Built-in tools for task tracking and progress management.

TodoTools allows AI agents to break down complex tasks, track progress, and maintain focus on multi-step operations. Todos are automatically injected into the system prompt when enable_todos: true.

Setup

Enable todos when creating an agent:

agent = Nous.new("lmstudio:qwen3-vl-4b-thinking-mlx",
  instructions: "You are a helpful assistant",
  enable_todos: true,  # Enable todo tracking
  tools: [
    &TodoTools.add_todo/2,
    &TodoTools.update_todo/2,
    &TodoTools.complete_todo/2,
    &TodoTools.list_todos/2
  ]
)

Initial todos (optional):

{:ok, result} = Nous.run(agent, "Build a REST API",
  deps: %{todos: []}  # Start with empty todo list
)

How It Works

1. Todos are stored in ctx.deps.todos
2. Tools return __update_context__ to update the todo list
3. AgentRunner merges updates back into context
4. Before each model request, todos are injected into system prompt
5. AI sees current progress and can self-organize

Example

# AI receives complex task
{:ok, r1} = Nous.run(agent, "Analyze codebase and create report")

# AI breaks it down:
# - Calls add_todo("Read all source files")
# - Calls add_todo("Analyze dependencies")
# - Calls add_todo("Write report")

# AI starts working:
# - Calls update_todo(id: 1, status: "in_progress")
# - Reads files...
# - Calls complete_todo(id: 1)

# System prompt automatically shows:
# ✅ Completed (1): Read all source files
# ⏳ In Progress (0):
# 📝 Pending (2): Analyze dependencies, Write report