SupportSignal

Camada de inteligência de suporte: temas, SLA, risco de reembolso e causa raiz.

Support intelligence layer: topics, SLA, refund risk and root-cause actions.

PT-BR · English · Live Demo · Stack · Architecture · Quick Start · Author

Live Demo · Repo · Portfolio · LinkedIn

Lab / demo notice: frontend-first lab with a synthetic inbox seed. Configurable rules + explainable scores — not an autonomous AI helpdesk, not a full ticketing system, and not production support automation.

---

PT-BR

Visão geral

O SupportSignal classifica mensagens, mede SLA de primeira resposta, pontua risco de reembolso e transforma sintomas de suporte em causas raiz e backlog de ações.

Problema

Suporte vira caixa-preta: volume sobe, mas a operação não vê temas, breaches de SLA, risco de chargeback nem causa de produto.

Para quem

Founders, ops de suporte e analistas que precisam de inteligência sobre o canal — sem trocar o helpdesk no dia 1.

Funcionalidades

• Cockpit de inteligência + classificador de temas (regras configuráveis)
• Dashboard de SLA (primeira resposta, breaches)
• Board de risco de reembolso com drivers explicáveis
• Root-cause explorer, weekly memo e action backlog
• Import/seed de mensagens demo
• Engine client-side (apps/web/lib/engine) + API FastAPI opcional

Escopo e limites (honestos)

• Não é helpdesk completo nem IA que “resolve” atendimento
• Classificação por regras/heurísticas no MVP — revisão humana em risco alto
• Demo pública usa seed sintético no browser
• Sem Gmail/Zendesk/WhatsApp oficiais no MVP

---

English

Overview

SupportSignal classifies messages, measures first-response SLA, scores refund risk and turns support symptoms into root causes and an action backlog.

Problem

Support becomes a black box: volume rises, but ops cannot see topics, SLA breaches, chargeback risk or product root causes.

Who it is for

Founders, support ops and analysts who need intelligence on the channel — without replacing the helpdesk on day one.

Features

• Intelligence cockpit + topic classifier (configurable rules)
• SLA dashboard (first response, breaches)
• Refund-risk board with explainable drivers
• Root-cause explorer, weekly memo and action backlog
• Demo message import/seed
• Client-side engine (apps/web/lib/engine) + optional FastAPI API

Scope and honest limits

• Not a full helpdesk or “AI that resolves support”
• Rule/heuristic classification in the MVP — human review for high risk
• Public demo uses a synthetic browser seed
• No official Gmail/Zendesk/WhatsApp integrations in the MVP

---

Live Demo

Surface	URL
Public lab	https://supportsignal-lab-lake.vercel.app
GitHub	https://github.com/BarujaFe1/SupportSignal

How to try: open the lab → load the synthetic seed → inspect topic classifier + SLA → open refund-risk board → read root-cause explorer + weekly memo + action backlog.

Alternate alias also live: https://supportsignal-lab.vercel.app (README Live Demo follows the GitHub homepage field).

---

Screenshots

Intelligence cockpit	Topic classifier
SLA dashboard	Refund risk board
Root-cause explorer	Weekly memo
Action backlog	Message import

---

Stack

Layer	Technology
Web	Next.js 15, React 19, TypeScript, Recharts, Lucide
Engine (browser)	TypeScript analyzer in apps/web/lib/engine
API (optional)	FastAPI, Pandas, NumPy, pytest

---

Architecture

apps/
  web/
    app/                 Next.js UI
    lib/engine/          analyzer + types (client-side)
    public/data/         support_inbox_demo.json
  api/
    app/services/        analyzer, report, demo_data
assets/

Flow: inbox seed/CSV → normalize → topic rules → SLA flags → refund risk score → root causes → weekly memo + backlog.

---

Quick Start

Prerequisites: Node.js 20+, Python 3.10+ (optional), Git.

Frontend-only (same as Vercel)

cd apps/web
npm install
npm run dev

Windows integrated

.\start.bat

FastAPI

cd apps/api
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
uvicorn app.main:app --reload --port 8000

---

Technical decisions

• Intelligence layer, not helpdesk — keep agents where they already work
• Explainable refund risk with drivers instead of opaque “AI score”
• Client-side engine on Vercel for a reliable lab without hosting Pandas
• Human review for high risk — no autonomous refund decisions

---

Roadmap

• Official connectors (helpdesk / email) with OAuth
• Stronger topic taxonomy versioning
• Multi-inbox workspaces
• Alerting for SLA breach spikes
• Deeper parity tests between web engine and API

---

Author

Felipe Alirio Baruja — data / product / full-stack portfolio.

• Portfolio: https://barujafe.vercel.app/
• GitHub: https://github.com/BarujaFe1
• LinkedIn: https://www.linkedin.com/in/barujafe/

---

License

MIT — see LICENSE.