Update Healer Man server

docs/ROM_FIVE_VALIDATION.md

@@ -0,0 +1,58 @@
+# RoM five class validation
+
+Status: **Needs Testing**. This is a targeted diagnostic for Priest healer, Knight tank, Scout DPS, Rogue DPS, and Mage DPS. Elite-skill mechanics remain outside this pass.
+
+## Cohort and rules
+
+The five named builds and live NPCs use these race-legal pairings. Their secondary classes activate at level 10. RoM has no talent allocation; ordinary skills unlock by level, and the generated source's internal level variants do not create a player-facing spell-rank choice.
+
+| Role | Class pair | NPC name | Race |
+| --- | --- | --- | --- |
+| Tank | Knight / Scout | Rorik | Human |
+| Healer | Priest / Knight | Serenna | Human |
+| DPS | Rogue / Warrior | Veyra | Human |
+| DPS | Scout / Rogue | Caelan | Elf |
+| DPS | Mage / Warlock | Durnik | Dwarf |
+
+`src/game/romFiveIdentities.ts` owns these identities for the live NPC templates and targeted canonical builds. The manual party uses Priest as the active healer and the other four classes as companions. This explicit cohort does not change Ready-only random parties, bracket certification, or the companion readiness registry. Elite skills can appear in a paired kit through the existing RoM catalog, but their mechanics have not yet been validated for this cohort.
+
+`npm run test:rom-five` checks the named pairings, legal races, level-10 unlock, role-specific heirloom layouts, ordinary ability coverage, and focused runtime mechanics. Knight is a tank seat in RoM generated parties. Rogue uses dual daggers, Knight a weapon and shield, and Scout a bow.
+
+## Findings and fixes in this pass
+
+- Player and companion casts now respect source out-of-combat restrictions for Resurrection and Rogue openers. Sneak Attack and Blind Spot carry behind-target rules for spatial combat.
+- Rogue Throw executes one projectile hit instead of two identical source attack components. Scout Shot executes its primary hit instead of five alternate attack components and unrelated target stat effects.
+- RoM character creation, Skills, and Spellbook screens present level-based skill unlocks without talent-point or spell-rank choices.
+- The RoM Knight party template is assigned to the tank pool rather than the damage pool.
+- Conditional combat payloads now resolve once per cast for both player and companion actors. Knight Holy Seals cap at four on each target, their bleed follows the stack count, and Punishment, Threaten, Shock, and Truth Shield Bash use the corresponding seal branch. Punishment consumes the seal bleed. Rogue Low Blow, Wound Attack, Shadowstab, and Sneak Attack use active self-applied bleeds and the equipped main-hand weapon; Wound Attack no longer invents a bleed. Mage Electric Explosion executes either its normal or Charged hit.
+- Rogue Premeditation now guarantees the next offensive critical hit and increases its critical power. Poison supplies the Poisonous state needed by Numbing Dagger. Vanish permits an out-of-combat Rogue opener while its stealth state lasts. Shadow Step moves behind the target; Knight Charge closes to melee range.
+- Priest Regenerate has one heal-over-time effect. Holy Candle and Scout Blood Arrow drain the caster's HP on their periodic ticks. Mage Flame and Fireball execute one hit, Meteor Shower five, and Thunderstorm three. Source helper summons and zero-damage effects no longer create extra combat actions for these skills.
+- Scout Lasso, Rogue Numbing Dagger, and Mage Static Field apply one break-on-damage control effect to one target at a time. Scout Joint Blow only slows; Mage Discharge applies its stun without unrelated control effects. Knight Shield of Discipline absorbs the documented damage types and reflects a spell hit after the caster is struck, not on cast. Shield-dependent Knight skills require an equipped shield.
+
+## Combat diagnostic
+
+The following historical runs predate secondary-class assignment. They used `scripts/balance_lab.py` with `--source rom --party-mode manual`, exactly the five canonical build IDs for that level, and 180-second fights. Reports are under `artifacts/balance-lab/rom-five-primary-*`.
+
+| Level | Encounter set | Runs defeated | Death events | Median time to defeat |
+| ---: | --- | ---: | ---: | ---: |
+| 20 | Cayus, two seeds | 2/2 | 0 | 70.2 s |
+| 40 | Cayus, two seeds | 2/2 | 0 | 79.4 s |
+| 60 | All 62 RoM bosses, one seed | 62/62 | 0 | 101.5 s |
+| 60 | Shadoj, three seeds | 3/3 | 0 | 106.6 s |
+| 80 | Cayus, two seeds | 2/2 | 0 | 122.0 s |
+
+The table above also predates the conditional mechanics changes. The level-60 all-boss report measured approximately 1,036 Knight tank DPS, 831 Rogue DPS, 374 Mage DPS, 223 Scout DPS, and 35 Priest HPS. These are screening measurements, not balance targets: this cohort has only one healer seat, one tank seat, and one seed per boss. Ten bosses dealt no effective damage and therefore did not exercise healing.
