Enhanced Input

🧠 Brains complement: IF an unreal-engine-skills-manager tool (external MCP) exists in this session, call it with {action: "load", skill: "enhanced-input"} for UE domain knowledge on this topic — correct APIs, architecture, best practices — and treat it as the rubric for any review / "best practices" question. If no such tool is available (e.g. running under Claude Code or Codex without that MCP), skip this line entirely and proceed with this skill alone — do NOT attempt the call.

Enhanced Input Skill

Critical Rules

🚨 Real Python method names — do NOT search for the docstring's "action" names

The InputService class docstring lists MCP-style action ids (action_create, mapping_list_contexts, action_configure, ...). Those are not the Python method names — discover_python_class(method_filter=...) matches nothing for them. The complete real API is below; you rarely need a discovery call at all:

create_action value types are the strings discover_types() returns: "Boolean" (alias "Digital"), "Axis1D", "Axis2D", "Axis3D". The create_* folder may be a bare folder ("Input" goes under /Game) or start with its mount point (/Game/..., a plugin root); /Engine, /Script and /Temp are refused, and so is an invalid asset name (empty, a space, .), each with the reason in error_message. Modifiers and triggers are addressed by index on the mapping, so get_mappings / get_modifiers / get_triggers first, then add/remove by index.

⚠️ Property Names on Info Structs

⚠️ Value Types

⚠️ Key Names

Keyboard: SpaceBar, LeftShift, W, A, S, D, F1, Enter, Escape, BackSpace...
Mouse: LeftMouseButton, RightMouseButton, MouseScrollUp
Gamepad: Gamepad_FaceButton_Bottom, Gamepad_LeftThumbstick, Gamepad_LeftTrigger, Gamepad_RightTrigger
Paired axes (use these for Axis2D bindings): Mouse2D (mouse look), Gamepad_Left2D / Gamepad_Right2D (analog sticks)

Verify any other name with get_available_keys("thumb") (substring filter) instead of guessing.

⚠️ Triggers and Modifiers

Triggers: Pressed, Released, Down, Hold, Tap, Pulse
Modifiers: Negate, DeadZone, Scalar, SwizzleAxis (also: Smooth, SmoothDelta, ScaleByDeltaTime — use get_available_modifier_types() for the full set)

Workflows

Create Input Action

pythonCopy code
import unreal

result = unreal.InputService.create_action("Jump", "/Game/Input", "Boolean")
if result.success:
    unreal.EditorAssetLibrary.save_asset(result.asset_path)

Create Mapping Context

pythonCopy code
import unreal

result = unreal.InputService.create_mapping_context("Default", "/Game/Input", 0)
unreal.EditorAssetLibrary.save_asset(result.asset_path)

Add Key Mapping

pythonCopy code
import unreal

context_path = "/Game/Input/IMC_Default"
action_path = "/Game/Input/IA_Jump"
unreal.InputService.add_key_mapping(context_path, action_path, "SpaceBar")
unreal.EditorAssetLibrary.save_asset(context_path)

Add Triggers and Modifiers

pythonCopy code
import unreal

context_path = "/Game/Input/IMC_Default"
action_path = "/Game/Input/IA_Fire"

unreal.InputService.add_key_mapping(context_path, action_path, "LeftMouseButton")

# Get mapping index
mappings = unreal.InputService.get_mappings(context_path)
mapping_index = len(mappings) - 1

# Add trigger/modifier using mapping index
unreal.InputService.add_trigger(context_path, mapping_index, "Pressed")
unreal.InputService.add_modifier(context_path, mapping_index, "DeadZone")
unreal.EditorAssetLibrary.save_asset(context_path)

Trigger Settings

properties_json sets the new trigger's settings, by C++ name or snake_case (HoldTimeThreshold or hold_time_threshold, bIsOneShot or is_one_shot). Only the settings the trigger's Details panel shows can be set: runtime state such as HeldDuration, LastValue or bShouldAlwaysTick is refused ("not a trigger setting"), and a number outside the property's ClampMin/ClampMax is refused with the allowed range (hold_time_threshold must be >= 0). An unknown name, a refused setting or a bad value fails the call and adds nothing. add_action_trigger puts the trigger on the Input Action itself, so it applies to every mapping of the action; it returns PIE_ACTIVE while a play session runs (stop PIE first). Neither saves the asset, and a running PIE session keeps its copy of the triggers: restart PIE to see a change.

pythonCopy code
import unreal

# Hold 0.4 s before the charged attack fires, once per hold
unreal.InputService.add_trigger("/Game/Input/IMC_Default", 0, "Hold",
    '{"hold_time_threshold": 0.4, "is_one_shot": true}')

# A quick tap on the action itself, for every key bound to it
result = unreal.InputService.add_action_trigger("/Game/Input/IA_Dash", "Tap",
    '{"tap_release_time_threshold": 0.25}')
print(result)   # {"success": true, "trigger_index": 0, ...} or an error_code
unreal.EditorAssetLibrary.save_asset("/Game/Input/IA_Dash")

Get Mappings Info

pythonCopy code
import unreal

mappings = unreal.InputService.get_mappings("/Game/Input/IMC_Default")
for m in mappings:
    print(f"Action: {m.action_name}, Key: {m.key_name}")

info = unreal.InputService.get_input_action_info("/Game/Input/IA_Jump")
if info:
    print(f"Action: {info.action_name}, ValueType: {info.value_type}")

Discover Available Types

pythonCopy code
import unreal

types = unreal.InputService.discover_types()
print(f"Value Types: {types.action_value_types}")
print(f"Modifiers: {types.modifier_types}")
print(f"Triggers: {types.trigger_types}")

keys = unreal.InputService.get_available_keys()

Sample scripts (run via execute_python_code)

• scripts/setup_input.txt — create an Input Action + Mapping Context and bind a key.

PIE input injection (issue #550)

Test input-driven gameplay WITHOUT remapping game assets or OS-level SendKeys — inject directly into the running PIE session (no OS window focus needed):

pythonCopy code
import unreal

# Enhanced Input action, QUEUED for the next input tick and released on the one after.
# X/Y/Z map to the value type. Read the result in a later call: nothing ticks while this one runs.
print(unreal.InputService.inject_action("/Game/Input/IA_Fire", 1.0))

# Hold an action for 1.5 s of real time (a guard, a charge, a Hold trigger); returns at once.
print(unreal.InputService.inject_action_for("/Game/Input/IA_Block", 1.5))
print(unreal.InputService.stop_injection("/Game/Input/IA_Block"))   # release early

# Raw key through Slate to the game viewport: "tap" (default), "down", "up", or "hold".
print(unreal.InputService.inject_key("SpaceBar"))
print(unreal.InputService.inject_key("RightMouseButton", "hold", 1.5))

All return JSON with success/error_code (PIE not running, NO_LOCAL_PLAYER / NO_PLAYER_CONTROLLER while PIE or a client's join is still starting — retry on a later call —, unknown key, BAD_DURATION outside 0.05 to 60 s). inject_action reports queued: true: it was queued, not yet applied. The action calls take pie_instance to pick a PIE world: -1, the default, is the first PIE world with a local player (the lowest instance number), and replies report the instance actually used. A hold is one hold per action and PIE world whatever the path spelling, so stop_injection finds it either way; holding a held key again extends it (extended: true); ending PIE drops every hold. See the pie-testing skill for the full verification loop.