Deterministic Generation by Default | Seed Control for Stochastic Generation | Input/Output Hash Verification | Hashes with explore() | Automatic Documentation | Best Practices for Reproducible Research | 1. Always Set Seeds | 2. Log Your Environment | 3. Use Document Functions for Audit Trails | 4. Share Hashes for Verification | 5. Version Control Your Models | Summary

Installation Issues | "Backend library is not loaded" error | Installation fails on my platform | "Library already installed" but functions don't work | Model Download Issues | "Download lock" or "Another download in progress" error | Download times out or fails | "Model not found" when using cached model | Private Hugging Face model fails | Memory Issues | R crashes when loading a model | "Memory check failed" warning | Context creation fails with large n_ctx | GPU Issues | GPU not being used | GPU runs out of memory | Generation Issues | Backend prints too many log messages | Output is garbled or nonsensical | Output contains strange tokens like <|eot_id|> | Generation stops too early | Same prompt gives different results | Performance Issues | Generation is very slow | Parallel processing isn't faster | Compatibility Issues | "GGUF format required" error | Model works in Ollama but not localLLM | Common Error Messages | Getting Help | Quick Reference

Installation | Step 1: Install the R package | Step 2: Install the backend library | Your First LLM Query | Text Classification Example | Processing Multiple Prompts | Finding and Using Models | GGUF Format | Loading Different Models | Managing Cached Models | Customizing Generation | Next Steps

Why Parallel Processing? | Using generate_parallel() | Basic Usage | Progress Tracking | Text Classification Example | Sequential vs Parallel Comparison | Sequential (For Loop) | Parallel | Benchmark: Multiple Models | Using quick_llama() for Batches | Performance Considerations | Context Size and n_seq_max | Memory Usage | Batch Size Recommendations | Error Handling | Complete Workflow | Summary | Tips | Next Steps

The Core Workflow | Step 1: Loading a Model | Model Loading Options | Step 2: Creating a Context | Context Parameters | Step 3: Formatting Prompts with Chat Templates | Multi-Turn Conversations | Step 4: Generating Text | Generation Parameters | Complete Example | Tokenization | Tips and Best Practices | 1. Reuse Models and Contexts | 2. Size Your Context Appropriately | 3. Controlling Log Output (verbosity) | 4. Use GPU When Available | Next Steps

The explore() Function | Creating Structured Prompts | Template Builder Format | Running the Comparison | Viewing Results | Validation Against Ground Truth | Confusion Matrices | Reliability Metrics | Alternative Prompt Formats | Character Vector | Custom Function | Model-Specific Prompts | Computing Metrics Separately | Intercoder Reliability | Complete Example | Summary | Next Steps

Discovering Ollama Models | Loading Ollama Models | By Model Name | By Tag | By SHA256 Prefix | Interactive Selection | Using with quick_llama() | Ollama Reference Trigger Rules | Common Workflows | Check Available Models First | Load Specific Model | Model Comparison with Ollama | Ollama Directory Structure | Troubleshooting | Model Not Found | Ollama Not Installed | Multiple Matches | Benefits of Ollama Integration | Complete Example | Summary | Next Steps