​
LLMMessagesFrame

Frame containing a list of messages for LLM chat completions.

​
Properties

messages
List[dict]
required

List of message dictionaries representing the conversation history. Each message typically contains ‘role’ and ‘content’ keys.

​
Notes

Messages property is mutable and may be updated by ResponseAggregator processors.

​
LLMMessagesAppendFrame

Frame for adding new messages to the current context.

​
Properties

messages
List[dict]
required

List of new messages to append to the current context

​
LLMMessagesUpdateFrame

Frame for replacing the current context with new messages.

​
Properties

messages
List[dict]
required

List of messages that will replace the current context

​
LLMSetToolsFrame

Frame for configuring function calling capabilities.

​
Properties

tools
List[dict]
required

List of tool definitions, typically in JSON Schema format

​
LLMEnablePromptCachingFrame

Frame for controlling prompt caching behavior.

​
Properties

enable
bool
required

Whether to enable or disable prompt caching

​
LLMFullResponseStartFrame

Control frame indicating the start of an LLM response sequence.

​
LLMFullResponseEndFrame

Control frame indicating the end of an LLM response sequence.

​
Usage Examples

​
Basic Message Management

# Initial conversation context
context = LLMMessagesFrame(messages=[
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Hello!"},
])

# Append new messages
append_frame = LLMMessagesAppendFrame(messages=[
    {"role": "assistant", "content": "Hi! How can I help?"},
    {"role": "user", "content": "What's the weather?"},
])

# Update entire context
update_frame = LLMMessagesUpdateFrame(messages=[
    {"role": "system", "content": "New system prompt"},
    {"role": "user", "content": "New conversation"},
])

​
Function Calling Setup

# Configure tools/functions
tools_frame = LLMSetToolsFrame(tools=[{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get current weather",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {
                    "type": "string",
                    "description": "City name"
                }
            },
            "required": ["location"]
        }
    }
}])

​
Common Pipeline Usage

# Typical LLM pipeline
pipeline = Pipeline([
    context_manager,        # Manages LLMMessagesFrame
    llm_service,            # Processes messages, produces responses
    response_handler,       # Handles LLM responses
    tool_executor           # Handles function calls
])

​
Frame Flow

​
Notes

  • Messages should follow the chat completion format of your LLM provider
  • Tool definitions must match the schema expected by your LLM service
  • Response sequences are bounded by Start/End frames
  • Prompt caching can be controlled dynamically during pipeline execution
Text FramesFunction Calling Frames