new AI for VGCraftBot -- feedback wanted

[Xeon]

This thread is only for feedback on the new AI code

If you have problems with the released script, please post them on the other thread.

There are two Extra Quality Control frames, one for Easy (easy and lower) and one for Hard (Difficult and Moderate) work orders.

The APLimit slider is calculated as the totalStepsLeft / AP left. It is used to force the bot into progress only mode if you start getting low on Action Points. The lower the slider, the sooner Progress Only mode will kick in.

The MaxQ slider will limit the amount of Quality the bot will try to push. The higher the slider, the more Quality it will try to squeeze out.

The APLimit always overrides the MaxQ so if you reach the Quality you wanted, but are ending up with AP left over, try raising the APLimit.

If you are not getting the Quality you want, then raise the MaxQ slider and it will pump quality higher.

Please let me know how this system works for you. Exact setups, with what exact step it used incorrectly, etc are what I need feedback on. Also, if you think the numbers are too low or the sequence of steps is out of wack, that would also help.

You can download this new code here:

http://www.reality.net/svn/vanguard/VGCraftBot-0.992.zip

 

[gnits]

So far, i like it, i am completing difficult items a little more. I would like to see some work on the pathing logic. There needs to be some general identification of where the toon is, relative to its known points on the path. Because depending on things, it will go way out of its way (typically to near the craft supplier) to figure out where it is.


EDIT: Yes, the pathing logic needs work. Especially considering places where the taskmaster moves, and in some places he/she moves a fair distance. Probably to thwart this, but i think some serious work with how it chooses where to go should be rising up on the to-do list.

 

[IeU]

Nice Xeonx,
keep the updates coming !

i will test this new AI doing T5 Moderate, lets see how it peforms.

Btw, can you improve a bit the resources code ?
it buys wayyy tooooo much resources . . .

 

[aChallenged1]

There are items I like that I see here.
There are items I like about VGCrafterHelper.

Both seem to be very well done. I'd like to see you both work together on a single bot that has the best of both worlds.

Some suggestions for dealing with the moving taskmasters. First, it must be possible to see if they are moving. Secondly, it must be possible to tell which direction the taskmaster is moving. Most task masters only have 2 points, one at each end of their paths.

Instead of moving to the taskmaster at each point, why not simply target him at each point and mark those locations. Then, when you need to move to him for turning in or getting work orders you simply see if he is at one end or the other, and moving or not. Continuously check moving as you move towards him, and if he starts moving, wait for him at his next spot. Mark each usable workbench of the types of crafting you are focusing on. We should be able to create a set of databases of what items each work order requires, rather than go to the table, find out we don't have everything, go to the vendor and buy it and back to the table again.

In Khal for example, the refining TM moves about half the length of the hall and waits for a few minutes at each end. There is a refining bench near each of his stops. Why should I to to him at one end, go to the bench furthest from him, find out I don't have the items, and then go to the vendor and back to the table, which again, may not be the closest.

I really wish I understood how to make LavNav work. The concept is there, but it just is not sinking in. I may have to really see about understanding it and making it possible for crafters to not worry about paths, just which tasks they are going to work on.

Since this applies equally to VGCraftHelper I'll re-post in that feedback forum.

I know it is a lot, but with concerted effort, it can be done.

By the way, do NOT take this as criticism at any level. This is awesome work. I'm just trying to suggest (huge) improvements.

 

[Xeon]

the crafting AI has needed some serious work for a while. Now that this new code is done, I will have time to work on other areas... I suppose pathing should be next on the list.
I have some quick mods to the moveto code that should fix a lot of the problems that have been reported so far.

that said, I want to make sure the new crafting code works as designed, so please keep this thread on topic. Thanks!

 

[aChallenged1]

Instead of sticking with moveto, it would be better to figure out LavishNav

The old style pathing that we've been using from the days when EQ ruled and MQ2 was the best of the best just is not up to dealing with the needs we deal with these days. I'm trying to grasp how to use LavNav myself, but I keep hitting mental blocks.

