Resume Builder

AI-powered resume builder with LangGraph multi-agent orchestration and a visual regression CI pipeline.

The project migrated from a monolithic agent system (V1/agent-core) to LangGraph's state graph architecture (V2/agent-graph) for explicit state management, checkpointing, and SSE streaming. V2 is the default.

CI/CD Pipeline

Every pull request targeting main runs the full browser automation + visual regression suite:

PR opened └─ browser-automation-tests.yml ├─ pnpm type-check (all packages) ├─ skill-audit (shared composite action from ojfbot/github-actions@v1) ├─ Docker Compose: browser-app + api + browser-automation ├─ Comprehensive tests + visual regression tests ├─ Generate PR comment (test outcomes + baseline coverage) ├─ Deploy draw.io architecture viewer → GitHub Pages ├─ [if S3_BUCKET set] Upload baseline PNGs to S3 └─ [on main push] Commit updated draw.io canvas to repo

**Live architecture viewer:** Every CI run publishes a draw.io viewer (with real screenshots injected by the pipeline) to **[ojfbot.github.io/cv-builder](https://ojfbot.github.io/cv-builder/)**. Links are posted automatically in the PR comment.

Visual regression baselines live in `packages/browser-automation/test-baselines/cv-builder-visual/`. Regenerate by triggering the workflow with `update_baselines: true`. When AWS credentials are configured, the pipeline uploads baselines to S3 and injects public URLs into the draw.io canvas. See [docs/AWS_CI_SETUP.md](docs/AWS_CI_SETUP.md) for infrastructure setup.

## Project Structure

This is a monorepo with the following packages:

resume-builder/ ├── packages/ │ ├── agent-core/ # V1: Legacy agent system │ │ ├── src/ │ │ │ ├── agents/ # Specialized AI agents │ │ │ ├── models/ # Data models │ │ │ ├── utils/ # Utilities │ │ │ └── cli/ # CLI interface │ │ └── package.json │ │ │ ├── agent-graph/ # V2: LangGraph multi-agent system ⭐ DEFAULT │ │ ├── src/ │ │ │ ├── graphs/ # StateGraph definitions │ │ │ ├── nodes/ # Agent nodes │ │ │ ├── state/ # State management & checkpointing │ │ │ ├── rag/ # Vector stores & retrievers │ │ │ └── utils/ # Utilities │ │ └── package.json │ │ │ ├── api/ # Express API server │ │ ├── src/ │ │ │ ├── routes/ # API endpoints (V1 & V2, + GET /api/beads) │ │ │ ├── services/ # AgentManager & GraphManager │ │ │ └── middleware/ # Auth, validation, errors │ │ └── package.json │ │ │ ├── browser-app/ # React browser UI │ │ ├── src/ │ │ │ ├── components/ # React components │ │ │ ├── store/ # Redux state │ │ │ ├── api/ # API client │ │ │ └── services/ # Browser services │ │ └── package.json │ │ │ ├── browser-automation/ # Playwright visual regression + CI pipeline │ │ ├── src/ │ │ │ ├── drawio/ # draw.io URL injector │ │ │ └── storage/ # S3 uploader │ │ ├── templates/drawio/ # cvBuilder.drawio.xml + screenshot manifest │ │ ├── tests/ # Playwright test suites │ │ └── scripts/ # ci-screenshot-pipeline.ts │ │ │ └── visual-dashboard/ # Visual regression dashboard React app │ ├── src/ │ │ ├── components/ # DrawioCanvas, DiagramViewer, … │ │ └── utils/ # drawioParser.ts │ └── package.json │ ├── .github/workflows/ │ └── browser-automation-tests.yml # Full CI/CD pipeline │ ├── docs/ # Documentation │ ├── CI_CD_PIPELINE.md # Complete CI/CD pipeline reference │ ├── AWS_CI_SETUP.md # S3 + OIDC one-time setup guide │ ├── technical/ # Technical docs & ADRs │ └── how-to/ # Guides ├── V2_QUICKSTART.md # V2 quick start guide ├── docker-compose.yml # Docker orchestration ├── docker-compose.ci.yml # CI-specific Docker Compose ├── Dockerfile # Agent system container └── package.json # Root workspace config

## Getting Started

### Prerequisites

- Node.js 22.11.1+ (LTS)
- pnpm 9.0.0+
- Docker (optional)
- Anthropic API key
- fnm (recommended for Node version management)

