Pulse360 — 360° Performance Review

Pulse360 is a full-stack internal 360° performance review platform built for semi-annual/quarterly/monthly cycles. Built by me and the IBM Bob team at the Pulse Technologies Hackathon V1 — AI App Build IDE Challenge. Testing AI App Build IDE tools like IBM Bob, Codex, Cursor, Claude Code.

---

Table of Contents

1. Overview
2. Tech Stack
3. Team
4. Prerequisites
5. Getting Started
6. Environment Variables
7. Database
8. Running the App
9. Test Accounts
10. Project Structure
11. API Reference
12. Review Cycle Lifecycle
13. Features Implemented
14. Human-in-the-Loop Log
15. IBM Bob — Agentic AI Platform Notes
16. Useful Commands

---

Overview

Pulse360 replaces the original eWay 360° Review name with a modern product identity. It allows Pulse Technologies employees to conduct semi-annual peer performance reviews aligned to the eWay competency framework — covering 6 criteria (MASTERFUL, EXCELLENCE, EXECUTION, COMMITMENT, CONTRIBUTION, ADDITIONAL) across 28 questions.

Four roles — System Administrator, HR Administrator, Line Manager, and Employee — each see a tailored dashboard and workflow through an 8-phase review cycle: Draft → Nominate → Approve → Review → Calculation → Consultation → Accept → Closed. The System Administrator is a separate platform-operations persona for adoption analytics, audit metadata, AI usage, and operational health monitoring without exposing sensitive review scores or comments.

All features were built using IBM Bob, an enterprise agentic AI platform(Agentic SDLC Development partner), in a single hackathon day. Bob acted as Developer, Architect, QA Reviewer, Data Engineer, and Security Reviewer throughout the build. My role was to guide bob on the companies architecture, designs shift and other capabilities and feature that would make the product very useful, provide business value and create a differentiator with IBM BOB for Agentic SDLC Development by proving that it is possible to build a fully AI integrated system in a day with minimal ease saying developers the time on development and prototyping.

---

Tech Stack

Layer	Technology	Version
Framework	Next.js (App Router)	16.2.12
Language	TypeScript	5.x
Database	PostgreSQL	16
Container	Podman	5.7.1
ORM	Prisma	7.9.1
Auth	NextAuth.js	4.24.15
UI	Tailwind CSS	4.x
AI Text Generation	OpenAI SDK (GPT-4o)	7.4.0
Agentic AI Platform	IBM Bob	2
MCP Tooling	Model Context Protocol SDK	1.11.4
Runtime	Node.js	22.x

Note: This project uses Podman (not Docker). The docker-compose.yml in the root is kept for reference only. The database is started with podman run — see Database. My Docker env was running on a different machine for a BPMN agentic system build for an e-commerce client's order-processing with agentic capabilities.(projet to be released soon to the public)

---

Team

Name	Role
Romeo Ndlovu	Data and AI Engineer / Platform Engineer
IBM Bob	Agentic AI Developer

Want to test it out for your companies developers:

• Link to 30 day free trial: https://bob.ibm.com/

• IDC research Link (May 05, 2026): https://www-api.ibm.com/adobe/assets/urn:aaid:aem:f72132ce-a92b-43ea-8b57-f8127b773535/original/as/idc-paper-on-ibm-bob-for-agentic-sdlc-development.pdf

---

Prerequisites

Requirement	Version	Notes
Node.js	v20+	Required to run the Next.js app
Podman	v5+	Used to run the PostgreSQL container
npm	v9+	Bundled with Node.js

---

Getting Started

1. Clone the repository

git clone https://bitbucket.org/ovmobile/hackathon-v1-ibm.git
cd hackathon-v2-ibm

2. Start the PostgreSQL database

# Start the container (first time — creates and seeds the DB)
podman run -d `
  --name pulse360_db `
  -e POSTGRES_USER=pulse360_user `
  -e POSTGRES_PASSWORD=password123 `
  -e POSTGRES_DB=pulse360 `
  -p 5432:5432 `
  postgres:16-alpine

