feat: add adaptive Pest Control strategy

[JogOnJohn]

Summary

This overhauls Pest Control from a configurable NPC-priority loop into an adaptive, portal-first strategy designed to contribute aggressively to fast mass-world wins.

The plugin now treats vulnerable portals as the main objective, preserves activity through deliberate fallback combat, handles Pest Control's gates and Brawler obstructions, supports ordered multi-style loadouts, and exposes its runtime decisions through a substantially expanded overlay.

Strategy and state machine

• Replaces the old NPC Priority 1/2/3 configuration with explicit runtime states for initialization, travel, requeue, boat waiting, opening movement, portal pre-positioning, portal pursuit, Spinner combat, activity recovery, gate handling, Brawler avoidance, and error recovery.
• Confirms Auto Retaliate is disabled at startup and every round entry.
• Prioritizes active portal pressure, killing a Spinner healing the selected portal before resuming the portal attack.
• Uses a 40% activity-recovery trigger and continues recovery toward 70%.
• Activity recovery prefers a nearby Spinner, then a nearby Torcher for ranged-friendly recovery, then the lowest-level ordinary non-Brawler pest.
• Excludes Brawlers from routine activity fallback; they are attacked only when they physically prevent movement and flanking cannot clear the route.
• Uses a randomized opening target with configurable Purple, Blue, and Yellow weights. Red is intentionally excluded as an opening side.
• Pre-positions at the remaining shielded portal when the round reaches a sole pending shield.

Adaptive portal targeting

• Tracks shield drops from Pest Control game messages.
• Independently verifies vulnerability from each portal NPC's Attack action, allowing recovery from missed chat messages or mid-round startup.
• Counts nearby players and joins an active portal group instead of selecting solely by distance.
• Retains a live target to avoid oscillation between portals as crowd counts fluctuate.
• Uses Purple as the tie-break preference.

Movement, gates, and obstruction recovery

• Favors direct NPC interaction when the target is visible and interactable.
• Uses short, non-blocking canvas/minimap movement steps rather than blocking full-map walks during rounds.
• Stages ranged and magic loadouts in front of portals at normal attack distance instead of repeatedly forcing the player behind them.
• Steers the camera once toward newly selected portal objectives.
• Routes east/west travel around the outer perimeter where appropriate.
• Detects and opens intact lane gates, commits through confirmed openings, and suppresses repeated gate clicks.
• Recognizes the totally destroyed Pest Control gate variant as passable even when its composition still advertises Open.
• Detects Brawlers intersecting movement routes, attempts reachable compass-direction flanks, and attacks the Brawler as a deterministic fallback.
• Includes a six-second movement/interaction watchdog that clears stale walking state and issues fresh commands based on observed progress.

Combat loadouts

• Adds ordered Style 1, Style 2, and Style 3 configuration for Ranged, Magic, and Melee.
• Keeps Style 1 mandatory and uses it as the fallback when an optional portal-specific weapon is blank or None.
• Treats a blank or None off-hand as an explicitly empty shield slot.
• Forces Ranged weapons onto Rapid and verifies the resulting combat mode.
• Supports powered-staff Magic or one-time startup autocast preparation.
• Supports Yellow portal Stab/Slash selection and a separate Red portal Crush loadout.
• Switches the matching Void helmet automatically when the complete enabled-style helmet set is available; otherwise leaves the head slot untouched for non-Void setups.
• Verifies weapon, off-hand, helmet, and combat mode before dispatching the final attack.
• Remembers verified weapon/style pairs so routine weapon switches do not repeatedly open the Combat interface.
• Adds independent portal-specific special-attack toggles.
• Restores Style 1 after a round once boat entry has been confirmed.

Round lifecycle and accounting

• Enables Moderate activity intensity when the plugin starts.
• Handles Quick Prayer once per round without blocking round movement.
• Clears stale walking routes immediately when a round ends.
• Prioritizes immediate reboarding and guards the gangplank interaction for three seconds to prevent duplicate boarding clicks.
• Preserves fast winning-team requeues where the next launch can occur within five seconds.
• Reconciles played, won, and lost counters so played = won + lost.
• Uses reward chat or a positive commendation-point delta as authoritative win evidence.
• Deduplicates point-award messages and preserves delayed round evidence across the boat transition.
• Tracks session points from award messages and the boat's total-points overlay without double-counting.

Diagnostics and configuration

• Publishes every runtime state through Microbot.status and logs meaningful state transitions.
• Expands the overlay with location, state detail, activity thresholds, target portal/crowd/source, portal counts, current weapon/style, Auto Retaliate state, boarding state, W/L/points, state age, progress age, and command age.
• Reorganizes the configuration panel into General, Combat Styles, Ranged, Magic, Melee, Opening Strategy, and Special Attack sections.
• Updates the public plugin documentation for the new strategy and loadout model.
• Bumps Pest Control from 2.3.4 to 2.5.7 and requires Microbot 2.6.15 or newer.