### Node Version Management

This project uses `.nvmrc` to pin the Node version. If you have `fnm` installed:

```bash
# Install the correct Node version
fnm use

# Or install if not present
fnm install

Installation

# Install pnpm globally (if not already installed)
corepack enable
corepack prepare pnpm@9.15.4 --activate

# Install dependencies
pnpm install

Configuration

You can configure the application using either env.json (recommended) or .env.local:

Option 1: env.json (Recommended)

Create env.json in packages/agent-core/:

cp packages/agent-core/env.json.example packages/agent-core/env.json
# Edit env.json and add your API key

Example env.json:

{
  "anthropicApiKey": "your_api_key_here",
  "directories": {
    "bio": "bio",
    "jobs": "jobs",
    "output": "output",
    "public": "public"
  },
  "model": "claude-sonnet-4-20250514"
}

Option 2: .env.local (Legacy)

Create .env.local file (using .env.local to avoid conflicts with Claude CLI):

cp .env.example .env.local
# Edit .env.local and add your API key

Example .env.local:

ANTHROPIC_API_KEY=your_api_key_here
VITE_ANTHROPIC_API_KEY=your_api_key_here

🔐 Security

IMPORTANT: API keys and secrets must NEVER be committed to git.

• env.json and .env.local are gitignored
• Pre-commit hooks scan for API keys
• Build artifacts (dist/, build/) are never committed
• Run pnpm security:verify to check for security issues

See SECURITY.md for detailed security policies and incident reporting.

Development

Run Full Stack (V1 - Legacy)

pnpm dev:all        # API server + Browser UI (agent-core)

Run Full Stack (V2 - LangGraph) ⭐ DEFAULT

pnpm dev:v2         # API server + Browser UI (agent-graph)
# Or use pnpm dev:all - V2 is now the default mode in the browser UI

This uses the new LangGraph-based architecture with:

• 🔄 Multi-agent orchestration
• 💾 State persistence (checkpointing)
• 🧵 Thread-based conversations
• 📡 Streaming support (SSE)

Note: V2 (LangGraph) mode is now enabled by default in the browser UI. Users can toggle between V1 and V2 modes using the toggle in the dashboard header.

See V2_QUICKSTART.md for details.

Individual Services

pnpm dev            # Browser UI only (port 3000)
pnpm dev:api        # API server only (port 3001)

CLI Agent System

pnpm cli            # Interactive CLI mode
pnpm cli:headless   # Headless mode

Docker

Build

pnpm docker:build

Run

docker-compose up

Documentation

Comprehensive documentation is available in the /docs directory:

Getting Started

• Setup Guide - Detailed setup and configuration
• Quick Start - Get up and running quickly
• Architecture - System architecture overview

Development

• Agents Guide - Working with AI agents
• Badge Actions - Interactive UI actions
• Navigation System - Tab navigation
• Browser Integration - Browser app integration

Deployment & CI/CD

• CI/CD Pipeline - Complete pipeline reference (GitHub Actions, GitHub Pages, draw.io canvas)
• AWS CI Setup - S3 + OIDC one-time infrastructure setup
• Docker Guide - Docker setup and deployment

Reference

• Technical Documentation - In-depth technical guides
• How-To Guides - Step-by-step tutorials
• Archive - Historical documentation and migration guides

License

MIT

Frame OS Ecosystem

Part of Frame OS — an AI-native application OS.

Repo	Description
shell	Module Federation host + frame-agent LLM gateway
core	Workflow framework — 30+ slash commands + TypeScript engine
cv-builder	AI-powered resume builder with LangGraph agents (this repo)
blogengine	AI blog content creation platform
TripPlanner	AI trip planner with 11-phase pipeline
core-reader	Documentation viewer for the core framework
lean-canvas	AI-powered lean canvas business model tool
gastown-pilot	Multi-agent coordination dashboard
seh-study	NASA SEH spaced repetition study tool
daily-logger	Automated daily dev blog pipeline
purefoy	Roger Deakins cinematography knowledge base
MrPlug	Chrome extension for AI UI feedback
frame-ui-components	Shared component library (Carbon DS) — published as @ojfbot/frame-ui-components on npm
github-actions	Shared composite GitHub Actions (skill-audit CI, etc.)
asset-foundry	Asset pipeline with dual Blender transports — Frame MF remote at :3035