# Apply all init scripts in order
Get-ChildItem database\init\*.sql | Sort-Object Name | ForEach-Object {
  Write-Host "Applying $($_.Name)..."
  Get-Content $_.FullName | podman exec -i pulse360_db psql -U pulse360_user -d pulse360
}

If the container already exists from a previous run:

podman start pulse360_db

3. Install app dependencies

cd pulse360
npm install

4. Configure environment variables

Create pulse360/.env with the following (never committed to git):

DATABASE_URL="postgresql://pulse360_user:password123@localhost:5432/pulse360"
NEXTAUTH_SECRET="pulse360-dev-secret-change-in-production-32chars"
NEXTAUTH_URL="http://localhost:3000"
OPENAI_API_KEY="sk-..."    # Optional — app falls back to stub responses if not set

5. Generate Prisma client

cd pulse360
npx prisma generate

6. Start the development server

cd pulse360
npm run dev

App is available at http://localhost:3000 locally

Deployed App is on: https://pulse360-gkt8.onrender.com

---

Environment Variables

Variable	Required	Description
DATABASE_URL	✅	PostgreSQL connection string
NEXTAUTH_SECRET	✅	Random string for JWT signing (min 32 chars)
NEXTAUTH_URL	✅	Base URL of the app (http://localhost:3000 locally)
OPENAI_API_KEY	⚠️ Optional	Enables live AI Comment Suggestions, Theme Summary, Improvement Plan, and PDF Report narrative. App falls back to realistic stub responses if not set.

---

Database

Init scripts (database/init/)

Scripts are applied in order on first container start:

File	Purpose
01_extensions.sql	Enables pgcrypto and citext extensions
02_enums.sql	All PostgreSQL ENUM types
03_schema.sql	11 tables, triggers, and constraints
04_indexes.sql	Performance indexes
05_seed_criteria.sql	6 eWay criteria + 28 questions
06_seed_departments.sql	10 departments
07_seed_employees.sql	57 synthetic employees with manager relationships
08_add_password_hash.sql	bcrypt password hashes for all 57 employees

Data Model (17 tables)

Table	Purpose
department	10 org departments
employee	All users — System Admin, HR Admin, Line Manager, Employee
review_cycle	Semi-annual review cycles (8-phase lifecycle)
pulse_criterion	The 6 eWay competency pillars
pulse_question	28 individual sub-questions (RATING / TEXT / BOOLEAN)
cycle_criteria	Junction: which criteria are active per cycle
nomination	Peer nominations with manager auto-lock
review	Submitted reviews per reviewer per subject
review_rating	Individual question answers (score or text)
review_result	Computed aggregates written at CALCULATION phase
audit_log	Append-only raw governance event log
auth_event	Login success/failure projection for adoption and access monitoring
profile_event	Profile update projection showing which fields changed without exposing old/new values
ai_usage_event	AI feature usage, status, model, and token metadata
ai_hitl_decision	Human-in-the-loop accept/edit/discard decisions for AI-generated content
nomination_event	Nomination lifecycle events with employee/reviewer and department snapshots
review_event	Review draft/submission metadata without review comments or scores

The employee table now also carries HR-managed workforce dimensions: job grade, employment type, conversion-hire status, gender, and ethnicity. System Admin analytics read from dedicated event tables and aggregated workforce dimensions. Ethnicity is intentionally kept for HR demographic reporting and is not displayed on the System Admin dashboard.

Verify the database

# Connect with psql
podman exec -it pulse360_db psql -U pulse360_user -d pulse360

# Check all tables
\dt

# Check seeded employees
SELECT first_name, last_name, role FROM employee LIMIT 10;

# Check criteria and questions
SELECT c.name AS criterion, COUNT(q.id) AS questions
FROM pulse_criterion c
JOIN pulse_question q ON q.criterion_id = c.id
GROUP BY c.name, c.sort_order
ORDER BY c.sort_order;

---

Running the App

# Development (with hot reload)
cd pulse360
npm run dev

# Production build check
cd pulse360
npx next build

# Production server
cd pulse360
npm start

---

Test Accounts

All passwords are bcrypt-hashed in the database — never stored in plain text.

