apply_chat_template: Apply Chat Template to Format Conversations

In localLLM: Running Local LLMs with 'llama.cpp' Backend

Apply Chat Template to Format Conversations

Description

Formats conversation messages using the model's built-in chat template or a custom template. This is essential for chat models that expect specific formatting for multi-turn conversations.

Usage

apply_chat_template(model, messages, template = NULL,
  add_assistant = TRUE)

Arguments

model	A model object created with model_load
messages	List of chat messages, each with 'role' and 'content' fields. Role should be 'user', 'assistant', or 'system'
template	Optional custom template string (default: NULL, uses model's built-in template)
add_assistant	Whether to add assistant prompt suffix for response generation (default: TRUE)

Value

Formatted prompt string ready for text generation

See Also

model_load, quick_llama, generate

Examples

## Not run: 
# Load a chat model
model <- model_load("path/to/chat_model.gguf")

# Format a conversation
messages <- list(
  list(role = "system", content = "You are a helpful assistant."),
  list(role = "user", content = "What is machine learning?"),
  list(role = "assistant", content = "Machine learning is..."),
  list(role = "user", content = "Give me an example.")
)

# Apply chat template
formatted_prompt <- apply_chat_template(model, messages)

# Generate response
response <- quick_llama(formatted_prompt)

## End(Not run)

localLLM documentation built on Nov. 3, 2026, 1:09 a.m.