Developer Onboarding Accelerator

Turn any GitHub repository into a complete onboarding package in minutes — powered by IBM Bob.

Built for the IBM Bob Dev Day Hackathon 2026 · Theme: Turn idea into impact faster

Quick start (fastest)

# Runs via the published npm package
npx repo-onboarder https://github.com/expressjs/express
# Output written to ./output

If IBM Bob CLI is installed and authenticated, it will be used automatically.
If not, the tool falls back to Groq (configure GROQ_API_KEY).

GROQ_API_KEY=your_key npx repo-onboarder <repo-url>

🎬 Demo Video 🎬

Click the thumbnail below to watch the demo on YouTube

📸 Screenshots

View all screenshots in bob_sessions/

| IBM Bob Shell | CLI — All 4 files via Bob | Web UI | Bob in Action | |---|---|---|---| | | | | |

The Problem

Every developer has been there: you join a new project, clone the repo, and stare at thousands of lines of unfamiliar code. Where do you start? What does this thing actually do? Which files matter?

Onboarding a developer to a new codebase takes 4–8 hours on average. That's time spent:

• Reading through files with no clear entry point
• Reverse-engineering architecture from code
• Hunting for setup instructions scattered across wikis and Slack
• Writing tests from scratch just to understand what functions do

Multiply that by every new hire, every open source contributor, every team member switching projects — and you have a massive, invisible tax on developer productivity.

The Solution

Developer Onboarding Accelerator is a CLI + web tool that takes any GitHub repository URL and uses IBM Bob to automatically generate a complete onboarding package in minutes.

One command:

node cli.js https://github.com/expressjs/express

Four files, ready to use:

| File | What it contains | |------|-----------------| | README-GENERATED.md | Plain English explanation of what the repo does, its purpose, and key features | | ARCHITECTURE.md | How the code is organized, main components, design patterns, and why the structure exists | | GETTING-STARTED.md | Where new developers should start, key files to read first, setup steps, and common workflows | | TESTS-GENERATED.js | Unit tests for the main functions — ready to run, covering typical use cases and edge cases |

📂 View example output files generated by IBM Bob

IBM Bob reads the actual code, understands it, and writes documentation that would take a developer hours to produce manually.

Impact

| Metric | Before | After | |--------|--------|-------| | Time to get an actionable orientation | 4–8 hours | ~5–10 minutes | | Documentation written manually | Hours of work | Zero | | Barrier to open source contribution | High | Low | | New hire productivity | Day 3–5 | Day 1 |

Who benefits:

• New hires — get up to speed on day one instead of day five
• Open source maintainers — lower the barrier for contributors
• Engineering teams — reduce the onboarding tax across every project switch
• Solo developers — return to old projects without re-reading everything

AI Benchmark: IBM Bob vs Others

We ran the same 4 prompts against the same repo (expressjs/express) using IBM Bob, ChatGPT 4o, Perplexity Sonar, and Manus AI.

| Dimension | IBM Bob | ChatGPT 4o | Perplexity | Manus AI | |-----------|---------|------------|------------|----------| | README quality | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐ | ⭐⭐⭐ | | Architecture depth | ⭐⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐ | ⭐⭐⭐⭐ | | Getting started | ⭐⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐ | ⭐⭐⭐⭐ | | Tests generated | ⭐⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ | | Automation | ⭐⭐⭐⭐⭐ | ⭐ | ⭐ | ⭐ |

IBM Bob produced the deepest, most actionable output — and it's fully automated in one command.

📊 Full comparison with detailed analysis →

How IBM Bob Is Used

IBM Bob is the primary AI engine for this tool. Bob is not a helper or an add-on — it is the core of what makes this work.

For each repository, the tool:

1. Clones the repo and analyzes the top 20 most relevant files (entry points + core modules + key configs)
2. Builds a structured codebase snapshot (code + file structure)
3. Calls bob do "<prompt>" four times with targeted, role-based prompts:
   • "You are a technical documentation expert. Given this codebase snapshot, generate a comprehensive README..."
   • "You are a software architect. Generate an architecture overview explaining how the code is organized..."
   • "You are a developer onboarding specialist. Generate a getting started guide for new developers..."
   • "You are a test engineer. Generate unit tests for the main functions..."
4. Captures Bob's output and writes it directly to files

Bob's ability to understand code context, infer intent, and write in natural language is what makes the output genuinely useful — not just a file listing or a summary, but real documentation a developer can act on immediately.

Fallback chain: If Bob times out or is unavailable (e.g., in cloud deployment), the tool automatically falls back to a Groq LLM chain:
llama-3.3-70b-versatile → meta-llama/llama-4-scout-17b-16e-instruct → llama-3.1-8b-instant → allam-2-7b.

