Productionization Approach

How to take PULSE as an accelerator and deploy to customers

Accelerator Philosophy

PULSE is designed as a configurable accelerator, not a one-off solution. Core logic remains unchanged across deployments. Only configuration and branding change per customer.

80%

Reusable Core

15%

Configuration

Custom Extensions

Accelerator Components

Core Accelerator

Excel Parsing Engine

Generic XLSX parser with configurable schema mapping

Anomaly Detection Framework

Pluggable detection rules with threshold configuration

Pipeline Builder UI

Visual workflow designer for QC pipelines

Context Search Agent

Hypothesis-driven web search with customizable buckets

Integration Layer

Jira MCP Connector

Pre-built ticket creation with field mapping

Email Service Adapter

Template-based email generation

SSO/Auth Framework

Pluggable authentication providers

Webhook System

Event-driven notifications to external systems

UI Components

Spreadsheet Viewer

Excel-like display with cell selection

Anomaly Dashboard

Summary cards, charts, and drill-down views

Review Workflow UI

Human-in-loop approval components

Configuration Panels

Admin UI for rules and thresholds

Configuration Layers

All customer-specific settings are externalized into configuration files. No code changes required for standard deployments.

Environment Config

API keys, service URLs, feature flags

.env.production.env.staging.env.local

# Customer-specific environment
NEXT_PUBLIC_CUSTOMER_ID=circana
JIRA_PROJECT_KEY=CIRQC
JIRA_HOST=circana.atlassian.net
EMAIL_FROM=qcbot@circana.com
ANOMALY_THRESHOLD_VARIANCE=0.02

Branding Config

Colors, logos, fonts, terminology

branding/config.jsontailwind.config.ts

{
  "name": "PULSE",
  "primaryColor": "#5B2E91",
  "logo": "/circana-logo.svg",
  "terminology": {
    "anomaly": "variance",
    "ticket": "case"
  }
}

Detection Rules

Anomaly types, thresholds, categorization

config/rules.jsonconfig/thresholds.json

{
  "rules": [
    {
      "id": "store_dropout",
      "threshold": 0.05,
      "severity": "critical",
      "action": "create_ticket"
    }
  ]
}

Integration Config

Jira fields, email templates, webhooks

config/integrations.json

{
  "jira": {
    "project": "CIRQC",
    "issueType": "Bug",
    "customFields": {
      "dataSource": "customfield_10001"
    }
  }
}

Deployment Options

Vercel (SaaS)

Pros

Zero infrastructure management
Auto-scaling
Global CDN
Preview deployments

Considerations

- Less control
- Egress costs at scale

Best for: POC, pilot programs, small teams

AWS (Self-Hosted)

Pros

Full control
VPC integration
Compliance ready
Custom scaling

Considerations

- More ops overhead
- Longer setup

Best for: Enterprise, regulated industries

Customer Cloud

Pros

Data stays in customer env
SSO integration
Audit compliance

Considerations

- Deployment complexity
- Support overhead

Best for: Highly regulated, data-sensitive clients

Customer Onboarding Process

Standard onboarding takes 10-15 working days from kickoff to go-live.

Discovery

1-2 days

Gather requirements

Map data sources

Define anomaly types

Identify integrations

Configuration

2-3 days

Set up environment

Configure branding

Define detection rules

Map Jira fields

Data Mapping

2-4 days

Analyze Excel structure

Create schema mappings

Define hierarchies

Set up test data

Integration

2-3 days

Connect Jira/ServiceNow

Configure email service

Set up SSO

Test webhooks

Testing

2-3 days

End-to-end testing

UAT with pilot users

Performance validation

Security review

Go-Live

1 day

Production deployment

Monitoring setup

Documentation handoff

Support transition

Repository Structure

Core Accelerator (Shared)

qc-bot-accelerator/
├── src/
│   ├── app/              # Next.js pages
│   ├── components/       # Reusable UI components
│   │   ├── demo/        # Pipeline, anomaly UI
│   │   ├── layout/      # Header, nav, footer
│   │   └── ui/          # Shared components
│   ├── lib/
│   │   ├── agents/      # Detection, context agents
│   │   ├── pipeline/    # Pipeline builder core
│   │   └── utils/       # Excel parsing, helpers
│   └── data/            # Schema definitions
├── config/
│   └── defaults/        # Default configurations
└── package.json

Customer Instance

qc-bot-circana/
├── config/
│   ├── branding.json     # Colors, logo, terms
│   ├── rules.json        # Detection rules
│   ├── integrations.json # Jira, email config
│   └── thresholds.json   # Per-metric thresholds
├── public/
│   └── logo.svg          # Customer logo
├── .env.production       # API keys, URLs
└── package.json          # Extends accelerator

Version Management

Customer instances reference specific accelerator versions via npm/git. Updates to core are pulled in via dependency updates, allowing customers to stay on stable versions while benefiting from improvements.

CI/CD Pipeline

Push to main

Run Tests

Build

Deploy Staging

E2E Tests

Deploy Prod

GitHub Actions (Recommended)

Pre-built workflows for test, build, deploy. Works with Vercel, AWS, Azure.

Environment Promotion

Dev → Staging → Production with approval gates and rollback support.

Quick Start: New Customer Deployment

# 1. Clone accelerator template
npx create-qc-bot my-customer-qcbot

# 2. Configure branding
cp branding/circana.json config/branding.json
# Edit config/branding.json with customer colors/logo

# 3. Set up environment
cp .env.example .env.production
# Add customer-specific API keys and URLs

# 4. Configure detection rules
# Edit config/rules.json with customer thresholds

# 5. Test locally
npm run dev

# 6. Deploy
npm run deploy:production

With proper configuration externalization, a new customer deployment can be completed in 2-3 weeks including testing and training.