LLMAgent (llm_agent v0.2.0)

LLMAgent is an abstraction library for building domain-specific intelligent agents based on Large Language Models (LLMs). It provides a core architecture and behavior definitions that simplify the development of specialized agents.

LLMAgent is built on top of AgentForge's signal-driven architecture and provides LLM-specific interaction patterns, signal types, message processing handlers, tool integration, and conversation management.

Key Features

LLM-specific signal types for agent interactions
Predefined handlers for processing LLM signals
Standard flow compositions for common agent patterns
Plugin-based extensions for LLM providers
Task management for long-running operations

Usage Example

# Create agent with system prompt and basic tools
{flow, initial_state} = LLMAgent.Flows.conversation(
  "You are a helpful assistant that can answer questions and use tools.",
  [
    %{
      name: "search",
      description: "Search the web for information",
      parameters: %{
        "type" => "object",
        "properties" => %{
          "query" => %{
            "type" => "string",
            "description" => "Search query"
          }
        },
        "required" => ["query"]
      },
      execute: &MyApp.Tools.search/1
    }
  ]
)

# Process a user message
{:ok, result, new_state} = AgentForge.process(
  flow,
  LLMAgent.Signals.user_message("What is the capital of France?"),
  initial_state
)

Summary

Functions

new(system_prompt, tools \\ [], options \\ [])

Creates a new LLM agent with the given system prompt and tools.

process(flow, state, message, options \\ [])

Process a user message through an agent flow.

version()

Returns the current version of LLMAgent.

Functions

new(system_prompt, tools \\ [], options \\ [])

Creates a new LLM agent with the given system prompt and tools.

This is a convenience function that creates a flow and initial state for a conversational agent.

Parameters

system_prompt - The system prompt that defines the agent's behavior
tools - A list of tools that the agent can use
options - Additional options for configuring the agent

Returns

A tuple containing the flow and initial state for the agent.

Examples

iex> {flow, state} = LLMAgent.new("You are a helpful assistant.", [])
iex> is_function(flow) and is_map(state)
true

process(flow, state, message, options \\ [])

Process a user message through an agent flow.

Parameters

flow - Flow to process the message with (function or list of handlers)
state - Current agent state
message - User message to process
options - Processing options

Returns

Result of processing the message through the flow.

Examples

iex> {flow, state} = LLMAgent.new("You are a helpful assistant.", [])
iex> {:ok, result, _} = LLMAgent.process(flow, state, "Hello")
iex> result.type == :response
true

version()

Returns the current version of LLMAgent.

Examples

iex> LLMAgent.version()
"3.0.0"