Compatibility note

Melee mode detection matches visible Combat-tab option text. The intended Rapid, Dragon scimitar Slash, and DDS Stab options use matching names. A configured crush weapon whose button is named Pummel or Smash rather than Crush should be live-verified before relying on that switch.

Validation

• PestControlPluginJar packaged successfully against Microbot 2.6.15.
• The packaged 2.5.6 JAR was installed and smoke-tested; the final 2.5.7 compatibility bump packaged successfully against the same client baseline.
• An operator smoke test completed one full round successfully on Microbot 2.6.16.
• Focused regression coverage was added for combat-plan resolution, opening selection, loadout fallback, Void helmet requirements, reward-backed accounting, counter reconciliation, point deduplication, and destroyed-gate recognition.

[chsami]

I validated that PestControlPluginJar compiles against Microbot 2.6.19. I’m holding this PR because the remaining work is substantial:

1. Hub feature work should target development, not main; please rebase onto current development and retarget the PR.
2. PestControlCombatPlanTest is a class with a main() method and no JUnit @Test methods. As written, ./gradlew test completes without executing these assertions. Please convert it to JUnit tests (or wire an explicit verification task into check) and confirm the test report contains the cases.
3. This changes +4,588/-419 lines, including a ~3,500-line runtime script and round accounting/requeue behavior. Please add live multi-round evidence covering reboarding, delayed reward/point messages, activity recovery, and at least one loadout switch. The stated single-round smoke test is a good start, but it does not exercise most of the new lifecycle.
4. The PR itself notes that crush weapons whose option is named Pummel/Smash may not match. Please either handle those option labels or clearly constrain/reject incompatible configured weapons before merge.

Once the base, executable tests, and runtime coverage are addressed, this can receive a meaningful final review.

[JogOnJohn]

I validated that PestControlPluginJar compiles against Microbot 2.6.19. I’m holding this PR because the remaining work is substantial:

1. Hub feature work should target development, not main; please rebase onto current development and retarget the PR.
2. PestControlCombatPlanTest is a class with a main() method and no JUnit @Test methods. As written, ./gradlew test completes without executing these assertions. Please convert it to JUnit tests (or wire an explicit verification task into check) and confirm the test report contains the cases.
3. This changes +4,588/-419 lines, including a ~3,500-line runtime script and round accounting/requeue behavior. Please add live multi-round evidence covering reboarding, delayed reward/point messages, activity recovery, and at least one loadout switch. The stated single-round smoke test is a good start, but it does not exercise most of the new lifecycle.
4. The PR itself notes that crush weapons whose option is named Pummel/Smash may not match. Please either handle those option labels or clearly constrain/reject incompatible configured weapons before merge.

Once the base, executable tests, and runtime coverage are addressed, this can receive a meaningful final review.

Ill get to this eventually. I dont need pest control anymore. The plugin served it purpose for me. Ill try run an hour of it with these points addressed.

[chsami]

The previously reported blockers are unchanged. This still targets main instead of development; PestControlCombatPlanTest is still a main-method harness rather than JUnit tests executed by Gradle; the multi-thousand-line runtime rewrite still only reports a single-round smoke test; and the stated Pummel/Smash crush-label incompatibility remains. Please retarget and rebase on development, make the regression suite execute under Gradle test, cover multi-round reboarding/accounting/activity/loadout behavior live, and handle or explicitly reject incompatible crush weapons.

[chsami]

Requested changes summary: please retarget and rebase this PR onto development, convert PestControlCombatPlanTest into tests that execute under Gradle test, and add multi-round live evidence for reboarding, delayed reward accounting, activity recovery, and a loadout switch. Please also handle or explicitly reject crush weapons whose combat option is Pummel or Smash rather than Crush.

[JogOnJohn]

@chsami @coderabbitai

All requested blockers have been actioned; requesting a fresh review.

• Rebased the branch on upstream/development (64d1d0f) and retargeted this PR to development.
• Converted PestControlCombatPlanTest from its main() harness to JUnit Jupiter, wired useJUnitPlatform(), and verified the Gradle XML report: 19 tests, 0 failures, 0 errors using gradlew.bat test PestControlPluginJar -PpluginList=PestControlPlugin --console=plain.
• Crush-mode matching now accepts the visible Crush, Pummel, and Smash labels. It retains exact matching for Stab, Slash, and Rapid.
• Added the single-melee-entry fallback: either Yellow or Red can be configured alone; the other portal reuses it. With neither configured, startup emits the explicit missing-melee validation.