Just a suggestion.

 

[thezig]

I found a bug that seems to pretty consistently botch the new quality/ap system. For whatever reason, this new AI insists on choosing a low quality/high progress step for raising quality. The result, of course, is that the step gets completed very quickly and I gain almost no quality.

 

[Xeon]

I found a bug that seems to pretty consistently botch the new quality/ap system. For whatever reason, this new AI insists on choosing a low quality/high progress step for raising quality. The result, of course, is that the step gets completed very quickly and I gain almost no quality.

Ok, found and fixed. You will need to re-download 0.992 zip again.

Thanks!

 

[gnits]

Ok, Xeon, the logic is working MUCH better after last nights (this mornings) change. Also, i discovered that part of my problem was a failure to update my clothing to current level gear... i had been putting it off until i hit 31.

Now i am doing much better. A pair +45 finishing skills boots are helping, too.

NOTE: It seems there is a marked difference between what i can complete in mineralogy and artificing. I can do far better, getting into the Bs with it. But with mineralogy i do not do as well, it may have something to do with the difference in the steps

 

[slipy00]

Hey there I just installed the new ai and found a couple bugs.

Firstly, I accidentally made the last point to get back to the station to far away, so the program deemed it nessesary to move itself closer. Usually it just takes a few steps forward, but what this bot did was run the entire length of the circle ( supplier, wo npc, station ) and then take a few steps forward.

The second bug I found was when I got a complication I usually fix but was not able to at that time. For instance, I usually have it add cleaner to my table, it just so happens that this time I used up my cleaner at an earlier time in the rec and when I got a complication requiring cleaner the program just sat around trying to use rigging tools all day. (rigging tools being the other solution)

 

[gnits]

The mineralogy ones in tier 4 have 3 steps in stage 2 and once it gets to stage 3 does nothing but progress.

 

[gnits]

Complication logic still will not choose treated water to fulfill a water request, even though manually it will work. (there are others...)

 

[Blazer]

AI Logic

LavishNav is definately the way to go with the pathing. I still havnt figured it all out myself. I agree with AC, should perhaps combine both efforts into one. I spoke to Shaba regarding the AI for crafting, and this was the way I was going to do it, if I had time to do the craft bot.
If you want the perfect craftbot no matter what class, or skills/stats you choose then this is how it should be done.

This will only be really effective from level 20+
The bot needs to learn when it crafts. It starts with the lowest actions, and saves its progress, what percentage it gains for the action used vs the success* (i.e poor, fair, moderate etc) which is dependant on Action Points. You would set a custom number of iterations through each action, and record an average and rate it. Obviously the higher the iteration the better the result.
When it comes to actions that use Progress and Quality you will be able to measure this value with the individual actions.
You would also need to take into consideration complications that can affect the progress and quality.
Also, with the average percentage gain for a skill, you would find a more optimal action to use to end a step.

I see some actions better used than higher ones.

*Amadeus needs to add in a member to see what the result is for an action (i.e poor, fair, good, etc)

 

[Xeon]

The second bug I found was when I got a complication I usually fix but was not able to at that time. For instance, I usually have it add cleaner to my table, it just so happens that this time I used up my cleaner at an earlier time in the rec and when I got a complication requiring cleaner the program just sat around trying to use rigging tools all day. (rigging tools being the other solution)

Yeah, I need to change the way it decides if it has the correct ingredients for corrections to check the table.

 

[Xeon]

The bot needs to learn when it crafts. It starts with the lowest actions, and saves its progress, what percentage it gains for the action used vs the success* (i.e poor, fair, moderate etc) which is dependant on Action Points.

yes, I am currently storing all the actions in a LavishSetting set and tracking the average Progress/Quality. Right now I just compare the cost vs static rating (high/med/low/etc) but I'm going to switch it around to use each action at least once to get the average started. Then I will use those for the comparison.

