Michael

This project automates a task that was traditionally manual, slow, and error-prone for game administrators.

It processes a massive C++ codebase (over six million lines) where quest drop logic is hidden inside random math functions such as gg::Rand, which calculates drop probability, followed by calls like myself::GiveItem1, where the actual item quantity is defined when a required monster is killed. Modifying these values used to require binary decompilation and hand-editing.

At the core is AI.obj, a NASC-compiled object loaded by the NPC Manager that defines NPC and monster behavior. The tool integrates with the NASC compiler/decompiler pipeline, parses the decompiled output, detects drop-related patterns, and adjusts mission parameters before rebuilding the object.

Earlier projects like Elhaym preserved editable Script Maker sources, making these changes trivial. Once that layer disappeared, customization became painful. This project restores that flexibility through automation.

What once took hours or days of reverse engineering can now be done safely and consistently at scale.

NASC AI is event-driven. Each AI class defines handlers for specific events such as TALKED (when a player interacts with an NPC) or MY_DYING (when the NPC dies).

The MY_DYING handler is the critical one: it checks whether the killer has an active quest tied to that NPC and then uses a random function to decide whether a quest item is dropped.

The tool analyzes each AI class and looks for three concrete patterns:

• The NPC is involved in a quest-related kill condition.
• The MY_DYING handler contains a random math function.
• A quest item is granted after that random check.

Once all conditions are met, the tool reconstructs the original drop chance, applies a configurable multiplier, and injects a named parameter into the class, for example:

int 100_30_trial_of_the_seeker_medusas_scales = 30;

This parameter is placed in the class’s local parameters section, which remains editable even after the AI is compiled into NASC bytecode (similar to assembly and mostly unreadable). As a result, administrators can fine-tune quest drop rates without reverse engineering or recompiling the entire AI logic.