PRODUCTION BUG 2026-04-30 11:50:
VLM returned action_bar_button_back (desc='Back') for 'tap profile tab'
intent on OTHER_PROFILE screen. This caused account_switcher to navigate
AWAY from the profile instead of opening the account selector bottom
sheet, halting the session.
Root cause: _visual_discovery pre-filter did not exclude navigation
buttons (Back, Close) when the intent was for a tab element.
Fix:
- Added filter_navigation_conflicts() method to IntentResolver
- Guards tab intents by excluding nodes with 'back' in resource_id
or content_desc == 'Back'
- Does NOT apply for 'tap back button' or 'press back' intents
TDD:
- RED: test_intent_resolver_back_guard.py — 2 tests that call
filter_navigation_conflicts on real OTHER_PROFILE XML
- GREEN: filter_navigation_conflicts method + wired into _visual_discovery
- E2E: test_workflow_account_switch.py, test_workflow_vlm_tab_confusion.py
31/31 tests passing.
🔴 RED: Tests proved obstacle_guard silently ignored system permission dialogs
and foreign app overlays, causing the bot to get stuck on Android modals.
🟢 GREEN: obstacle_guard now dismisses OBSTACLE_SYSTEM and OBSTACLE_FOREIGN_APP
with immediate back-press, preventing the interaction chain from running
against non-Instagram UI elements.
🔵 REFACTOR: Fixed broken import in resonance_evaluator (was importing
humanized_scroll from utils instead of physics.humanized_input).
Added a 'Structural Navigation Guard' in IntentResolver to map critical bottom navigation intents (e.g., 'tap profile tab') directly to their stable resource-ids. This bypasses the VLM entirely, guaranteeing 100% deterministic clicks and resolving the issue where the VLM failed to locate the profile tab, causing the edge to become masked and trapping the bot on the home feed.
Included TDD proof.
The assertion choices["goal_A"] > choices["goal_C"] fails sporadically because goal_A has a very low weight (2 vs 100) and can easily be chosen 0 times just like goal_C (0 vs 100). Changed to >= to handle valid 0 == 0 scenarios.
Also fixes "Failed to forget path" warning where _get_id was used instead of generate_uuid.
Fixes:
1. Rewrote run_competitive_benchmark.py to test BOTH capabilities:
- Telepathic (JSON element extraction)
- Brain (Free-text action extraction with format_json=False)
2. Normalizes scores by averaging per-scenario (fixes score inflation
where models with more scenarios tested scored higher but were marked unsuitable).
3. Added 4 new brain_action scenarios to ensure the 'think=false' code path
is actively benchmarked going forward.
4. Added test_benchmark_integrity.py to lock in scenario format rules.
5. Cleared stale llm_benchmarks.json data to force clean re-evaluations.
ROOT CAUSE: qwen3.5 (reasoning model) returns response='' with thinking
block containing all reasoning. llm_provider.py line 352 silently
substituted the thinking block as the response via:
content = raw_response or raw_thinking or ''
The Brain then extracted random actions from the reasoning text.
FIXES:
1. llm_provider.py: Conditional thinking isolation
- format_json=True (SAE/perception): thinking fallback preserved
- format_json=False (Brain): thinking NEVER substituted
- Added think=false for Ollama free-text calls to force direct response
2. planner.py: No-Op Guard strips tab actions that navigate to
the current screen (e.g. 'tap profile tab' on OWN_PROFILE)
3. test_brain_live.py: Stochastic testing (5 runs, 60% min valid)
to handle non-deterministic LLM behavior reliably
4. tests/integration/test_llm_provider_pipeline.py: NEW test layer
mocking at HTTP level (requests.post) to exercise the FULL
llm_provider → Brain pipeline. This would have caught the
thinking substitution bug from day one.
Suite: 168 passed, 0 failed
- Add permanent mock ban guard in root conftest.py that fails any test
importing unittest.mock at COLLECTION TIME (before execution)
- Add 8 brain output contract tests reproducing the exact production bug:
LLM thinks 'press back' but parser extracts 'tap messages tab' from
the <think> block
- Add UI change noise threshold (MIN_UI_CHANGE_BYTES=50) to prevent
false-positive 'ui_changed' from 1-byte XML diffs (timestamps/whitespace)
- Verify planner correctly strips masked actions from Brain prompt
- Added strict TDD coverage for all autonomous changes.
- Implemented GrowthBrain.get_current_goal to select high-level objectives.
- Replaced procedural orchestrator with GoalExecutor in bot_flow.
- Purged hardcoded resource-ids in dm_engine in favor of ScreenIdentity.
- Removed regex parsing in unfollow_engine in favor of telepathic semantic extraction.
- Fixed get_plugin_config AttributeError in MockConfigs and FakeConfig
- Adjusted test_carousel_zero_percent to assert on can_activate
- Explicitly delete missing mock config args in E2E tests for getattr coverage
ROOT CAUSE: The bot was systematically sabotaging its own navigation.
The HD Map planned correct 2-step routes (Home→Profile→FollowList),
but _execute_action() validated each INTERMEDIATE step against the
FINAL goal. Step 1 (tap profile tab → OWN_PROFILE) got rejected
because the goal said 'following', triggering aversive learning
that permanently burned the only valid route.
SIX COMPOUNDING FIXES:
1. SSOT Consolidation (_is_goal_achieved):
Replaced 12-line hardcoded if-chain with ScreenTopology.goal_to_target_screen().
Single source of truth for all goal→screen mappings.
2. Step-Aware Navigation Validation (_execute_action):
Replaced goal_screen_map keyword-scan with
ScreenTopology.expected_screen_for_action(action, pre_action_screen).