Demo

CLI:

node cli.js https://github.com/expressjs/express

🚀 Developer Onboarding Accelerator

Repository: https://github.com/expressjs/express

✔ Cloning repo...
   Cloned to .temp-repos/express
✔ Analyzing codebase...
   Found 146 relevant files, analyzing top 20
✔ Generating README-GENERATED.md...   ✅ Generated via IBM Bob
✔ Generating ARCHITECTURE.md...       ✅ Generated via IBM Bob
✔ Generating GETTING-STARTED.md...    ✅ Generated via IBM Bob
✔ Generating TESTS-GENERATED.js...    ✅ Generated via IBM Bob

✅ Done! 4 files generated in ./output

Web UI:

node server.js
# Open http://localhost:3000

Paste any GitHub URL, click Run Bob, and watch the progress stream live. Results appear in 4 tabs with markdown rendering and syntax highlighting.

Setup

Prerequisites

• Node.js 14+
• IBM Bob CLI installed and authenticated (bob do "hello" should work)
• Git

Install

git clone https://github.com/RandomProjects-db/Developer-Onboarding-Accelerator.git
cd Developer-Onboarding-Accelerator
npm install

Configure

cp .env.example .env

Edit .env:

GROQ_API_KEY=your_groq_api_key_here   # Free at https://console.groq.com

IBM Bob is used automatically if you have the Bob CLI installed and authenticated locally. GROQ_API_KEY is only needed for the fallback.

NPM Package

repo-onboarder is published on npm and can be used without cloning this repository.

Installation & Usage

# Option 1: Run directly with npx (no installation required)
npx repo-onboarder https://github.com/expressjs/express

# Option 2: Install globally
npm install -g repo-onboarder
repo-onboarder https://github.com/expressjs/express

Configuration

Create a .env file in your current directory with your Groq API key:

# .env
GROQ_API_KEY=your_groq_api_key_here

Or set it as an environment variable:

GROQ_API_KEY=your_key npx repo-onboarder <repo-url>

How It Works

1. Primary: Uses IBM Bob CLI if installed locally (bob do command)
2. Fallback: Uses Groq API with model chain:
   • llama-3.3-70b-versatile
   • meta-llama/llama-4-scout-17b-16e-instruct
   • llama-3.1-8b-instant
   • allam-2-7b

Output files are saved to ./output/ in your current directory.

Docker

Run the tool in a container without installing Node.js locally. The Docker image is automatically published to GitHub Container Registry.

Pull and Run

# Pull the latest image
docker pull ghcr.io/randomprojects-db/developer-onboarding-accelerator:latest

# Run the web UI
docker run -p 3000:3000 \
  -e GROQ_API_KEY=your_key_here \
  ghcr.io/randomprojects-db/developer-onboarding-accelerator:latest

# Open http://localhost:3000

Run CLI in Docker

# Run CLI and save output to local directory
docker run --rm \
  -v $(pwd)/output:/app/output \
  -e GROQ_API_KEY=your_key_here \
  ghcr.io/randomprojects-db/developer-onboarding-accelerator:latest \
  node cli.js https://github.com/expressjs/express

Build Locally

# Build from source
docker build -t dev-onboarder .

# Run
docker run -p 3000:3000 -e GROQ_API_KEY=your_key dev-onboarder

Usage

CLI

node cli.js <github-repo-url>

Output files are saved to ./output/.

Web UI

node server.js

Open http://localhost:3000, paste a GitHub URL, click Run Bob.

Project Structure

dev-onboarder/
├── cli.js          # Entry point — accepts GitHub URL as CLI argument
├── index.js        # Orchestrator — clone → analyze → generate
├── analyzer.js     # Walks directory tree, builds codebase snapshot
├── generator.js    # Calls Bob CLI, captures output, writes 4 files
├── server.js       # Express server — wraps CLI, streams via SSE
├── public/
│   ├── index.html  # Single page web UI
│   ├── style.css   # Dark theme styling
│   └── app.js      # Frontend — SSE client, markdown rendering
└── .env.example    # Environment variable template

Running the Generated Tests

The generated TESTS-GENERATED.js uses Mocha + Supertest. Tests are fully runnable out of the box:

npm test

Or run directly:

npx mocha output/TESTS-GENERATED.js --timeout 5000

Example output:

  Express App
    ✔ should return 200 for GET /
    ✔ should return 404 for GET /unknown
    ✔ should return 200 for POST / with middleware
    ✔ should return 500 for GET / with error
    ✔ should return 200 for GET / with route parameter
    ✔ should return 200 for GET / with query parameter
    ✔ should return 200 for GET / with header

  7 passing (78ms)

Self-healing tests: The generator automatically runs the tests after generation. If any fail, it feeds the errors back to the AI to fix them — producing all-passing tests without manual intervention.

Note: Generated tests are a starting point. They validate core functionality of the analyzed repo. You may want to extend them with project-specific edge cases.

Why This Matters

The theme of this hackathon is "Turn idea into impact faster." Developer onboarding is one of the biggest hidden costs in software development — and it's solvable with AI.

This tool doesn't just demonstrate what IBM Bob can do. It solves a real problem that every developer faces, every time they touch a new codebase. The impact is immediate, measurable, and scales across teams and open source.

bob_sessions

See the bob_sessions/ folder for screenshots and logs of IBM Bob generating documentation during development and testing of this tool.

License

MIT