Compare commits
2 Commits
5fbbe3d273
...
6cd068f951
| Author | SHA1 | Date | |
|---|---|---|---|
| 6cd068f951 | |||
| f46b0b7bcb |
@@ -496,8 +496,12 @@ class GoalExecutor:
|
||||
return False
|
||||
else:
|
||||
# action_success is None (INCONCLUSIVE)
|
||||
# We decay the memory so it unlearns if it repeatedly fails to produce a definitive success.
|
||||
logger.warning(f"⚠️ [GOAP Execute] Applying AGGRESSIVE PENALTY for inconclusive action '{action}'.")
|
||||
engine.decay_click(action)
|
||||
# Double penalty to burn ambiguous paths faster (outer loop adds +1, so total +2 = instantly hits MAX_RETRIES)
|
||||
self.action_failures[(pre_action_screen_type, action)] = (
|
||||
self.action_failures.get((pre_action_screen_type, action), 0) + 1
|
||||
)
|
||||
return False
|
||||
|
||||
def _execute_recalled_path(self, steps: List[Dict], goal: str) -> bool:
|
||||
|
||||
@@ -156,17 +156,8 @@ class GoalPlanner:
|
||||
)
|
||||
return None
|
||||
|
||||
# ── 2. Brain-Driven Decision Making (Primary Strategy) ──
|
||||
# The user explicitly wants the AI to be the primary driver of goals.
|
||||
from GramAddict.core.navigation.brain import ask_brain_for_action
|
||||
|
||||
brain_action = ask_brain_for_action(goal, screen_type.name, available, avoid_actions)
|
||||
if brain_action:
|
||||
logger.info(f"🧠 [Brain] Decided to execute: '{brain_action}' (to achieve: '{goal}')")
|
||||
return brain_action
|
||||
|
||||
# ── 2. HD Map Routing (Fallback) ──
|
||||
# If the Brain doesn't know what to do, try the deterministic topological map.
|
||||
# ── 2. HD Map Routing (Primary Strategy for Navigation) ──
|
||||
# Ground UI transitions in structural invariants. If the topological map knows the route, use it.
|
||||
target_screen = ScreenTopology.goal_to_target_screen(goal)
|
||||
if target_screen and target_screen != screen_type:
|
||||
route = ScreenTopology.find_route(screen_type, target_screen, avoid_actions=avoid_actions)
|
||||
@@ -187,6 +178,15 @@ class GoalPlanner:
|
||||
f"🛡️ [HD Map] Route action '{next_action}' already explored and failed. Skipping HD Map."
|
||||
)
|
||||
|
||||
# ── 3. Brain-Driven Decision Making (Fallback / Discovery) ──
|
||||
# For non-navigation goals or when the HD Map is incomplete.
|
||||
from GramAddict.core.navigation.brain import ask_brain_for_action
|
||||
|
||||
brain_action = ask_brain_for_action(goal, screen_type.name, available, avoid_actions)
|
||||
if brain_action:
|
||||
logger.info(f"🧠 [Brain] Decided to execute: '{brain_action}' (to achieve: '{goal}')")
|
||||
return brain_action
|
||||
|
||||
# ── 2. Learned Knowledge (Qdrant) ──
|
||||
required_screens = self.knowledge.get_requirements(goal)
|
||||
|
||||
|
||||
@@ -167,7 +167,8 @@ class ActionMemory:
|
||||
and "row_feed_button_like" not in post_xml_lower
|
||||
and "clips_viewer" not in post_xml_lower
|
||||
):
|
||||
return None # Still on grid, inconclusive
|
||||
logger.warning(f"⚠️ [ActionMemory] Still on grid after trying to '{intent}'. Verification FAIL.")
|
||||
return False # Still on grid, definitely failed
|
||||
|
||||
state_toggles = ["like", "save", "follow", "heart"]
|
||||
is_toggle = any(t in intent_lower for t in state_toggles)
|
||||
@@ -269,7 +270,10 @@ class ActionMemory:
|
||||
if diff > 0:
|
||||
logger.debug(f"🧠 [ActionMemory] Structural delta detected for toggle '{intent}'. Verification PASS.")
|
||||
return True
|
||||
else:
|
||||
logger.warning(
|
||||
f"⚠️ [ActionMemory] Zero structural shift (diff={diff}) for state-toggle '{intent}'. Verification FAIL."
|
||||
)
|
||||
return False
|
||||
# If the intent is an abstract goal (like "find customers"), diff > 50 is NOT enough.
|
||||
# We must force visual VLM confirmation because clicking the wrong thing (like "Create highlight")
|
||||
# also produces a large diff but achieves the wrong goal.
|
||||
@@ -317,6 +321,12 @@ class ActionMemory:
|
||||
logger.warning(f"⚠️ [ActionMemory] Cannot visually verify abstract intent '{intent}'. Failing safe.")
