TD - Technical Director Workspace with LLM Integration

Workspace tổng hợp cho Technical Artist/Director với AI, LLM, Houdini, và Production Pipeline.

🚀 START HERE - GETTING_STARTED.md

---

🎯 What's New - LLM Integration

Dự án đã được cấu trúc lại và phát triển với LLM integration:

✅ What Was Added

1. 🤖 FastAPI Backend (api-server/)

   • LLM endpoints (chat, code generation, debug)
   • Houdini integration endpoints
   • Pipeline automation APIs
   • WebSocket support
   • RAG system với ChromaDB

2. 🌐 Next.js Frontend (web-dev/llm-assistant/)

   • Modern chat interface
   • Houdini remote control
   • Pipeline dashboard
   • Real-time updates
   • Responsive design

3. 📚 Enhanced Documentation

   • GETTING_STARTED.md - Complete setup guide
   • PROJECT_STRUCTURE.md - Architecture details
   • WORKSPACE_GUIDE.md - Workflows & examples

4. 🎨 Houdini Integration (houdini-integration/)

   • Python panels structure
   • HDA templates
   • API client for Houdini

---

📁 Project Structure

TD/
├── 📚 TC/                          # 16-week AI Course (131 modules)
│   ├── week08*/                    # ⭐ LLM content (base for this project)
│   └── [Complete documentation]
│
├── 🤖 api-server/                  # FastAPI LLM Backend ✨ NEW
│   ├── main.py
│   ├── routers/                    # LLM, Houdini, Pipeline
│   ├── services/                   # Business logic
│   └── core/                       # Config & logging
│
├── 🌐 web-dev/llm-assistant/       # Next.js Frontend ✨ NEW
│   ├── src/app/                    # Pages & components
│   └── package.json
│
├── 🎨 houdini-integration/         # Houdini Tools ✨ NEW
│   ├── python/                     # Python panels
│   └── hdas/                       # Digital assets
│
├── 📊 data/                        # Data & Models
│   ├── models/                     # LLM models
│   ├── prompts/                    # Prompt templates
│   └── knowledge-base/             # RAG database
│
├── 🐍 scripts-local/               # Python automation
├── 🎬 houdini-files/               # Scene files
└── 📁 projects/                    # Active projects

---

🚀 Quick Start

Prerequisites

• Python 3.9+ (Backend)
• Node.js 18+ (Frontend)
• Houdini 19.5+ (Optional, for integration)
• PostgreSQL 15+ (Optional, for production)
• Redis 7+ (Optional, for caching)

1. API Server Setup

cd api-server

# Create virtual environment
python -m venv venv
source venv/bin/activate  # macOS/Linux
# venv\Scripts\activate   # Windows

# Install dependencies
pip install -r requirements.txt

# Configure environment
cp .env.example .env
# Edit .env with your API keys:
# OPENAI_API_KEY=sk-...
# ANTHROPIC_API_KEY=sk-ant-...

# Start server
uvicorn main:app --reload --host 0.0.0.0 --port 8000

API Server: http://localhost:8000
API Docs: http://localhost:8000/docs
Health Check: http://localhost:8000/health

2. Web Frontend Setup

cd web-dev/llm-assistant

# Install dependencies
npm install

# Configure environment
cp .env.example .env.local
# Edit .env.local:
# NEXT_PUBLIC_API_URL=http://localhost:8000

# Start development server
npm run dev

Web App: http://localhost:3000

3. Database Setup (Optional)

# PostgreSQL
docker run -d --name postgres \
  -e POSTGRES_PASSWORD=password \
  -e POSTGRES_DB=llm_assistant \
  -p 5432:5432 postgres:15

# Redis
docker run -d --name redis \
  -p 6379:6379 redis:7-alpine

4. Learn Foundation

cd TC
./setup.sh && source venv/bin/activate
jupyter notebook week08*/  # Study LLM modules

📖 Full Guide: GETTING_STARTED.md

---

✨ Key Features

Feature	Description	API Endpoint
🤖 LLM Chat	Natural language interface for Houdini questions	POST /api/llm/chat
💻 Code Generation	Generate Python, VEX, HScript code	POST /api/llm/generate-code
🐛 Debugging	AI-powered code debugging	POST /api/llm/debug-code
⚡ Code Optimization	Optimize code for performance/readability	POST /api/llm/optimize-code
🎨 Houdini Control	Remote Houdini control via web	POST /api/houdini/execute-code
🏗️ Node Creation	Create Houdini nodes programmatically	POST /api/houdini/create-node
🌐 Network Generation	Generate node networks from description	POST /api/houdini/generate-network
⚙️ Pipeline Automation	Automated workflows	GET /api/pipeline/jobs
📊 Monitoring	Real-time job monitoring	WebSocket /ws/{client_id}
🔍 RAG Search	Context-aware responses with ChromaDB	Integrated in LLM service
🎯 Fine-tuning	Custom model training	Future feature
📝 Text Completion	Streaming text generation	POST /api/llm/completion