Primary Test Accounts (Delivery & Technology)

Role	Email	Password
HR Administrator	lerato.mkhize@techcorp.co.za	Pulse360!Admin
Line Manager	hlanganani.oosthuiz@techcorp.co.za	Pulse360!Manager
Employee	mpho.zulu@techcorp.co.za	Pulse360!Employee

Secondary Test Accounts (Marketing — different department)

Role	Email	Password
Line Manager	zanele.mohamed@techcorp.co.za	Pulse360!Manager
Employee	dikeledi.mkhize@techcorp.co.za	Pulse360!Employee

Zanele and Dikeledi are in the Marketing department. They can only nominate peers within Marketing (department-scoped peer selection). Zanele, as Dikeledi's line manager, appears as her mandatory auto-locked reviewer.

---

Project Structure

hackathon-v2-ibm/
├── README.md
├── .env.example
├── docker-compose.yml              # Reference only — use Podman
├── .bob/
│   └── mcp.json                    # MCP server registration (pulse360-mcp)
├── database/
│   └── init/                       # 8 SQL init scripts (applied in order)
│       ├── 01_extensions.sql
│       ├── 02_enums.sql
│       ├── 03_schema.sql
│       ├── 04_indexes.sql
│       ├── 05_seed_criteria.sql
│       ├── 06_seed_departments.sql
│       ├── 07_seed_employees.sql
│       └── 08_add_password_hash.sql
└── pulse360/                        # Next.js application
    ├── .env                         # Local env (not committed)
    ├── next.config.ts
    ├── prisma.config.ts             # Prisma v7 datasource config
    ├── prisma/
    │   └── schema.prisma            # 11 models, 7 enums
    ├── mcp-server/                  # ✨ Custom MCP server (TypeScript)
    │   ├── src/index.ts             # Two tools: write_csv, generate_pdf_report
    │   ├── build/index.js           # Compiled output (registered in .bob/mcp.json)
    │   └── package.json
    └── src/
        ├── proxy.ts                 # Next.js 16 middleware (RBAC route guards)
        ├── components/
        │   ├── Sidebar.tsx          # Role-aware navigation
        │   ├── CycleList.tsx        # Cycle phase stepper (client component)
        │   ├── AISummaryPanel.tsx   # AI Feedback Theme Summary panel
        │   └── SelfAssessmentPanel.tsx  # ✨ Self-assessment + gap analysis + improvement plan
        ├── lib/
        │   ├── auth.ts              # NextAuth v4 credentials provider
        │   ├── auth.types.ts        # Session type augmentation
        │   └── prisma.ts            # PrismaClient singleton with pg adapter
        └── app/
            ├── layout.tsx
            ├── page.tsx             # Root redirect → /dashboard or /login
            ├── globals.css
            ├── login/
            ├── api/
            │   ├── auth/[...nextauth]/
            │   ├── ai/
            │   │   ├── suggest-comments/     # POST — AI comment suggestions (GPT-4o)
            │   │   ├── theme-summary/        # POST — AI feedback theme analysis (GPT-4o)
            │   │   └── improvement-plan/     # ✨ POST — AI self-improvement plan + gap analysis
            │   ├── analytics/
            │   │   ├── heatmap/              # GET — dept × criterion heatmap + trends + pie data
            │   │   └── report/               # ✨ POST — AI-generated PDF report HTML
            │   ├── mcp/
            │   │   ├── write-csv/            # ✨ POST — MCP tool proxy: write improvement plan CSV
            │   │   └── generate-report/      # ✨ POST — MCP tool proxy: write approved report HTML
            │   ├── approvals/[id]/approve/   # POST
            │   ├── approvals/[id]/reject/    # POST
            │   ├── approvals/bulk-approve/   # POST
            │   ├── criteria/                 # GET (active), POST (create new)
            │   ├── criteria/full/            # GET (all + questions)
            │   ├── criteria/[id]/            # PUT (edit / toggle active)
            │   ├── cycles/                   # GET, POST
            │   ├── cycles/list/              # GET (lightweight list)
            │   ├── cycles/[id]/advance/      # POST (advance phase + auto-calculate scores)
            │   ├── departments/              # GET
            │   ├── employees/                # GET, POST
            │   ├── employees/list/           # GET (dept-scoped for non-HR roles) ✨ Updated
            │   ├── employees/managers/       # GET
            │   ├── employees/[id]/           # GET, PUT
            │   ├── nominations/              # GET, POST (dept-scoped validation) ✨ Updated
            │   ├── nominations/cycle/        # GET (effective min/max per pool size) ✨ Updated
            │   ├── nominations/submit/       # POST (effective min per pool size) ✨ Updated
            │   ├── nominations/[id]/         # DELETE
            │   ├── reviews/                  # POST (save/submit)
            │   └── reviews/draft/            # GET (load draft)
            └── (app)/                        # Authenticated app shell
                ├── layout.tsx               # Sidebar + session wrapper
                ├── analytics/               # ✨ Rebuilt: cycle selector, pie chart, trend bars, heatmap, AI report
                ├── approvals/               # HR Admin + Manager — nomination approvals + bulk approve
                ├── criteria/                # HR Admin — view/edit/add/deactivate criteria
                ├── cycles/                  # HR Admin — list cycles + advance phases
                ├── cycles/new/              # HR Admin — create new cycle
                ├── dashboard/               # All roles — role-specific dashboards
                ├── employees/               # HR Admin — employee list
                ├── employees/[id]/          # HR Admin — add/edit employee
                ├── manager/results/         # Manager — team results + AI summary
                ├── my-results/              # ✨ Employee — own results + AI summary + self-assessment
                ├── nominations/             # Employee + Manager — nominate peers (dept-scoped)
                ├── results/                 # HR Admin — all employee results
                ├── reviews/                 # All roles — pending review list
                └── reviews/[employeeId]/    # All roles — 28-question form + AI suggestions

