debtors in spreadsheets, and calculating cash flow projections.

I asked myself: What if merchants didn't have to navigate menus or build complex BI dashboards at all? What if they could converse with their financial ecosystem in plain English, and have someone observe their data and act ?

This question inspired Uno and agent Warren.

What it does

Uno lives on 3 pillars:

Observe: Live Telemetry & Briefing Automatically probes PayPal Subscriptions, Invoicing, Disputes, and Reporting to construct a real-time operating profile without manual data entry. There are charts for visual summary. Settlement Ledger (AG Grid): Real-time interactive table tracking customer transactions with live 1-click reminder triggers. The Business Glossary is a built-in reference with category filters explaining formulas, benchmarks, and key metrics.
Act: Agent Warren Generates and registers official PayPal invoices from natural language ("Bill Maya Lin $340 for Enterprise Integration") via the PayPal Invoicing API. 1-Click Overdue Reminders: Identifies overdue accounts and triggers live payment reminder emails via PayPal's invoicing API. On-the-fly generation and rendering of interactive charts inside the chat window.
Predict: Runway Telemetry Blends historical invoice velocity with recurring subscription run-rates to project a 90-day trajectory. Evaluates churn risk, exposed MRR, specific risk drivers (e.g., card declines), and concrete recovery actions across all active subscription tiers.

How I built it

Frontend: Built with Next.js and TypeScript, simple to use and intuitive UI. The visuals are powered by AG Charts API and AG Grid. Backend: FastAPI (Python 3.10+) microservice managing OAuth 2.0 token refreshes, asynchronous PayPal API calls, and telemetry pipelines. AI Engine: Gemini 2.5 Flash and Jev integrated through LangChain, LangGraph, and paypal-agent-toolkit. Gemini handles intent classification, entity extraction (debtors, amounts, line items), and structured JSON generation for AG Charts. PayPal Sandbox Seeder: A customized automated seeder provisioning 70+ invoices to stress-test real-world volume and dispute states.

Challenges I ran into

Bridging unstructured conversational answers with typed AG Chart options on the fly without breaking client-side rendering.
Ensuring the AI model never hallucinated invoice IDs or amounts when executing operations. I implemented strict Pydantic schema validation and entity grounding against live ledger records before any write requests were dispatched.

Accomplishments that I am proud of

i believe this application (product) actually has the potential solving a real problem for merchants. I did not build some agent that could purchase something through PayPal or some different kind of store, instead built a limited functions but necessary application.

What I learned

Learning how to design agentic workflows where AI does not just summarize information, but takes verified, low-friction actions.

What's next for Uno

Sending morning briefings and urgent overdue warnings directly to merchants on Whatsapp.
Actually testing this in the real-world with merchants, getting feedback and improving the experience.