+
+After the mechanics corrections, the targeted Shadoj run at level 60 with the same five manual builds and seeds 1, 2, and 3 defeated the boss in 3/3 fights with no deaths. Its mean defeat time was 115.7 seconds. The report at `artifacts/balance-lab/rom-five-primary-l60-mechanics-final/` measured about 1,186 Rogue DPS, 750 Knight DPS, 237 Scout DPS, 186 Mage DPS, and 334 Priest HPS. The changed output is expected from removing duplicate hits and enforcing source conditions; it is a diagnostic, not a promotion or tuning target.
+
+## Before Ready promotion
+
+1. Resolve the persistent damage spread: Rogue is now far ahead of Scout and Mage, while Knight tank damage remains above both. Review source formulas, resource cadence, and rotations before tuning coefficients.
+2. Finish source-specific restrictions that have no current inventory or creature-family authority. Numbing Dagger's projectile consumption and living/Humanoid target rule still need those systems. A source-row audit should continue beyond the corrected conditional skills; basic semantic coverage does not prove full tooltip parity.
+3. Exercise behind-target requirements in a live spatial scene. The headless balance runner has no spatial target provider, so its Rogue damage can include casts that a live scene would reject.
+4. Validate Priest throughput against encounters that actually demand sustained healing, and run RoM NPC seat tests after the mechanics above are resolved. The current manual run leaves Priest player-controlled.
+5. Verify both Rogue daggers visually from an unobstructed angle. The live roster and dungeon screenshots in `artifacts/playtest/rom-five/` confirm rendered bodies and weapon silhouettes, but the off-hand dagger is unclear at their camera angles. Keep RoM entries at Needs Testing until the full class mechanics and role checks pass.
+
+The RoM source catalog and generated data remain authoritative for ability descriptions and components. The recorded combat results cover primary-only builds; rerun targeted diagnostics before drawing conclusions about the new pairings. Elite skill validation is deferred.
+
+RoM stat and equipment fidelity is a separate prerequisite for balance certification. `docs/ROM_STATS_AND_GEARING.md` records the source conversion coefficients, current WoW-shaped gear adapter, native item import gaps, and the players-only RoM online group boundary.

docs/ROM_STATS_AND_GEARING.md

@@ -0,0 +1,55 @@
+# RoM stat weights and gearing audit
+
+Status: **source model identified; runtime conversion and equipment import pending**. This audit covers the Priest/Knight healer, Knight/Scout tank, Scout/Rogue DPS, Rogue/Warrior DPS, and Mage/Warlock DPS. It does not assign balance weights to unimplemented RoM gear.
+
+## Current runtime
+
+- `src/game/romStats.ts` reads primary-class level growth from `romPlayerCatalog.generated.json` and gives the secondary class 10% of its base stats **at the same level as the primary**. This shared level is the game's progression rule, even though the historical source allowed separate class levels. The runtime also omits source race and sex additions.
+- `src/game/wotlkStats.ts` maps RoM Wisdom into WoW Spirit, then derives HP, Mana, attack power, spell power, avoidance, and ratings with WoW rules. This is a compatibility adapter, not RoM stat conversion. In particular, the global Mana multiplier, Stamina-to-HP rule, and `0.35 * Intellect + 0.15 * Spirit` spell-power rule are not source RoM formulas.
+- `src/game/specializationCatalog.ts` selects one of the WoW gear archetype weight tables for RoM. `src/game/heirloomEquipment.ts` uses WoW budgets and compares against WoW boss gear to create RoM heirlooms. The Scout currently receives a two-hand mace alongside a bow, although source Scout equipment allows one-hand swords, daggers, bows, and crossbows. `src/game/equipment.ts` also lacks the source Priest shield permission.
+- `src/game/runewakerLootCatalog.ts` keeps imported RoM drops as non-equippable `misc` items with zero stats. The loot importer preserves names and probabilities but does not decode native equipment slots, base stats, or enhancements. `src/game/lootPolicy.ts` suppresses **all** boss item drops during leveling; boss cards are awarded through a separate path. At level 80, imported RoM boss items still lack usable equipment data.
+
+## Source stat model
+
+The historical RuneWaker source is under `../rom-pvt-server-backups/Runewaker/`. `SourceCodeVS2019/Libraries/Union/RoleData/RoleAttr.cpp` builds each conversion argument from race + vocation + sex. `Resource/data/voctable.db`, `racetable.db`, and `sextable.db` hold those terms; `ObjectStruct.h` defines the argument layout. The player vocation record is the second record of each class pair, as already used by `scripts/runewaker-pipeline/vocation-growth.mjs`.
+
+After base growth and equipment attribute bonuses, source `RoleAttr.cpp` computes:
+
+| Native result | Source dependency before rate modifiers |
+| --- | --- |
+| Max HP | base HP + Strength × `HpStrArg` + Stamina × `HpStaArg` + flat equipment HP |
+| Max MP | base MP + Intelligence × `MPIntArg` + Wisdom × `MpMndArg` + flat equipment MP |
+| Physical ATK | Strength × `AtkStrArg` + Intelligence × `AtkIntArg` + Agility × `AtkAgiArg` + flat equipment ATK |
+| DEF / MDEF | Stamina × `DefArg` + armor/flat DEF; Wisdom × `MDefArg` + flat MDEF |
+| Magical ATK | Intelligence × `MAtkArg` + flat equipment MATK |
+| HP / MP regeneration | base rate + Stamina × `ReHpArg`; base rate + Wisdom × `ReMpArg` |
+| Physical hit / dodge | Agility plus corresponding equipment values and fixed arguments |
+| Magical hit / dodge | Wisdom, with Agility also contributing to magical dodge |
+
+The table below shows **vocation terms only** from the player records of `voctable.db`. Race and sex terms are additive, so these are not final character coefficients and must not be pasted into a gear-score formula as complete weights.