I also use a rating system to decide if Quality should have a higher "cost" than progress. This should force it to use Progress only actions even if a combined Progress+Quality rating of an action would normally be higher.

I was saving these values, but I keep getting random characters inserted into the .XML file.... really wierd.

 

[kingfisher]

It still does not clear complications w/ 3 effects.

 

[csyxceven]

While doing T4 outfitter WO's the finishing step(7) does not use the 3rd action if there are less then 100 AP but more then(or equal to) 50 AP. The actions available for finishing at T4 (not sure about T3)are:
1 - 100 AP
2 - 0 AP (loss of quality)
3 - 50 AP (small loss of quality)

a few times it meant dropping to a C instead of scrapping by with a B grade

Wonderful script anyway

 

[tesla]

I was looking at your layout and it would seem better if you had control of each WO quality seperatly. I made this VERY CRUDE layout copy and pasting your existing picture into this chopped up mess.

This will give the user full control of each WO seperatly. Rather than a range.

I have found that the best and easiest xp is moderate on c it would seem for some odd reason moderate on b yeilds less xp and has a VERY high fail rate.

I have found that there is a step effect.

Moderate C quality xp is comparable to Easy at B quality. So if u had control on exactly where each grades quality went to that would be great.

Also XP aside once you get into very easy and trivial, you want A quality for the rewards since xp is near nothing at these levels.

Just a thought.

 

[aChallenged1]

I was looking at your layout and it would seem better if you had control of each WO quality seperatly.

This will give the user full control of each WO seperatly. Rather than a range.

I have found that the best and easiest xp is moderate on c it would seem for some odd reason moderate on b yeilds less xp and has a VERY high fail rate.

I have found that there is a step effect.

Moderate C quality xp is comparable to Easy at B quality. So if u had control on exactly where each grades quality went to that would be great.

Also XP aside once you get into very easy and trivial, you want A quality for the rewards since xp is near nothing at these levels.

Just a thought.

I second this!

 

[don'tdoit]

I first this

http://www.isxgames.com/forums/showthread.php?t=712&page=27#262

 

[thezig]

I don't know if this is a bug or not, but the AI seems to be totally ignoring my quality settings and just doing whatever it wants to. I've got the AP at 40, MaxQ at 51, and the thing is repeatedly pumping quality to 75+. Granted, it has NOT actually failed to make an item yet, but I have said to myself "man, if I got a comp right now I'd be fucked" quite a few times.

Anyway, it does seem to be a lot smarter about stuff, and I'm happy to see it finally using tool/station actions that raise quality. Great stuff.

Edit: Okay, apparently AP 40 was too high, as it finally botched an item. I lowered it to 30 and it seems to be doing much better.

One thing I've noticed is that if the utility step is first, it'll use pure quality actions to raise quality until it reaches the appropriate threshold. If the utility step is last, it'll use partial quality/progress station and tool actions. I think the efficiency of the AI could be enhanced greatly if it would take into account potential quality-raising tool/station actions that occur after the utility step. Ideally, it would (based on difficulty and ap/qual sliders) use less pure-quality and more mixed actions, and would even use mixed-effect actions at the end if it finds it has more AP than it expected.

For example, it stops using quality actions at 1500AP and 50 quality because it hit the threshold, but it ends up with 700AP left in the last section (which, for this example, contains a tool and finishing action), possibly due to getting lucky with comps. Instead of going pure progress in this case and ending up with ~400 AP left over, it would choose a partial-quality action. Heck, if you wanted to get really fancy, you could also have the program be slightly more aggressive when you're just a hair under a quality level, for example choosing a partial-quality action with low AP when you're at 74% quality.

 

[Xeon]

One thing I've noticed is that if the utility step is first, it'll use pure quality actions to raise quality until it reaches the appropriate threshold. If the utility step is last, it'll use partial quality/progress station and tool actions. I think the efficiency of the AI could be enhanced greatly if it would take into account potential quality-raising tool/station actions that occur after the utility step. Ideally, it would (based on difficulty and ap/qual sliders) use less pure-quality and more mixed actions, and would even use mixed-effect actions at the end if it finds it has more AP than it expected.

