Creates a batch processing flow.

A batch processing flow iterates over a collection of items and applies the batch handler to each item.

Parameters

• items - The collection of items to process
• batch_handler - The handler function to apply to each item
• options - Additional options for batch processing

Returns

A function that can be used as a flow.

Examples

iex> items = [1, 2, 3]
iex> handler = fn signal, state -> {{:emit, signal}, state} end
iex> flow = LLMAgent.Flows.batch_processing(items, handler)
iex> is_function(flow)
true

Creates a standard conversation flow with the given system prompt and tools.

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.Flows.conversation("You are a helpful assistant.", [])
iex> is_function(flow) and is_map(state)
true

Maps a flow to a new flow by applying a transformation function.

Useful for creating derived flows that modify signal or state handling.

Parameters

• flow - The original flow
• transform_fn - A function that transforms the flow's result

Returns

A new flow that applies the transformation.

Examples

iex> flow = fn signal, state -> {{:emit, signal}, state} end
iex> transform = fn {result, state} -> {result, Map.put(state, :transformed, true)} end
iex> new_flow = LLMAgent.Flows.map_flow(flow, transform)
iex> is_function(new_flow)
true

Creates a flow for a simple question-answering agent.

This is a specialized flow for handling basic question-answering without tools, providing a streamlined implementation for simple use cases.

Parameters

• system_prompt - The system prompt that defines the agent's behavior
• options - Additional options for configuring the agent

Returns

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

Examples

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

Creates a task flow with the given task definition.

A task flow executes a series of primitive operations defined in the task definition.

Parameters

• task_definition - The task definition as a list of AgentForge primitives
• options - Additional options for the task flow

Returns

A function that can be used as a flow.

Examples

iex> task_def = [
...>   fn signal, state -> {:ok, "Step 1", state} end
...> ]
iex> flow = LLMAgent.Flows.task_flow(task_def)
iex> is_function(flow)
true

Creates a flow for a tool-using 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> tools = [%{name: "get_time", execute: fn _ -> %{time: DateTime.utc_now()} end}]
iex> {flow, state} = LLMAgent.Flows.tool_agent("You are a helpful assistant.", tools)
iex> is_function(flow) and is_map(state)
true

Adds middleware to a flow.

Middleware functions are executed before and after the main flow, allowing for consistent pre- and post-processing.

Parameters

• flow - The original flow
• middleware - A function that receives the signal, state, and a continuation

Returns

A new flow with the middleware applied.

Examples

iex> flow = fn signal, state -> {{:emit, signal}, state} end
iex> middleware = fn signal, state, continue ->
...>   {result, new_state} = continue.(signal, state)
...>   {result, Map.put(new_state, :middleware_applied, true)}
...> end
iex> new_flow = LLMAgent.Flows.with_middleware(flow, middleware)
iex> is_function(new_flow)
true