---

🎓 Learning Path

Based on TC Week 08 Content

Module	Learn	Apply To
m01	LLM basics	Architecture
m02	Houdini AI Bot	Base implementation ⭐
m03	Fine-tuning	Custom models
m04	Transformers	Technical depth
m05	Pipeline assistants	Automation
m06	LLM techniques	Prompt engineering
m07	Synthetic data	Data generation
m10	Digital Human	Production workflow ⭐

---

🔄 Example Workflows

1. Generate VEX Code

User → "Create VEX scatter points with noise"
  ↓
Web UI → API (/api/llm/generate-code)
  ↓
GPT-4/Claude → Generate optimized VEX
  ↓
User ← Formatted code + explanation
  ↓
Apply to Houdini Point Wrangle

2. Automated Digital Human

JSON spec → API → Houdini Integration
  ↓
1. ML Mesh topology generation
2. Diffusion model textures
3. Body matching (scan library)
4. Clothing system
  ↓
USD Export → Deployment

3. Debug Python Script

Error in Houdini
  ↓
Send code + error to API
  ↓
LLM analyzes & fixes
  ↓
Return: analysis + fixed code + explanation

---

🛠️ Tech Stack

Backend

Technology	Version	Purpose
FastAPI	0.109.0	Modern Python web framework
Python	3.9+	Programming language
Uvicorn	0.27.0	ASGI server
Pydantic	2.5.3	Data validation
PostgreSQL	15+	Relational database
Redis	7+	Cache & sessions
ChromaDB	0.4.22	Vector database (RAG)
SQLAlchemy	2.0.25	ORM
Alembic	1.13.1	Database migrations

Frontend

Technology	Version	Purpose
Next.js	14+	React framework (App Router)
TypeScript	5+	Type safety
Tailwind CSS	3+	Utility-first CSS
Radix UI	Latest	Accessible components
React Query	5+	Server state management
Zustand	4+	Client state management

AI/ML

Service	Model	Context Window	Purpose
OpenAI	GPT-4 Turbo	128K	Primary LLM
OpenAI	GPT-3.5 Turbo	16K	Fast responses
Anthropic	Claude 3 Opus	200K	Advanced reasoning
Anthropic	Claude 3 Sonnet	200K	Balanced performance
Local	Mistral 7B	8K	Offline inference
Embeddings	all-MiniLM-L6-v2	-	Vector embeddings

Houdini Integration

• Houdini Python API (hou) - Node manipulation
• Houdini RPC Server - Remote execution (Port 9090)
• HDAs - Digital assets
• Python Panels - In-app UI

---

📖 Documentation

Setup & Usage

• GETTING_STARTED.md ← START HERE
• ARCHITECTURE.md ← Kiến trúc hệ thống ⭐
• PROJECT_STRUCTURE.md - Project structure
• WORKSPACE_GUIDE.md - Workflows

TC Course

• TC/00-START-HERE.md
• TC/INDEX.md - 131 modules
• TC/TOPICS.md - A-Z lookup

API Documentation

• Swagger UI: http://localhost:8000/docs - Interactive API documentation
• ReDoc: http://localhost:8000/redoc - Alternative API docs
• OpenAPI Schema: http://localhost:8000/openapi.json

API Endpoints Summary

LLM Endpoints (/api/llm)

• POST /chat - Chat with LLM assistant
• POST /completion - Text completion (with streaming)
• POST /generate-code - Generate code (Python, VEX, HScript)
• POST /debug-code - Debug code with AI assistance
• POST /optimize-code - Optimize code
• GET /models - List available LLM models

Houdini Endpoints (/api/houdini)

• POST /execute-code - Execute Python code in Houdini
• POST /create-node - Create Houdini node
• GET /scene-info - Get current scene information
• POST /generate-network - Generate node network from description

Pipeline Endpoints (/api/pipeline)

• GET /jobs - List pipeline jobs
• POST /jobs - Create new job
• GET /jobs/{id} - Get job details