Now validates: 'did this step land where IT should?' not 'did I reach my goal yet?'
3. Topology Guard (aversive learning):
ScreenTopology.is_structural_action() prevents burning HD Map actions.
VLM may fail to find the element, but the route itself is structurally valid.
4. Fast-Path Threshold Fix (telepathic_engine):
100% match threshold now applies only to NAV_TAB_KEYWORDS (actual tabs),
not NAV_ZONE_BYPASS_KEYWORDS. 'tap following list' was blocked because
'following' triggered the tab threshold on a non-tab action.
5. navigate_to_screen SSOT:
Replaced 8-entry goal_map dict with ScreenTopology.screen_name_to_goal().
6. QNavGraph Name Map Dedup:
Replaced 2 inline screen_name_map/name_to_screen dicts with
ScreenTopology.SCREEN_NAME_MAP reverse lookups.
Before: 4 competing goal→screen mappings. Self-sabotage loop.
After: 1 SSOT. Step-aware validation. Topology-protected routing.
141 unit tests pass. Zero regressions.
FUNDAMENTAL ARCHITECTURE OVERHAUL:
1. ScreenTopology HD Map (NEW):
Pure-data BFS pathfinding between Instagram screens.
Zero runtime dependencies. The GOAP planner's GPS.
Knows: HOME_FEED → tap profile tab → OWN_PROFILE → tap following list → FOLLOW_LIST
2. Graph-Aware GOAP Planning:
GoalPlanner._plan_navigation() now consults ScreenTopology FIRST.
From HOME_FEED, goal 'open following list' returns 'tap profile tab'
(intermediate hop) instead of blind 'open following list' (impossible).
Autonomous discovery and Qdrant knowledge kept as fallbacks.
3. NAV Keyword Split:
NAV_INTENT_KEYWORDS → NAV_ZONE_BYPASS_KEYWORDS + NAV_TAB_KEYWORDS
Ends the Structural Guard civil war where 'following' was both
'allowed in nav zone' AND 'must be at bottom' simultaneously.
4. QNavGraph Deduplication:
_find_path() delegates to ScreenTopology.find_route() (SSOT).
core_nodes seed generated from ScreenTopology.TRANSITIONS.
115 unit tests pass. Zero regressions.
Root cause: VLM Structural Guard enforced 'must be at bottom' for ALL
NAV_INTENT_KEYWORDS including 'following'/'follower'. But these are
profile stats at Y≈246 (top of screen), not nav tabs. The inner
_structural_sanity_check correctly checks for 'tab' in intent before
enforcing bottom-zone requirement — the VLM guard was inconsistent.
Fix 1: Align VLM guard with inner guard — only enforce bottom-zone
requirement for intents explicitly containing 'tab'.
Fix 2: Add back-press circuit breaker (MAX_CONSECUTIVE_BACK=3). If GOAP
presses back 3 times on the same screen without any transition,
abort immediately to prevent exiting Instagram entirely.
95 unit tests pass.
- Extract NAV_INTENT_KEYWORDS constant (DRY) with 'direct message', 'inbox',
'dm', 'notification', 'heart icon' to fix structural guard self-sabotage
where 'tap direct message icon inbox' was rejected as non-nav intent
- Add goal-achieved pre-check in GOAP _execute_recalled_path to skip
stale paths when the bot is already on the target screen
- Add already-there detection in _execute_action to prevent false unlearning
when navigation produces no UI change because goal is already met
- Implement 3-tier ad escape cascade: normal skip -> double scroll -> GOAP
force-navigate to HomeFeed after 6+ consecutive ad cycles
- 92 unit tests pass, 223/224 integration tests pass (1 pre-existing flaky)
- Fixed a bug where `humanized_scroll(is_skip=True)` could trigger random 'Doomscroll Corrections' (scrolling backwards) or 'reading pauses'.
- This caused the Anti-Stuck loop for ads to fail because the aggressive skip would occasionally just pause or scroll back into the ad.
- Added strict TDD coverage in `test_physics_humanized.py` to verify `is_skip` generates strictly forward gestures without pauses.
- Hardened `NavigationKnowledge` and `TelepathicEngine` to prevent premature blacklisting on high-latency grid taps by returning `None` (inconclusive) when grid markers are still present.
- Updated `_execute_action` in `goap.py` to respect inconclusive verification states and avoid polluting the blacklist.
- Refined `Adaptive Snap` in `timing.py` to detect `ProgressBar` loading spinners, extending timeouts for slow connections.
- Prevented brittle Adaptive Snap `wobble` routines from firing while the bot is still trapped on the Explore Grid.
- Stabilized Qdrant collections by converting destructive delete-only wipes into safe `wipe_collection` routines that instantly recreate the index.
- Maintained 100% TDD pass-rate by aligning E2E grid navigation tests with the new inconclusive states.
- implemented self-healing unlearn for Qdrant false positives
- centralized testing logic in conftest
- documented core rules, ai standards, and goap philosophy
- purged old dev scratchpads
- Separated obstacle detection from feed marker validation
- Prevented blind BACK button presses when markers are missing mid-scroll
- Added TDD verification for feed navigation stability
- Cleaned up debug artifacts and temporary test output
Summary of work:
- Resolved mass SystemExit: 2 failures by hardening Config against pytest CLI args.
- Fixed state leakage in test suite by implementing aggressive cache wiping in conftest.py.
- Fixed TypeErrors and UnboundLocalErrors in TelepathicEngine and bot_flow.
- Aligned MockTelepathicEngine signatures to resolve Mock Drift.
- Achieved 100% pass rate across 498 tests.
