Skip to content

Traigent SDK Demo Videos

This directory contains animated terminal demos showcasing Traigent's LLM optimization capabilities.

Demo Videos

Demo Description Key Concepts
optimize Core LLM Agent Optimization @traigent.optimize(), tuned variables, objectives, evaluation datasets
hooks Optimization Callbacks OptimizationCallback, lifecycle hooks, progress tracking, parameter importance
github-hooks Agent Configuration Hooks traigent-validate, cost/performance constraints, push rejection on violations

Quick Start

cd docs/demos

# Make scripts executable
chmod +x record-demos.sh scripts/*.sh

# Generate demos
./record-demos.sh

Output Files

After running record-demos.sh:

  • output/optimize.cast - asciinema recording (Video 1)
  • output/hooks.cast - asciinema recording (Video 2)
  • output/github-hooks.cast - asciinema recording (Video 3)
  • output/optimize.svg - animated SVG (if svg-term available)
  • output/hooks.svg - animated SVG (if svg-term available)
  • output/github-hooks.svg - animated SVG (if svg-term available)

Prerequisites

Required

  • Python 3.x
  • Bash shell

Optional (for SVG generation)

# Install svg-term-cli for animated SVG output
npm install -g svg-term-cli

Directory Structure

docs/demos/
├── README.md               # This file
├── record-demos.sh         # Master generation script
├── scripts/
│   ├── generate-cast.py    # Python cast file generator
│   ├── demo-optimize.sh    # Video 1: Core optimization
│   ├── demo-hooks.sh       # Video 2: Optimization callbacks
│   └── demo-github-hooks.sh # Video 3: GitHub hooks and CI/CD
├── output/
│   ├── *.cast              # asciinema format
│   └── *.svg               # Animated SVGs for GitHub
└── test_agents/            # Sample agents + traigent.yml for hooks demo

Embedding in README

## See Traigent in Action

### LLM Agent Optimization
![Traigent Optimization](output/optimize.svg)

### Optimization Callbacks
![Traigent Callbacks](output/hooks.svg)

### GitHub Hooks & CI/CD
![Traigent GitHub Hooks](output/github-hooks.svg)

Key Messages

Video 1: Core Optimization

  • Zero code changes - just add the decorator
  • Tuned variables = parameters that affect agent behavior (model, temperature, etc.)
  • Objectives = what you want to optimize (accuracy, cost, latency)
  • Evaluation-driven = uses test datasets to measure improvement
  • Results = find the best configuration automatically

Video 2: Optimization Callbacks

  • Lifecycle hooks for monitoring optimization progress
  • Built-in callbacks for common use cases (progress bar, logging, statistics)
  • Custom callbacks for notifications (Slack, email, dashboards)
  • Statistics for parameter importance analysis

Video 3: Agent Configuration Hooks

  • traigent-validate blocks pushes with invalid agent configurations
  • Cost constraints enforce budget limits (max cost per query, monthly budget)
  • Performance constraints ensure accuracy/latency meet baseline
  • Model constraints allow/block specific models for production
  • Push rejection prevents expensive mistakes before they reach production

Manual Playback

You can preview cast files with asciinema:

# Install asciinema
pip install asciinema

# Play a recording
asciinema play docs/demos/output/optimize.cast

Reference

See DEMO_VIDEO_GUIDE.md for implementation details and customization options.