Feature: Agent Memory Lock and Operator Error Propagation
  As a developer
  I want agents to handle event loop edge cases and propagate operator errors correctly
  So that agent memory locks and streaming operators are robust under concurrent and error conditions

  Background:
    Given a fresh agents base final test context (abf)

  Scenario: AgentWithMemory _memory_lock handles RuntimeError from get_event_loop
    Given a fresh agents base final test context (abf)
    When I access _memory_lock from a thread with no event loop (abf)
    Then a new event loop should be created by _memory_lock (abf)

  Scenario: AgentWithMemory _memory_lock creates new loop when loop is closed
    Given a fresh agents base final test context (abf)
    When I access _memory_lock after closing the current event loop (abf)
    Then _memory_lock should create and set a new event loop (abf)

  Scenario: AgentWithMemory dispose clears lock and calls parent dispose
    Given a fresh agents base final test context (abf)
    When I dispose an AgentWithMemory instance (abf)
    Then _memory_lock_instance should be set to None (abf)
    And parent dispose should be called (abf)

  Scenario: StreamableAgent as_operator propagates on_error
    Given a fresh agents base final test context (abf)
    When I use as_operator with an error-producing source (abf)
    Then the error should propagate through on_error (abf)

  Scenario: StreamableAgent map_operator processes value through agent
    Given a fresh agents base final test context (abf)
    When I use map_operator in a pipeline with a value (abf)
    Then the value should be processed and result set on the future (abf)

  Scenario: StreamableAgent filter_operator applies condition to output
    Given a fresh agents base final test context (abf)
    When I use filter_operator with a condition and value (abf)
    Then the condition should be applied to agent output (abf)