Yes, I really need to add some look ahead code to see what Quality steps are available where and then figure out where it wants to use those. The problem is that there are so many combinations of steps for different types of classes that if I get to fancy, I may screw one type of recipe to make others kick ass. Keeping it generic enough for all classes is a tough balancing act.

For example, it stops using quality actions at 1500AP and 50 quality because it hit the threshold, but it ends up with 700AP left in the last section (which, for this example, contains a tool and finishing action), possibly due to getting lucky with comps. Instead of going pure progress in this case and ending up with ~400 AP left over, it would choose a partial-quality action. Heck, if you wanted to get really fancy, you could also have the program be slightly more aggressive when you're just a hair under a quality level, for example choosing a partial-quality action with low AP when you're at 74% quality.

I will be putting in code today that takes the level of desired Progress vs Quality into account in helping to select the best action. That means if Progress has a weight of 10 and Quality has a weight of 1, then a Very High, pure Progress action would be selected instead of a High Progress + High Quality action.
That should also allow the script to select Low Progress + High Quality actions, which it simply won't use right now.

Great feedback, thanks!

 

[thezig]

Thanks for the response, Xeon.

Another potential bug I found, but haven't reproduced, is that another player sending you a group invite will cause the error "pathing loop" or whatever. I can pause and resume, and the bot will run correctly, but it throws the error and stops the minute it reaches a destination (station/WO/supplier).

I've also heard that being sent a trade invite breaks certain things, too, but I'm not sure if that occurs with your particular program.

 

[Xeon]

Thanks for the response, Xeon.

Another potential bug I found, but haven't reproduced, is that another player sending you a group invite will cause the error "pathing loop" or whatever. I can pause and resume, and the bot will run correctly, but it throws the error and stops the minute it reaches a destination (station/WO/supplier).

I've also heard that being sent a trade invite breaks certain things, too, but I'm not sure if that occurs with your particular program.

Hmm, havn't heard of the group invite problem before. I'll fix the 'pathing loop' bug on not reseting when you hit Pause/Resume.

The script now checks and rejects Tade Requests... At least the code is in, I still need someone to test that to see if it works.

 

[gnits]

Thanks for the response, Xeon.

Another potential bug I found, but haven't reproduced, is that another player sending you a group invite will cause the error "pathing loop" or whatever. I can pause and resume, and the bot will run correctly, but it throws the error and stops the minute it reaches a destination (station/WO/supplier).

I've also heard that being sent a trade invite breaks certain things, too, but I'm not sure if that occurs with your particular program.

See the following link.
http://www.vanguardcrafters.com/forums/viewtopic.php?t=4741

It is very enlightening. We should find a way to put some randomness in the points we go to. I can see this being a problem. The warpath is being sought.

 

[gnits]

Speaking of pathing bugs, it appears that in some cases the bot will spin in place when it is trying to go somewhere. It happens a lot for me when i go from the craft station to the supplier (actually, right when he leaves.)

 

[aChallenged1]

That is exactly why we need LavishNav and not using the old style pathing.

Please, someone better skilled take up the burden of putting together a system that uses LavishNav, and gives us a way to add in new points, such as places to avoid, like cliffs, walls, etc.

There is nothing more noticable than a bot that looks like a bot.

 

[gnits]

Stage 2 - pt 1
Corrode - 75 mod - none
dissolve - 100 non - high
complete removal 125 - vhigh - none
deep cleaning 100 - mod - vlow
full cleansing 150 - high - low
Stage 2 - pt 3
Basic level 75 - mod - none
perfect adjustment 150 - vhigh - v low

and a stage 2 pt 2 that seems greyed out...

and all it seems to do is a little quality and thats it.. always getting Ds or most of the time.

btw its cost - prog - qual - (above)

 

[DrmChld]

Tried the new AI, the bot still does not add extra ingredients to the table unless it is set to Full Auto. Otherwise it adds nothing.

