feat(android): add native controls and continuous multi-touch

tests/unit/agents/test_operator_transcript.py

@@ -35,6 +35,7 @@ from langchain_core.messages import AIMessage, HumanMessage, SystemMessage, Tool
 from artemis.agents.operator.operator import OperatorNode
 from artemis.agents.operator.prompts import load_operator_prompts
 from artemis.agents.operator.prompts import (
+    OPERATOR_PROMPT_TOOLSET,
     PLAN_HISTORY_STATIC_POINTER,
     PLAN_HISTORY_TEMPLATE_SECTION,
     PromptBuilder,
@@ -54,8 +55,8 @@ from artemis.memory.transcript import (
 
 # SHA-256 snapshots of the legacy system message with the fixed inputs below.
 # Update these when an intentional template change alters the rendered prompt.
-GOLDEN_EMPTY_PLAN = "23c3d6b5fc3694275e7015f7bd0901bde54aae54fec7a8c68d4f8529c65ef4bc"
-GOLDEN_SENTINEL_PLAN = "370c53f1d088a4120c63fbaea6d895f121c683b8d3c8cc7d61105d6e1d223128"
+GOLDEN_EMPTY_PLAN = "f5ee50f4923fed4cb943cb80087d39ffffe66f259cfae4a54f27a244d7b06453"
+GOLDEN_SENTINEL_PLAN = "bb725e2d096f6aa0cf1773809b2d0e672475589782b9c72348d9eaf8481230b9"
 
 SCREENSHOT_B64 = base64.b64encode(b"fake-jpeg-bytes").decode("utf-8")
 
@@ -84,6 +85,42 @@ async def test_legacy_prompt_bytes_match_chinese_prompt_golden():
     assert hashlib.sha256(sentinel.encode("utf-8")).hexdigest() == GOLDEN_SENTINEL_PLAN
 
 
+@pytest.mark.asyncio
+@pytest.mark.parametrize("transcript", [False, True])
+async def test_legacy_and_transcript_prompts_omit_all_unavailable_native_capabilities(transcript):
+    native_tools = {
+        "pinch",
+        "touch_gesture",
+        "read_control",
+        "control_action",
+        "wait_until",
+        "scroll_to",
+    }
+    available_tools = OPERATOR_PROMPT_TOOLSET - native_tools
+    ctx = SimpleNamespace(
+        execution_setup=None,
+        actuator=SimpleNamespace(capabilities=lambda: available_tools),
+        data_engine=None,
+    )
+    state = SimpleNamespace(initial_goal="FILTERED_FIXED_GOAL")
+    if transcript:
+        system = render_transcript_static_system(load_operator_prompts(), ctx, state)
+    else:
+        builder = PromptBuilder()
+        await TemplatePromptComponent()(
+            builder,
+            state,
+            ctx,
+            prompts=load_operator_prompts(),
+            template_name="main_template",
+        )
+        system = "".join(builder.system_parts)
+    assert "FILTERED_FIXED_GOAL" in system
+    assert "`click`" in system and "`drag`" in system
+    for tool in native_tools:
+        assert tool not in system
+
+
 def test_main_template_contains_exactly_one_plan_history_section():
     template = load_operator_prompts()["main_template"]
     assert template.count(PLAN_HISTORY_TEMPLATE_SECTION) == 1