|
||||
return False
|
||||
|
||||
# If diff <= 50 for non-toggle
|
||||
logger.warning(
|
||||
f"⚠️ [ActionMemory] Insufficient structural change (diff={diff}) for non-toggle '{intent}'. Verification FAIL."
|
||||
)
|
||||
return False
|
||||
|
||||
|
||||
def _intent_matches_node(intent: str, semantic_string: str) -> bool:
|
||||
"""Checks if the clicked element semantically matches the toggle intent.
|
||||
|
||||
@@ -62,7 +62,6 @@ class ScreenTopology:
|
||||
"press back": ScreenType.HOME_FEED,
|
||||
},
|
||||
ScreenType.OTHER_PROFILE: {
|
||||
"press back": ScreenType.HOME_FEED,
|
||||
"tap home tab": ScreenType.HOME_FEED,
|
||||
},
|
||||
ScreenType.UNKNOWN: {
|
||||
|
||||
@@ -72,28 +72,26 @@ def test_curiosity_navigates_to_homefeed_before_checking(
|
||||
nav_graph = cognitive_stack["nav_graph"]
|
||||
zero_engine = cognitive_stack["zero_engine"]
|
||||
dopamine = cognitive_stack["dopamine"]
|
||||
growth = cognitive_stack["growth_brain"]
|
||||
|
||||
# Force CHECK_CURIOSITY
|
||||
monkeypatch.setattr(growth, "evaluate_governance", lambda *args, **kwargs: "CHECK_CURIOSITY")
|
||||
|
||||
# We want it to exit cleanly after executing a few loops, without hanging forever.
|
||||
dopamine.session_limit_seconds = 0.1
|
||||
dopamine.session_start = time.time()
|
||||
|
||||
# 5. Run the loop (starting from other_profile)
|
||||
try:
|
||||
_run_zero_latency_feed_loop(
|
||||
device=device,
|
||||
zero_engine=zero_engine,
|
||||
nav_graph=nav_graph,
|
||||
configs=configs,
|
||||
session_state=session_state,
|
||||
job_target="homefeed",
|
||||
cognitive_stack=cognitive_stack,
|
||||
)
|
||||
except Exception:
|
||||
# It will likely crash when trying to 'tap heart icon notifications'
|
||||
# because the emulator state machine doesn't define what that button does.
|
||||
# This is expected and perfectly fine, because it proves the bot TRIED to
|
||||
# tap it AFTER navigating.
|
||||
pass
|
||||
# 5. Run the loop (starting from other_profile)
|
||||
_run_zero_latency_feed_loop(
|
||||
device=device,
|
||||
zero_engine=zero_engine,
|
||||
nav_graph=nav_graph,
|
||||
configs=configs,
|
||||
session_state=session_state,
|
||||
job_target="homefeed",
|
||||
cognitive_stack=cognitive_stack,
|
||||
)
|
||||
|
||||
# 6. Assertions: Must navigate to HomeFeed FIRST
|
||||
assert emulator.current_state == "home_feed", (
|
||||
|
||||
@@ -54,9 +54,9 @@ def test_goap_planner_avoids_infinite_loop_on_masked_edge():
|
||||
action_failures=action_failures,
|
||||
)
|
||||
|
||||
# The HD Map should fail, and because the planner is trapped, it forces a restart
|
||||
# The HD Map should fail, and because the planner is trapped, it falls back to back-tracking
|
||||
assert (
|
||||
action_avoided == "force start instagram"
|
||||
action_avoided == "press back" or action_avoided == "force start instagram"
|
||||
), "Planner routed BLIND into the dead end despite the edge being masked!"
|
||||
|
||||
|
||||
|
||||
@@ -855,14 +855,13 @@ class TestVerifySuccessStructuralDelta:
|
||||
result = memory.verify_success("like", pre_click_xml=pre_xml, post_click_xml=post_xml)
|
||||
assert result is True
|
||||
|
||||
def test_toggle_zero_diff_is_none(self):
|
||||
"""Zero XML change for a toggle = inconclusive (not confirmed)."""
|
||||
def test_toggle_zero_diff_is_false(self):
|
||||
"""Zero XML change for a toggle = fail."""
|
||||
memory, _ = self._make_memory()
|
||||
same_xml = '<node text="Like" />'
|
||||
|
||||
result = memory.verify_success("like", pre_click_xml=same_xml, post_click_xml=same_xml)
|
||||
# Zero diff, no markers → None (inconclusive) not True
|
||||
assert result is None or result is False
|
||||
assert result is False
|
||||
|
||||
def test_follow_success_with_requested_marker(self):
|
||||
"""
|
||||
@@ -888,7 +887,7 @@ class TestVerifySuccessStructuralDelta:
|
||||
result = memory.verify_success("view a post", pre_click_xml="<node/>", post_click_xml=post_xml)
|
||||
assert result is True
|
||||
|
||||
def test_view_post_still_on_grid_is_inconclusive(self):
|
||||
def test_view_post_still_on_grid_is_false(self):
|
||||
"""
|
||||
If after 'view a post' the XML still shows explore_action_bar
|
||||
WITHOUT post detail markers, navigation failed.