---

API Reference

Authentication

Method	Endpoint	Description
POST	/api/auth/callback/credentials	NextAuth email + password login

Cycles

Method	Endpoint	Role	Description
GET	/api/cycles	HR Admin	List all cycles
POST	/api/cycles	HR Admin	Create new cycle
GET	/api/cycles/list	HR Admin	Lightweight list for client refresh
POST	/api/cycles/[id]/advance	HR Admin	Advance to next phase; auto-calculates scores at REVIEW→CALCULATION

Nominations

Method	Endpoint	Role	Description
GET	/api/nominations	All	Get my nominations for current cycle
POST	/api/nominations	Employee	Add a nomination (dept-scoped validation)
DELETE	/api/nominations/[id]	Employee	Remove a nomination
GET	/api/nominations/cycle	All	Returns cycle with effectiveMinNominees / effectiveMaxNominees
POST	/api/nominations/submit	Employee	Submit all draft nominations (effective min enforced)

Approvals

Method	Endpoint	Role	Description
GET	/api/approvals	Manager / HR	List pending nominations scoped to role
POST	/api/approvals/[id]/approve	Manager / HR	Approve a nomination
POST	/api/approvals/[id]/reject	Manager / HR	Reject a nomination
POST	/api/approvals/bulk-approve	Manager / HR	Approve all pending nominations

Reviews

Method	Endpoint	Role	Description
POST	/api/reviews	All	Save draft or submit a review
GET	/api/reviews/draft	All	Load existing draft review

Employees

Method	Endpoint	Role	Description
GET	/api/employees	HR Admin	List all employees (paginated, filterable)
POST	/api/employees	HR Admin	Create new employee
GET	/api/employees/[id]	All	Get employee by ID
PUT	/api/employees/[id]	HR Admin	Update employee
GET	/api/employees/list	All	Dept-scoped list for nomination peer search
GET	/api/employees/managers	HR Admin	List employees who are managers

