TextileCV

AI-powered career toolkit — tailor resumes, write cover letters, and generate STAR interview answers, all from your own experience data. Built with LangChain + OpenAI + ChromaDB + a LaTeX resume compiler.

---

✨ Features

• Resume Tailor — Paste a job description, get a custom-tailored LaTeX resume matched to your real projects and skills.
• Cover Letter Architect — Narrative-driven cover letters that map your background to the role.
• STAR Generator — Behavioral interview answers built from your actual experience.
• LaTeX PDF Compilation — Server-side pdflatex compilation with fallback template repair.
• RAG Pipeline — Chunked markdown ingested into ChromaDB, retrieved via MMR for diverse, relevant context.
• Profile Manager — Upload files (PDF, DOCX, TXT), view generation history, manage preferences.
• Encrypted Vault — Your OpenAI API key is encrypted at rest (AES-256-GCM) with a master password and stored locally; it never leaves your machine.
• CLI — Install dependencies, start the server, and manage config from the terminal.

---

🛠️ Tech Stack

Layer	Technology
Runtime	Node.js, Express, TypeScript (ESM)
AI	LangChain.js, OpenAI (gpt-5.4-mini, text-embedding-3-small)
Vector Store	ChromaDB (Python, external)
LaTeX	node-latex + pdflatex (MiKTeX / TeX Live)
Database	sql.js (SQLite WASM) for logs, preferences, and encrypted vault
Client	React 18, Vite, Tailwind CSS
CLI	Commander.js, picocolors
Security	AES-256-GCM vault (Node crypto) for API-key encryption at rest

---

🚀 Quick Start

# 1. Clone and install
git clone https://github.com/MinnKhantZ/textilecv.git
cd textilecv
npm install

# 2. Install system dependencies (ChromaDB, Python, LaTeX)
npx textilecv install

# 3. Start the web interface
npx textilecv start
# → opens http://localhost:3001

On first launch the browser asks you to create a master password (this encrypts your API key at rest). Then set your OpenAI key in the Settings tab and upload your experience files in the Profile & Files tab.

---

📋 Prerequisites

Dependency	Purpose	Auto-installed by CLI?
Node.js ≥ 20	Runtime	No
npm ≥ 10	Package manager	No
Python 3	ChromaDB runtime	Yes (textilecv install)
ChromaDB	Vector store	Yes (textilecv install)
MiKTeX / TeX Live	LaTeX PDF compilation	Yes (textilecv install)
OpenAI API Key	Embeddings + chat	No (set via Settings / encrypted vault)

---

🏗️ Architecture

textilecv/
├── packages/
│   ├── server/        # Express API + LangChain + ChromaDB + LaTeX
│   ├── client/        # React + Vite + Tailwind SPA
│   └── cli/           # Commander.js CLI (bundles server+client for production)
├── scripts/           # Build scripts (copy dist into CLI bundle)
├── package.json       # npm workspaces root
└── README.md

• Development: run server and client separately with hot-reload.
• Production: CLI bundles server and client dist, serves the SPA from a single Express process.

---

💻 Development

Two terminals:

# Terminal 1 — API server (port 3001, hot-reload)
npm run dev:server

# Terminal 2 — Vite dev server (port 5173, hot-reload)
npm run dev:client

The Vite dev server proxies /generate-*, /uploads, /logs, /profile, /vault, /settings, and /health to localhost:3001.

Data Files

Place your experience data in packages/server/data/ (in development) — in production this lives at ~/.textilecv/data/:

File	Purpose
master_experience.md	Your projects, skills, work history (markdown, split on # headers)
about.md	Optional personal summary, preferences, career goals

Uploading files through the web interface automatically re-ingests them into ChromaDB — there is no separate manual ingestion step.

API Key & Vault

The OpenAI API key is stored encrypted (AES-256-GCM) inside the SQLite database and unlocked with a master password set on first launch. Set and manage it from the Settings tab in the web UI (works in both dev and production).

Environment Variables (optional)

packages/server/.env (dev) or ~/.textilecv/.env (production) — all optional:

CHROMA_URL=http://localhost:8000  # ChromaDB endpoint
PORT=3001                        # API server port
ALLOWED_ORIGINS=http://localhost:5173  # Comma-separated CORS origins

---

📟 CLI Commands

textilecv install          # Install Python, ChromaDB, LaTeX
textilecv start            # Start ChromaDB + Express server + SPA
textilecv start -p 8080    # Override port for this session
textilecv config --list    # Show current config
textilecv config --set port 8080
textilecv config --get host
textilecv uninstall        # Remove TextileCV data (prompts for Python/ChromaDB/LaTeX)
textilecv uninstall -y     # Skip prompts, uninstall everything
textilecv uninstall --keep-data  # Keep config and data files

Config is stored at ~/.textilecv/config.json. All user data (SQLite DB, uploads, ChromaDB vector store) lives under ~/.textilecv/.

---

📡 API Endpoints

Method	Endpoint	Description
GET	/vault/status	Vault setup/unlock state (public)
POST	/vault/setup	Create master password (first launch)
POST	/vault/unlock	Unlock the vault
POST	/vault/lock	Lock the vault
POST	/vault/change-password	Change master password
GET	/settings	AI settings (key redacted)
PATCH	/settings	Update AI provider/model/base URL/key
POST	/settings/ai/test	Test the configured API key
POST	/generate-resume	Tailor resume to a job description
GET	/generate-resume/pdf/:id	Download compiled PDF
GET	/generate-resume/source/:id	Download raw LaTeX
POST	/generate-cover-letter	Generate cover letter
POST	/generate-star-answers	Generate STAR answers for questions
POST	/uploads/:fileType	Upload a file (PDF, DOCX, TXT)
GET	/uploads/status	Upload + ingest status
GET	/logs	Generation history
GET	/health	Health check

---

📦 Build & Publish

# Full production build (client → server → CLI bundle)
npm run build

# Publish CLI to npm
npm run publish:cli

Build compiles the client, copies dist into the CLI package, compiles the server, copies dist into the CLI package, then compiles the CLI.

---

📄 License

MIT