Live validation evidence

A 30-minute Bizza development hot-reload session completed 10 fully accounted rounds with the updated PestControlPlugin-2.5.7.jar:

Pest Control session: 10 played, 7 won, 3 lost, 28 points gained
Pest Control round ended: 11
Pest Control round accounted: 10
Pest Control activity recovery started: 6
Pest Control activity recovered: 4
Pest Control watchdog recovery: 3
Pest Control ERROR: 0
Pest Control Exception: 0

The captured log also shows successful hot reload, a Dual macuahuitl Crush selection, full three-style Void recognition, portal transitions including Red/Yellow/Blue, result accounting after requeue, and activity recovery:

HotReloadService - Reloaded 1 plugin(s) from PestControlPlugin-2.5.7.jar
Pest Control attack style confirmed: Crush (Dual macuahuitl)
Pest Control Void helmet switching enabled for [MELEE, RANGED, MAGIC]
Pest Control round accounted: WON (... source POINT_DELTA, points +4 ...)

Packaged artifact: PestControlPlugin-2.5.7-ef664a3.jar
SHA-256: 22CCB3FF68FC99B0F92751E69E9C9828EE177192A4E46F5E1BD96040E10391DF

The complete captured smoke log is retained with the operator evidence; the concise session evidence above is included here for review.

[JogOnJohn]

Full captured 30-minute live smoke log is now committed to this PR: docs/pest-control/validation/2026-08-28-30-minute-live-smoke.log.

[chsami]

The earlier blockers are addressed on current head a9cd0bd: the PR targets development, the JUnit suite is wired into Gradle, the crush-label aliases are handled, the 10-round validation evidence is present, and CI is green. One merge-blocking loadout case remains.

PestControlCombatPlan.redMeleeLoadout() treats a missing Red weapon by rebuilding the configured Yellow weapon/off-hand with the CRUSH option. A normal Yellow-only weapon such as the test Abyssal dagger or the default Dragon scimitar does not expose Crush. When Red becomes vulnerable, prepareLoadout therefore fails combat-mode verification on every retry and the portal cannot be attacked. The new missingCrushLoadoutFallsBackToYellowWeapon test currently enshrines that incompatible request instead of proving a usable fallback.

Please either preserve the shared weapon actual configured attack option when it is reused for the other portal, or reject a one-entry setup at startup unless that weapon is verified to expose both requested modes. Add an executable regression that covers the resulting usable/rejected path. The one-entry configuration must not fail only after a Red portal is selected.

[JogOnJohn]

@chsami @coderabbitai

Addressed the remaining one-entry melee fallback case in cc29ce2; requesting another review.

When only the Yellow entry is configured, Red now reuses the complete configured Yellow loadout, including its actual Stab/Slash attack mode. It no longer reconstructs that weapon with an unverified Crush request when Red becomes vulnerable. The existing Crush-only -> Yellow fallback remains unchanged and preserves Crush.

The executable regression now verifies that a missing Red weapon with the configured Yellow Abyssal dagger keeps Stab rather than requesting incompatible Crush. gradlew.bat test PestControlPluginJar -PpluginList=PestControlPlugin --console=plain passed with 19 tests, 0 failures, 0 errors.

[chsami]

Maintainer cleanup pushed in 8849f69: stripped the accidental PowerShell file-object wrapper and contributor-local Windows profile path from all 300 validation-log lines while preserving the actual timestamped Pest Control runtime evidence. git diff --check passes, and the sanitized file no longer contains PSPath/PSParentPath/PSChildName/PSDrive/PSProvider/ReadCount or C:\Users\ metadata. A focused Gradle build is currently blocked by the repository-wide Farming Contracts Produce#getContractName() incompatibility already visible in this PR’s CI; I’ll revalidate after the base compatibility fix lands.

[chsami]

Base synchronization is now pushed in 33e241d. The only merge conflict was shared JUnit wiring in build.gradle; I retained the PR’s Jupiter engine dependency and the base’s broader tasks.withType(Test).configureEach { useJUnitPlatform() } configuration. Fresh ./gradlew clean build -PpluginList=PestControlPlugin passed on the merged tree; the report contains 19 Pest Control tests (plus the base’s 5 compatibility/snapshot tests), all with zero failures/errors. git diff --check also passes. Awaiting the new expected CI run before final review.

[chsami]

Maintainer follow-up is complete on f387c57: the committed validation log was sanitized, the prior loadout/JUnit blockers remain fixed, and the expected CI build passed. A fresh local ./gradlew clean build -PpluginList=PestControlPlugin also passed (19 Pest Control tests, plus the current base test set). Because I pushed these fixes and the PR still has an active changes-requested review, I am leaving it for independent review rather than approving or merging it. Note that development advanced again with #547 after this run, so the final reviewer should re-sync and require fresh CI before merge.

[JogOnJohn]

@chsami @coderabbitai

Final re-sync complete. The branch is now rebased onto current upstream/development through #547 (6ecb5e6), with the shared JUnit configuration conflict resolved by retaining the base tasks.withType(Test).configureEach { useJUnitPlatform() } plus the PR's Jupiter engine dependency.

Fresh focused verification on the rebased head passed:

gradlew.bat clean test PestControlPluginJar -PpluginList=PestControlPlugin --console=plain
PestControlCombatPlanTest: 19 tests, 0 failures, 0 errors

git diff --check is clean. A fresh PR CI run has been triggered from this head; requesting independent review once it completes.