Criteria

Method	Endpoint	Role	Description
GET	/api/criteria	All	List active criteria (for review form)
POST	/api/criteria	HR Admin	Create new criterion
GET	/api/criteria/full	HR Admin	All criteria + questions (for management page)
PUT	/api/criteria/[id]	HR Admin	Edit name/description or toggle active

Analytics & AI

Method	Endpoint	Role	Description
GET	/api/analytics/heatmap?cycleId=	HR Admin	Dept × criterion scores + allCycles, deptOveralls, trends
POST	/api/analytics/report	HR Admin	Generate AI narrative HTML report (GPT-4o or stub)
POST	/api/ai/suggest-comments	All	Generate AI draft feedback comments from ratings (GPT-4o)
POST	/api/ai/theme-summary	All	Generate AI theme summary from all peer comments (GPT-4o)
POST	/api/ai/improvement-plan	All	Generate AI self-improvement plan with gap analysis (GPT-4o)

MCP Tool Proxies

Method	Endpoint	Role	Description
POST	/api/mcp/write-csv	All (HITL)	Write approved improvement plan to CSV (called post-approval only)
POST	/api/mcp/generate-report	HR Admin (HITL)	Write approved HTML report to disk (called post-approval only)

---

Review Cycle Lifecycle

Each review cycle advances linearly through 8 phases. HR Administrator controls all phase transitions.

DRAFT → NOMINATE → APPROVE → REVIEW → CALCULATION → CONSULTATION → ACCEPT → CLOSED

Phase	Who Acts	What Happens
DRAFT	HR Admin	Creates cycle with name, dates, criteria
NOMINATE	Employee	Selects inbound/outbound peers; manager auto-locked; dept-scoped pool
APPROVE	Manager / HR	Approves or rejects individual nominations; bulk approve available
REVIEW	Employee	Rates each person on 28 questions (1–5) + 2 comments; AI suggestions available
CALCULATION	System	Auto-calculates per-criterion and overall averages (2 d.p.)
CONSULTATION	Manager / HR	Results visible to manager and HR; employee sees locked message
ACCEPT	Employee	Employee views own results, AI summary, and self-assessment gap analysis
CLOSED	HR Admin	Results become read-only historical record; visible in analytics history

---

Features Implemented

✅ MVP 1 — All Complete

Category	Features
Auth & RBAC	Email + bcrypt login, JWT sessions, role-based route guards, 4 roles
Employee Management	Add, edit, deactivate, set manager, 57 seeded employees, 10 departments
Review Cycles	Create, 8-phase advance, phase stepper UI, one active cycle enforced
Nominations	Inbound/outbound peer selection, manager auto-locked, draft + submit, min/max limits
Approvals	Manager approves/rejects per nominee; HR Admin bulk approve; scoped by role
Reviews	28-question form (25 RATING + 2 TEXT + 1 BOOLEAN), draft save, submit lock
Score Calculation	Auto-triggers at REVIEW→CALCULATION; per-criterion + overall averages to 2 d.p.
Results	Phase-gated: HR+Manager from CONSULTATION, Employee from ACCEPT only
Dashboards	Role-specific dashboards with progress rings, action banners, team rosters

✅ MVP 2 — Completed (10 of 15 features)