WebSocket

• WS /ws/{client_id} - Real-time communication

---

📊 Project Status

✅ Completed (v1.0.0 - 2026-01-06)

• Project restructuring
• API server architecture (FastAPI)
• FastAPI endpoints (LLM, Houdini, Pipeline, Assistant)
• WebSocket support for real-time communication
• Configuration management (Pydantic Settings)
• Logging system
• CORS middleware
• Comprehensive documentation
  • Architecture documentation
  • Getting started guide
  • Project structure guide
  • Workspace guide
• Integration design

🚧 In Progress

• Complete LLM service implementation
• Build RAG system with ChromaDB
• Create chat UI components (Next.js)
• Houdini Python panel
• Database migrations (Alembic)
• Authentication & authorization (JWT)

📋 Planned Features

• Production deployment (Docker)
• CI/CD pipeline
• Advanced monitoring (Prometheus, Grafana)
• Multi-tenant support
• Mobile app
• Fine-tuning pipeline
• Advanced RAG with fine-tuning

---

🎯 Use Cases

Technical Artists: Generate code, debug scripts, automate tasks Pipeline TDs: Monitor workflows, schedule jobs, optimize pipelines VFX Producers: Track progress, analyze metrics, generate reports

---

⚙️ Configuration

Environment Variables

API Server (.env)

# Server
HOST=0.0.0.0
PORT=8000
ENVIRONMENT=development
DEBUG=true

# LLM Providers
OPENAI_API_KEY=sk-...
OPENAI_MODEL=gpt-4-turbo-preview
ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_MODEL=claude-3-opus-20240229

# Local LLM
LOCAL_LLM_MODEL=mistral-7b-instruct
LOCAL_LLM_PATH=../data/models/llm/

# Vector Database
CHROMA_PATH=../data/knowledge-base/vector-db/chroma
CHROMA_COLLECTION=houdini_docs
EMBEDDING_MODEL=all-MiniLM-L6-v2

# Databases
DATABASE_URL=postgresql://user:password@localhost:5432/llm_assistant
REDIS_URL=redis://localhost:6379/0

# Security
SECRET_KEY=change-this-secret-key-in-production
ACCESS_TOKEN_EXPIRE_MINUTES=30

# Houdini
HOUDINI_HOST=localhost
HOUDINI_PORT=9090
HOUDINI_RPC_ENABLED=true

# Rate Limiting
RATE_LIMIT_PER_MINUTE=60

Frontend (.env.local)

NEXT_PUBLIC_API_URL=http://localhost:8000
NEXT_PUBLIC_WS_URL=ws://localhost:8000

💡 Getting Help

1. Start Here: GETTING_STARTED.md - Complete setup guide
2. Architecture: ARCHITECTURE.md - System architecture details
3. Structure: PROJECT_STRUCTURE.md - Project organization
4. Workflows: WORKSPACE_GUIDE.md - Usage examples
5. API Docs: http://localhost:8000/docs - Interactive API documentation
6. TC Course: Study TC Week 08 modules for LLM fundamentals

---

🔗 Links

• TC Course: https://github.com/aerovfx/TechnicalArtist.git
• FastAPI: https://fastapi.tiangolo.com/
• Next.js: https://nextjs.org/
• OpenAI: https://platform.openai.com/
• Houdini: https://www.sidefx.com/

---

📈 System Requirements

Minimum Requirements

• CPU: 4 cores
• RAM: 8GB
• Storage: 10GB free space
• Network: Internet connection for LLM APIs

Recommended Requirements

• CPU: 8+ cores
• RAM: 16GB+
• Storage: 50GB+ (for local models)
• GPU: NVIDIA GPU with 8GB+ VRAM (for local LLM inference)

Production Requirements

• CPU: 16+ cores
• RAM: 32GB+
• Storage: 100GB+ SSD
• Database: PostgreSQL 15+ with replication
• Cache: Redis 7+ cluster
• Load Balancer: Nginx or similar

🔐 Security Notes

⚠️ Important:

• Never commit .env files to version control
• Change SECRET_KEY in production
• Use strong passwords for databases
• Enable HTTPS in production
• Implement rate limiting
• Use environment-specific API keys

📝 License

This project is for educational and professional use by Technical Artists.

---

Built with ❤️ by Technical Artists, for Technical Artists

Version: 1.0.0
Last Updated: 2026-01-06

Ready to start? → GETTING_STARTED.md 🚀