for performance reason, it may be better to define a new collection (using gameState.updatingGlobalCollection as is done for example for civil center in headquarter.js) which would contain all relics, and use it instead on looping on enemy units.

I think you must also modify the functions which find the targets in attackPlan to only choose relics when the target is gaia, and abort the attack otherwise.

:)

playedTurn%400 is not ok as it depends on SP (0.2 * 8 * 400 = 640s) or MP (0.5 * 8 * 400 = 1600s). Rather use gameState.ai.elapsedTime, caching the time of the last try.
Although i've not yet tried this "capture the relics", these delays seem too much to me. What about 5 mn for both SP and MP ? and even maybe (5 - 0.5*(aidifficulty-3)) mn?
In addition, we can have a variant of tryCaptureGaiaRelic which would not wait for relics inside the player's territory (as the ai would certainly be able to see it) or those the ai already tried to capture but failed.

Instead of adding checks to all those functions in attackPlan.js, I think it would be better to add a check in the update function of the attack manager when it loops through startedAttacks, so that if a gaia relic is not the target when the attack starts (or if the target changes later), the attack will be aborted.

I'm not convinced that is the best choice:

• there are not so many attackPlan functions to modify: it seems to me that it is enough to add in getnearesttarget, in the loop to find the nearest target, something like if (player=0 && gametype = "capture_the_relic" && !ent.hasClass("Relic") continue
• in the future, we may add other attacks targeting gaia, so your changes in attackManager should anyway be done only if gameType = "capture_the_relic"

btw in defaultTargetFinder, gameState.getEnemyUnits(playerEnemy).filter(API3.Filters.byClass("Relic")) could be replaced by allRelics.filter(relic => relic.owner() === playerEnemy) which should be faster.

Sorry, i forgot to answer that one: my idea was to have two timers to look for relics:

• the one you've done (with a lapse time of 5 mn) looking for any relics
• another timer with a shorter lapse time (maybe 1 mn or even less, 30s ?) which would only look for relics we are already supposed to know the position (i.e. those inside the AI's territory or those that the AI already tried to capture but failed and which have not yet been captured by another player)

but this distinction can be in another patch.

Okay, I have moved the check.

It's just that I thought from a design perspective it might be better to have that check in attackManager.

I updated defaultTargetFinder.

Ah, okay I see :)

I've the impression this line would be better in attackManager with line 385 where we put tryCaptureGaiaRelic to false as it can take some time to start the attack. What's your opinion?

Yes, I agree, that sounds like a good idea.