|
||||
@@ -896,7 +895,7 @@ class TestVerifySuccessStructuralDelta:
|
||||
memory, _ = self._make_memory()
|
||||
post_xml = '<node resource-id="com.instagram.android:id/explore_action_bar" />'
|
||||
result = memory.verify_success("grid item", pre_click_xml="<node/>", post_click_xml=post_xml)
|
||||
assert result is None # Inconclusive
|
||||
assert result is False
|
||||
|
||||
def test_semantic_gate_blocks_wrong_toggle_element(self):
|
||||
"""
|
||||
|
||||
@@ -92,8 +92,7 @@ class TestBrainOutputParsing:
|
||||
explored_actions=set(),
|
||||
)
|
||||
assert result == "tap explore tab", (
|
||||
f"Brain extracted '{result}' instead of 'tap explore tab'. "
|
||||
f"Expected the last-mentioned action to win."
|
||||
f"Brain extracted '{result}' instead of 'tap explore tab'. " f"Expected the last-mentioned action to win."
|
||||
)
|
||||
|
||||
def test_brain_never_returns_avoided_action(self, monkeypatch):
|
||||
@@ -116,8 +115,7 @@ class TestBrainOutputParsing:
|
||||
)
|
||||
# The action MUST be None or one of the available actions — NEVER the masked one
|
||||
assert result != "tap messages tab", (
|
||||
"Brain returned an action that was not in available_actions! "
|
||||
"This means the masking layer has a hole."
|
||||
"Brain returned an action that was not in available_actions! " "This means the masking layer has a hole."
|
||||
)
|
||||
|
||||
|
||||
@@ -146,7 +144,7 @@ class TestBrainAvoidActionsParity:
|
||||
}
|
||||
|
||||
planner.plan_next_step(
|
||||
"open explore",
|
||||
"nurture community",
|
||||
screen,
|
||||
action_failures={"tap messages tab": 2}, # Masked!
|
||||
)
|
||||
@@ -159,8 +157,7 @@ class TestBrainAvoidActionsParity:
|
||||
if "available to you right now" in line:
|
||||
# The masked action must NOT be in the available actions list
|
||||
assert "tap messages tab" not in line, (
|
||||
f"Planner passed masked action 'tap messages tab' to the Brain as available!\n"
|
||||
f"Line: {line}"
|
||||
f"Planner passed masked action 'tap messages tab' to the Brain as available!\n" f"Line: {line}"
|
||||
)
|
||||
break
|
||||
else:
|
||||
@@ -254,10 +251,7 @@ class TestBrainEmptyResponse:
|
||||
available_actions=["tap explore tab", "scroll down"],
|
||||
explored_actions=set(),
|
||||
)
|
||||
assert result is None, (
|
||||
f"Brain returned '{result}' from a whitespace-only response! "
|
||||
f"Must return None."
|
||||
)
|
||||
assert result is None, f"Brain returned '{result}' from a whitespace-only response! " f"Must return None."
|
||||
|
||||
|
||||
class TestPlannerNoOpGuard:
|
||||
@@ -293,8 +287,7 @@ class TestPlannerNoOpGuard:
|
||||
for line in prompt.splitlines():
|
||||
if "available to you right now" in line:
|
||||
assert "tap profile tab" not in line, (
|
||||
f"Planner passed no-op action 'tap profile tab' to Brain on OWN_PROFILE!\n"
|
||||
f"Line: {line}"
|
||||
f"Planner passed no-op action 'tap profile tab' to Brain on OWN_PROFILE!\n" f"Line: {line}"
|
||||
)
|
||||
break
|
||||
else:
|
||||
@@ -328,8 +321,7 @@ class TestPlannerNoOpGuard:
|
||||
for line in prompt.splitlines():
|
||||
if "available to you right now" in line:
|
||||
assert "tap home tab" not in line, (
|
||||
f"Planner passed no-op 'tap home tab' to Brain on HOME_FEED!\n"
|
||||
f"Line: {line}"
|
||||
f"Planner passed no-op 'tap home tab' to Brain on HOME_FEED!\n" f"Line: {line}"
|
||||
)
|
||||
break
|
||||
else:
|
||||
|
||||
@@ -58,7 +58,7 @@ class TestVerifySuccessGridReels:
|
||||
</hierarchy>
|
||||
"""
|
||||
result = self.engine.verify_success("view a post", explore_xml)
|
||||
assert result is None, "verify_success should return None (inconclusive) when grid is still visible"
|
||||
assert result is False, "verify_success should return False when grid is still visible"
|
||||
|
||||
def test_profile_grid_reel_accepted(self):
|
||||
"""Profile grid → Reel must also be accepted."""
|
||||
|
||||
Reference in New Issue
Block a user