Traigent SDK Documentation

The first optimization platform that enhances your LLM applications without changing your code.

🚀 Quick Start

Install the current Traigent SDK release with pip install "traigent[recommended]", then run traigent onboard, traigent auth device-login, traigent quickstart, traigent first-prompt --agent claude|cursor|codex|copilot, or traigent mcp serve.

License: Traigent SDK is dual-licensed under AGPL-3.0-only OR LicenseRef-Traigent-Commercial.

import litellm
import traigent

@traigent.optimize(
    eval_dataset="examples/datasets/rag-optimization/evaluation_set.jsonl",
    objectives=["accuracy", "cost"],
    configuration_space={"model": ["gpt-4o-mini", "gpt-4o"], "temperature": [0.0, 0.7]},
)
def answer_question(question: str) -> str:
    cfg = traigent.get_config()  # Active trial/applied config
    response = litellm.completion(
        model=cfg.get("model"),
        temperature=cfg.get("temperature"),
        messages=[{"role": "user", "content": question}],
    )
    return response.choices[0].message.content

# Async-safe in Traigent
# results = await answer_question.optimize(max_trials=5)

📚 Documentation

🎯 Getting Started

• Installation - Install Traigent SDK
• Getting Started - Your first optimization in 5 minutes
• Minimum Integration - Smallest API/SDK surface to first successful trial

📖 User Guide

• Agent Optimization - Optimize AI agents and workflows
• Choosing Models - Select the right optimization model
• Injection Modes - Different ways to integrate Traigent
• Interactive Optimization - Real-time optimization workflows
• Evaluation Guide - Measure and improve performance
• Smart Optimization - Backend-routed smart optimizer strategies

🧭 Guides

• Optimization Routing - Default smart optimization, local search, portal sync, and no-egress runs
• Evaluation - Evaluation best practices and troubleshooting
• Parallel Configuration - Concurrency settings and tuning
• Secrets Management - Securely manage API keys

🔧 API Reference

• Complete Function Specification - Full API documentation
• Interactive Optimizer - Hybrid optimization API surface
• Thread Pool Examples - Context propagation in threads

🏗️ Architecture

• System Architecture - How Traigent works internally
• Project Structure - Codebase organization

🤝 Contributing

• Contributing Guide - How to contribute to Traigent
• Code of Conduct - Community guidelines
• Security Policy - Security reporting and practices

🎯 What is Traigent?

Traigent is a zero-code optimization platform that automatically finds the best configurations for your LLM applications. Simply add a decorator to your existing functions, and Traigent will:

• 🎯 Optimize multiple objectives (accuracy, cost, speed, etc.)
• 🔄 Test different configurations automatically
• 📊 Provide detailed analytics on performance
• 🚀 Improve your applications without code changes

💡 Key Features

• Zero Code Changes - Works with existing functions
• Multi-Objective Optimization - Balance accuracy, cost, and speed
• Framework Agnostic - Works with OpenAI, Anthropic, LangChain, and more
• Production Hardening Friendly - Structured for release validation and production deployment work
• Real-time Analytics - Monitor optimization progress live

🌐 Examples Navigator

Use Examples Guide, Start Here, and Quick Reference for runnable examples.

🧪 Example Guides

• Examples Guide - How the example tree is organized
• Start Here - Fast path to the right example
• Quick Reference - Common commands and knobs
• Troubleshooting - Common example run issues
• API Patterns - Reusable example implementation patterns

🚀 Next Steps

1. Install Traigent - Get set up in minutes
2. Follow the Getting Started Guide - Your first optimization
3. Use the Example Guides - Start with Examples Guide
4. Explore User Guides - Learn advanced features
5. Check API Reference - Detailed technical documentation

🧪 Mock Mode (No API Keys)

• Enable: TRAIGENT_MOCK_LLM=true (no API keys needed)
• Benefits: realistic scores, zero cost, quick iteration
• Example:
• TRAIGENT_MOCK_LLM=true python examples/core/rag-optimization/run.py
• Or copy a command from the Examples Navigator "Run" icon

---

Need help? Check our guides above or reach out to the community!