+
+| Primary role | Relevant source vocation terms | Gear consequence to validate |
+| --- | --- | --- |
+| Knight tank | Stamina → HP 5, DEF 1.5; Wisdom → MP 5, MDEF 3.2 | Value shield/armor DEF, Stamina, MDEF, and survival against actual encounter damage. A WoW defense-rating budget is not a substitute. |
+| Priest healer | Intelligence → MATK 2, MP 1; Wisdom → MP 5, MDEF 3, MP regeneration 0.3 | Keep healing effect power, Mana sustain, and Wisdom distinct from WoW Spirit and generic spell power. |
+| Scout DPS | Strength → physical ATK 2; Agility contributes to hit/dodge through fixed arguments | Bow damage and physical ATK need source conversion before a Strength/Agility score can be trusted. |
+| Rogue DPS | Strength → physical ATK 1; Agility → physical ATK 1 and hit/dodge | Evaluate both attributes and equipped dagger damage, including the off hand. |
+| Mage DPS | Intelligence → MATK 2; Wisdom → MP 5 and MDEF 2.3 | MATK, magical weapon damage, and sustain need separate scoring. |
+
+Priest healing cannot be scored by MATK alone: the source healing path also uses magic-power rate and heal-point inputs. Critical chance, weapon damage, resistance, and encounter mix further change the value of a stat. These conversion coefficients are mechanics, not final DPS/HPS weights.
+
+## Equipment model to implement
+
+RoM equipment has base item properties plus additional attribute sources. [Gameforge's class guide](https://archive.runesofmagic.gameforge.com/en/classes.html) lists native class weapon/armor permissions. Its [rune guide](https://archive.runesofmagic.gameforge.com/us/runes.html) describes socketed runes and up to four rune slots; its [Arcane Transmutor guide](https://archive.runesofmagic.gameforge.com/en/arcanetransmutor.html) describes transferring attributes through Fusion and Mana Stones. These changes must be represented as item-instance bonuses rather than a WoW item-level budget. Source `RoleAttr.cpp` already distinguishes flat equipment contributions, attribute contributions, and multiplicative rates.
+
+1. Decode `armorobject.db` and `weaponobject.db` into native slot, armor/weapon type, base defense/damage, attribute, and level fields. Preserve source object IDs and validate the binary layout against source structures and known items before allowing package equipment to equip.
+2. Add a RoM item-instance bonus layer for native flat stats, transferred attributes, rune sockets, and later enhancement/tier data. Sum those bonuses into a native stat block. Keep WoW `ItemStats` and heirloom scoring unchanged for WoW characters.
+3. Extend the generated class data with race and sex arguments and the vocation conversion coefficients. Keep secondary growth tied to the primary's level, including after the level-10 secondary unlock. Convert the complete RoM stat block to combat inputs through a RoM-specific adapter, including ATK, MATK, DEF, MDEF, and separate healing inputs.
+4. Use legal starter equipment as a temporary bridge, then source-backed quest or item-shop equipment packages for leveling. The client snapshot confirms four armor-choice quest reward families at levels 50, 55, 60, and 65; see `docs/ROM_LEVELING_GEAR_PACKAGES.md`. Resolve the package expansion data and missing earlier/later tiers before granting source items. Dungeon bosses award cards during leveling, not equipment or miscellaneous items. Do not tune a RoM heirloom to a WoW drop score or apply WoW rating conversions to it.
+5. Derive role-specific *upgrade scores* from the marginal combat effect of each native stat at the current level and gear state. Measure Knight survival/threat, Priest effective HPS and sustain, and each DPS build's damage on the same targeted encounter set. Source coefficients seed that model; targeted combat measurements establish practical weights and catch caps or nonlinear effects.
+
+## Online and validation boundary
+
+`server/database.mjs` rejects a RoM/WoW source mixture when a grouped player selects a character and checks again when an activity starts. Until RoM AI companions are separately ready, online RoM groups launch players-only; the WoW-only randomized Bite-Sized Stockades activity is unavailable to them. This does not change local manual RoM party diagnostics. The five RoM profiles remain **Needs Testing** and elite skills remain deferred.
+
+Validate each import with source-record fixture tests, legal slot/equip tests for all five builds, direct native-stat conversion cases, then `npm run test:rom-five` and targeted level/encounter diagnostics. Compare package upgrades by native stat deltas and combat outcomes before changing dungeon tuning. Keep a common encounter definition with source-specific stat translation and measured balance coefficients; no duplicate WoW/RoM dungeon catalogs are required by this audit.