Feature	Status	Notes
View / Add / Edit / Deactivate eWay criteria	✅ Done	Full CRUD with historical preservation
AI Comment Suggestions	✅ Done	GPT-4o generates "doing well" + "improve" draft; reviewer Accepts / Edits / Discards → live HITL
AI Feedback Theme Summary	✅ Done	GPT-4o surfaces top 3 strengths, top 3 growth areas, overall sentiment
Department Heatmap	✅ Done	Colour-coded grid (red→amber→green), org averages, top/bottom panels
PDF Report Generation	✅ Done	AI generates narrative HTML report; human reviews + edits; MCP tool writes to disk post-approval
Self-Assessment + Gap Analysis	✅ Done	Employee self-rates with sliders; AI computes gap vs peer score; labels Overrating/Underrating/Aligned
Historical Trend Comparison	✅ Done	SVG grouped bar chart across all closed cycles; per-criterion org averages compared
Bulk employee import (XLSX)	⬜ Not built	Add Employee UI exists for one-at-a-time entry
In-App Notification Bell	✅ Done	Bell icon in top bar; role-scoped notifications on every phase transition; red unread badge; mark-read per item or bulk; 30 s auto-poll
Mobile Responsive UI	✅ Done	Tested and confirmed — layout adjusts across all screen sizes
Calibration View	⬜ Not built	Department heatmap partially fulfils this requirement
System Admin Audit Trail & Platform Dashboard	✅ Done	Custom SYSTEM_ADMIN role and UI for platform adoption, audit metadata, AI usage, HITL decisions, approval health, and anonymized behavioral analytics

✨ Custom / Beyond-Spec Features

Feature	Description
AI Self-Improvement Plan	After gap analysis, AI generates a personalised weekly action / monthly goal / success metric per criterion. Human must approve before export.
MCP Server (pulse360-mcp)	Custom TypeScript MCP server with write_csv and generate_pdf_report tools, registered in Bob's MCP panel. Invoked only post human approval.
Department-Scoped Nominations	Peer selection restricted to same department for non-HR roles — enforced both in the employee list API and nomination POST validation.
Small-Department Adaptive Limits	effectiveMinNominees and effectiveMaxNominees auto-calculated from pool size, so teams of 2 can still complete the nomination workflow.
Cycle History Selector	Analytics dashboard lets HR choose any historical closed cycle to view, compare, and report on.
Pie Chart Visualisation	Pure SVG (no external charting lib) showing each department's share of the org-wide average score.
Bulk Approve Nominations	HR Admin can approve all pending nominations with one click (per employee or all at once).
CSV Export via MCP	Approved improvement plan written as Excel-compatible CSV via the write_csv MCP tool — human must click Approve first.
System Admin Platform Analytics	Dedicated System Admin experience for adoption, login activity, nomination/review completion, approval bottlenecks, anonymized behavioral network patterns, AI token metadata, and AI accept/edit/discard decisions.
Audit Trail Events	Captures key operational events such as logins, profile updates, nomination actions, approval/rejection decisions, review submissions, AI generation calls, and human-in-the-loop AI decisions.

---

Human-in-the-Loop Log

The hackathon rules require minimum 3 HITL interventions. We documented 10:

Development-Time HITL (Bob as Developer — Human as Reviewer)

#	Type	Description	Outcome
1	Architecture Decision	Bob suggested Office 365 SSO (OAuth). Human overrode to email + bcrypt — simpler, no external dependencies, demo-safe	Implemented email + bcrypt auth
2	Security Review	Bob generated fake bcrypt hashes in seed SQL. Human identified they would fail bcrypt.compare() at login	Replaced with real hashes generated via script
3	Code Review	Bob used NextAuth v5 beta which broke Next.js 16 route handler types	Downgraded to NextAuth v4.24.15
4	Business Logic	Bob made manager nomination removable. Spec requires manager to be auto-locked	Added isMandatory flag; manager locked in nominations UI
5	UX Override	Bob allowed review submission with partial ratings. Human enforced: all 28 questions + both comments ≥ 20 chars	Full validation before submit enables
6	Business Logic	Dept-scoped peer validation missing — Zanele and Dikeledi could not review within-dept peers due to missing pool-size adaptive limits	Added effectiveMinNominees / effectiveMaxNominees across three endpoints

In-App HITL (User as Reviewer — AI as Assistant)