 

[Xeon]

Tried the new AI, the bot still does not add extra ingredients to the table unless it is set to Full Auto. Otherwise it adds nothing.

Hmm, yeah, I took that out a while ago... Why would you want it to?

 

[Xeon]

Version 0.993 released

 * Version 0.993 -- March 21st, 2007
	Massive change to Crafting Action selection AI
	Changes to movement code to use Collision Detection
	Added Trade Invite reject
	Slow Turn while moving will NOT stop at each waypoint

http://www.reality.net/svn/vanguard/VGCraftBot-0.993.zip

I need someone to test the Trade Invite rejection code.

 

[DrmChld]

Hmm, yeah, I took that out a while ago... Why would you want it to?

I use it for everything else, just not to choose the WO's or do the pathing.

I try to stay cautious. I stay at my keyboard. Not having that in there makes it so I have to pause it, add the extra ingredients for each combine. As opposed to doing 15 combines before I have to do anything manually.

 

[aChallenged1]

I haven't really looked, but can this bot, as is, allow me to do the runs to buy, sell, obtain, complete WOs, and simply automate the actual work of doing the crafting?

 

[IeU]

I haven't really looked, but can this bot, as is, allow me to do the runs to buy, sell, obtain, complete WOs, and simply automate the actual work of doing the crafting?

everything.
leave there and go out with your girlfriend

 

[aChallenged1]

I meant manually, only doing the bench work using the bot only to do the actual bench work.

I'm not satisfied with the bots when it comes to movements.

 

[Xeon]

Could someone who is running the 0.993 code take a look at their scripts/vgcraft/save/ToonName/action-data.xml file and see if there are any weird characters in there?

I'm getting a strange char inserted in random spots through out the file and was wondering if anyone else is also...

<Set Name="Brief Polish">
      <Setting Name="Stage">244938752</Setting>
      <Setting Name="Name">Brief Polish</Setting>
      <Setting Name="ActionPointCost">0</Setting>
      <Setting Name="Progress">0</Setting>
      <Setting Name="Quality">0</Setting>
      <Setting Name="ToolRequired">FALSE</Setting>
      <Setting Name="ToolName">NONE</Setting>
      <Setting Name="ItemRequired">FALSE</Setting>
      <Setting Name="ItemName">NONE</Setting>
      <Setting Name="QualLoss">3</Setting>
      <Setting Name="NumUsed">0</Setting>
粑      <Setting Name="AvgProg">0</Setting>
      <Setting Name="AvgQual">0</Setting>
   </Set>

 

[slipy00]

I just installed the ai update after falling back to the official release due to a complication bug and in this new version I am still getting it.
Whenever I get the complication "Damaged Tool" the program will sit there and try to solve it using rigging tools, and I do not keep rigging tools on my toolbelt. It will continue like that forever (or until the kill time). the complication has two solutions, either rigging tools or cleaner, the program does not buy or put cleaner on my table so it fails every time to resolve this complication and I have to step in and do some clicking enough times to make the complication go away. Then the bot continues like normal.

I just got another one. I got another complication while crafting and the bot was working this one out, but then it ran out of mats to finish fixing the complication. So the bot sat there saying 'Out of mats' while trying to fix the complication with the 1 mat I had left (the solution required 2)

 

[IeU]

