feat: Add comprehensive LLM cheatsheet for evaluation framework

[mlpierce22]

Summary

• Add docs/llms.md as comprehensive reference guide for TrainLoop's evaluation framework
• Enables users to provide complete context to LLMs when seeking evaluation implementation help
• Covers setup, SDK integration, metrics/suites rules, benchmarking, and visualization

Contents

Core Documentation

• Setup Phase: Project initialization with trainloop init and folder structure
• SDK Integration: Multi-language patterns for Python, TypeScript, Go with zero-touch instrumentation
• Evaluation System: Complete rules for metrics (must return 1/0) and suites (tag-based combinations)
• Configuration: trainloop.config.yaml settings and workflow commands
• Benchmarking: Provider comparison for performance tracking
• Visualization: TrainLoop Studio for interactive analysis

Key Features

• Comprehensive Examples: End-to-end walkthrough with code samples
• Critical Rules: All evaluation requirements and constraints clearly documented
• Best Practices: Performance optimization and data management guidance
• Common Patterns: Ready-to-use examples for code quality, customer service, content validation

Use Case

This cheatsheet can be:

• Downloaded by users as a reference guide
• Provided to LLMs as context for automated evaluation assistance
• Used as complete documentation for evaluation framework implementation

The document covers everything from initial setup through advanced benchmarking scenarios, making it a one-stop resource for LLM evaluation with TrainLoop.

Test plan

• Verify file is accessible in docs directory
• Confirm markdown formatting renders correctly
• Validate all code examples match actual framework patterns
• Test downloadability from docs site