#	Feature	HITL Moment	What the Human Decides
7	AI Comment Suggestions	Reviewer sees AI-generated "doing well" and "improve" text before saving a review	Accept as-is / Edit / Discard — full control retained
8	AI Feedback Theme Summary	Employee or manager clicks Generate to surface top 3 strengths and growth areas	Human reads the output and decides how to act on it
9	AI PDF Report Generation	HR Admin generates AI narrative report → reviews it in a preview pane → optionally edits raw HTML	Must click Approve & Save before MCP tool writes the file
10	AI Self-Improvement Plan	Employee reviews weekly actions, monthly goals, success metrics → must click I've reviewed this — Approve	MCP write_csv tool only fires after explicit human approval

---

IBM Bob — Agentic AI Platform Notes

What Bob did in this project

IBM Bob was used as the primary development agent throughout the hackathon, playing multiple roles:

Role Bob Played	Examples
Full-Stack Developer	Scaffolded and implemented all Next.js pages, API routes, Prisma schema, and auth
Data Architect	Designed 11-table PostgreSQL schema, seeded 57 synthetic employees, modelled 8-phase cycle lifecycle
Security Reviewer	Identified bcrypt hash issue, enforced RBAC on every API route, added server-side nomination validation
QA / Tester	Ran tsc --noEmit after every change, caught TypeScript errors, validated API contracts
AI Feature Developer	Built all 4 AI endpoints (suggest-comments, theme-summary, improvement-plan, report) with GPT-4o + stub fallback
MCP Server Builder	Scaffolded, implemented, and registered the pulse360-mcp server with write_csv and generate_pdf_report tools
Technical Writer	Generated this README, the executive summary, and inline code documentation

Key IBM Bob capabilities demonstrated

• Multi-turn agentic reasoning — Bob maintained context across a full day of iterative feature requests without losing architectural context

• Human-in-the-Loop orchestration — Using Bob I built HITL checkpoints into the product itself (AI suggestions require human approval before persistence)

• MCP tool authoring — Using Bob I created a custom MCP server from scratch, registered it, and wired it into the app's approval flow (if you have a starterkit on .git use it)

• Text generation integration — All four AI features use GPT-5.1 for text generation, with Bob writing the prompts, parsing the responses, and building the stub fallback chain

• TypeScript discipline — Every change validated with tsc --noEmit before being considered complete

• Parallel task execution — Bob ran independent tasks in parallel (DB clear + API build + MCP build + page rebuild) to compress build time

• The best part is all this was done with full oversite and guidance from me through prompts and feature tunning.

MCP Server (pulse360-mcp)

Registered at .bob/mcp.json. Two tools:

Tool	Trigger	Guard
write_csv	User approves improvement plan → clicks Download	Post-approval only; description states "invoke ONLY after user review"
generate_pdf_report	HR Admin approves report → clicks Save to Downloads	Post-approval only; description states "invoke ONLY after human review"

---

Useful Commands

# ── Database ──────────────────────────────────────────────────────────────────

# Check container is running
podman ps --filter "name=pulse360_db"

# Start if stopped
podman start pulse360_db

# Connect to psql
podman exec -it pulse360_db psql -U pulse360_user -d pulse360

# Run a SQL script manually
Get-Content database\init\03_schema.sql | podman exec -i pulse360_db psql -U pulse360_user -d pulse360

# ── App ───────────────────────────────────────────────────────────────────────

# Install dependencies
cd pulse360; npm install

# Development server (http://localhost:3000)
cd pulse360; npm run dev

# Production build check
cd pulse360; npx next build

# Regenerate Prisma client (after schema changes)
cd pulse360; npx prisma generate

# ── MCP Server ────────────────────────────────────────────────────────────────

# Rebuild MCP server after changes
cd pulse360/mcp-server; npm run build

# ── Git ───────────────────────────────────────────────────────────────────────

# Check status
git status

# Push to Bitbucket
git add -A
git commit -m "feat: describe your change"
git push origin main

---

Built with IBM Bob · Pulse360 · Pulse Technologies Hackathon V1 · 2026