<?xml version='1.0' encoding='UTF-8'?>
<!-- Generated by LavishSettings v2 -->
<InnerSpaceSettings>
	<Set Name="Supplied Board">
		<Setting Name="Stage">131495936</Setting>
		<Setting Name="Name">Supplied Board</Setting>
		<Setting Name="ActionPointCost">0</Setting>
		<Setting Name="Progress">0</Setting>
		<Setting Name="Quality">0</Setting>
		<Setting Name="ToolRequired">FALSE</Setting>
		<Setting Name="ToolName">NONE</Setting>
		<Setting Name="ItemRequired">FALSE</Setting>
		<Setting Name="ItemName">NONE</Setting>
		<Setting Name="NumUsed">0</Setting>
		<Setting Name="AvgProg">0</Setting>
		<Setting Name="AvgQual">0</Setting>
	</Set>
	<Set Name="Refined Board">
		<Setting Name="Stage">131495936</Setting>
		<Setting Name="Name">Refined Board</Setting>
		<Setting Name="ActionPointCost">0</Setting>
		<Setting Name="Progress">0</Setting>
		<Setting Name="Quality">0</Setting>
		<Setting Name="ToolRequired">FALSE</Setting>
		<Setting Name="ToolName">NONE</Setting>
		<Setting Name="ItemRequired">FALSE</Setting>
		<Setting Name="ItemName">NONE</Setting>
		<Setting Name="NumUsed">0</Setting>
		<Setting Name="AvgProg">0</Setting>
		<Setting Name="AvgQual">0</Setting>
	</Set>
	<Set Name="Vice">
		<Setting Name="Stage">131495936</Setting>
		<Setting Name="Name">Vice</Setting>
		<Setting Name="ActionPointCost">75</Setting>
		<Setting Name="Progress">3</Setting>
		<Setting Name="Quality">0</Setting>
		<Setting Name="ToolRequired">FALSE</Setting>
		<Setting Name="ToolName">NONE</Setting>
		<Setting Name="ItemRequired">FALSE</Setting>
		<Setting Name="ItemName">NONE</Setting>
		<Setting Name="NumUsed">0</Setting>
		<Setting Name="AvgProg">0</Setting>
		<Setting Name="AvgQual">0</Setting>
	</Set>
	<Set Name="Clamp">
		<Setting Name="Stage">131495936</Setting>
		<Setting Name="Name">Clamp</Setting>
		<Setting Name="ActionPointCost">100</Setting>
		<Setting Name="Progress">4</Setting>
		<Setting Name="Quality">0</Setting>
		<Setting Name="ToolRequired">FALSE</Setting>
		<Setting Name="ToolName">NONE</Setting>
		<Setting Name="ItemRequired">FALSE</Setting>
		<Setting Name="ItemName">NONE</Setting>
		<Setting Name="NumUsed">0</Setting>
		<Setting Name="AvgProg">0</Setting>
		<Setting Name="AvgQual">0</Setting>
	</Set>
	<Set Name="Tighten">
		<Setting Name="Stage">131495936</Setting>
		<Setting Name="Name">Tighten</Setting>
		<Setting Name="ActionPointCost">50</Setting>
		<Setting Name="Progress">2</Setting>
		<Setting Name="Quality">0</Setting>
		<Setting Name="ToolRequired">FALSE</Setting>
		<Setting Name="ToolName">NONE</Setting>
		<Setting Name="ItemRequired">FALSE</Setting>
		<Setting Name="ItemName">NONE</Setting>
		<Setting Name="NumUsed">106</Setting>
		<Setting Name="AvgProg">0.000000</Setting>
		<Setting Name="AvgQual">NULL</Setting>
	</Set>
	<Set Name="Brace">
		<Setting Name="Stage">131495936</Setting>
		<Setting Name="Name">Brace</Setting>
		<Setting Name="ActionPointCost">125</Setting>
		<Setting Name="Progress">5</Setting>
		<Setting Name="Quality">0</Setting>
		<Setting Name="ToolRequired">FALSE</Setting>
		<Setting Name="ToolName">NONE</Setting>
		<Setting Name="ItemRequired">FALSE</Setting>
		<Setting Name="ItemName">NONE</Setting>
		<Setting Name="NumUsed">1787</Setting>
		<Setting Name="AvgProg">0.000000</Setting>
		<Setting Name="AvgQual">NULL</Setting>
	</Set>
	<Set Name="Fasten">
		<Setting Name="Stage">131495936</Setting>
		<Setting Name="Name">Fasten</Setting>
		<Setting Name="ActionPointCost">75</Setting>
		<Setting Name="Progress">2</Setting>
		<Setting Name="Quality">1</Setting>
		<Setting Name="ToolRequired">FALSE</Setting>
		<Setting Name="ToolName">NONE</Setting>
		<Setting Name="ItemRequired">FALSE</Setting>
		<Setting Name="ItemName">NONE</Setting>
		<Setting Name="NumUsed">0</Setting>
		<Setting Name="AvgProg">0</Setting>
		<Setting Name="AvgQual">0</Setting>
	</Set>
.
.
.
.
.
.

 

[Xeon]

Whenever I get the complication "Damaged Tool" the program will sit there and try to solve it using rigging tools, and I do not keep rigging tools on my toolbelt. It will continue like that forever (or until the kill time).

I just got another one. I got another complication while crafting and the bot was working this one out, but then it ran out of mats to finish fixing the complication. So the bot sat there saying 'Out of mats' while trying to fix the complication with the 1 mat I had left (the solution required 2)

Ok, both of those are fixed in the dev tree and for next release

 

[Xeon]

action-store.xml

Weird characters in action-store.xml file

Please only post a yes or no answer if you find any... You may have to look far down into the file to find them. If you open the file with Notepad or Wordpad, you will see them as a strange chinese character or a [] looking one.

Thanks!

 

[mordacai]

I try to stay cautious. I stay at my keyboard. Not having that in there makes it so I have to pause it, add the extra ingredients for each combine. As opposed to doing 15 combines before I have to do anything manually.

This is pretty much how I operate, Never truely AFK, Would be nice to get the table to auto load utilities. I dont bother loading extra utilities at the moment. seems to do alright on easy WO's

 

[Xeon]

This is pretty much how I operate, Never truely AFK, Would be nice to get the table to auto load utilities. I dont bother loading extra utilities at the moment. seems to do alright on easy WO's

I've gone through the code and changed it so that it will auto-load extra ingredients and do all available work orders even when not in full auto-mode.

I'll test it later tonight just to make sure it works correctly.

 

[gnits]

Hey Xeon, there are still issues with logic. It doesnt seem to be saving enough AP for a stage 4 recipe that requires a decent amount of AP. And again it seems to have a better time with artificing than with mineralogy.

Can we put in a box that causes it to favor one type over another? Send me a PM if you want a log of one that fails.

 

[IeU]

as you saw in my post, no strange chars.

and the ap/limit and the other setting, isnt saving after a shutdown of vgcraftbot.

 

[qevlhma]

I've gone through the code and changed it so that it will auto-load extra ingredients and do all available work orders even when not in full auto-mode.

I'll test it later tonight just to make sure it works correctly.

You have a divide by zero error in your new AI somewhere.

I am guessing it is where action pts = 0.

 

[fancy]

Quality Settings

Not sure if anyone else is experience this but with my blacksmith I can't seem to get the settings correct I have them turned up all the way and am barely making C items and have 500-700 points left over. This has only happened since the last update. My personal opinion was I like the D-A slider better myself but thats just one guys opinion.

 

[Theras]

The new AI is killing me. In general, I am failing a lot more than previously and also the WOs are taking much longer to complete. I've jerked with the settings until my hands bled.

I have two actions that, for 50pts, add a low amount of quality. Generally these come at the very last stage. The bot is saving up and doing all progress until this stage and then spamming quality for like 5 minutes.

How possible would it be to implement something similar to the Action Store that watches how we do WOs and tries to mimic them by using the same actions if possible, etc.? If I could sit down and do WOs manually for a day in order to make the bot 200% more efficient I gladly would. Heck, make the results sharable and we could all upload them to you and you can combine them to get a kind of master list.

Issues:
-Bot waits till very end and tries to pump quality from 10 or 20% to 100% using a very low AP action, which kills me because it'll get complications and be unable to finish.

-Bot pumps quality at beginning stages to 100%, resulting in wasted AP when it uses skills that add progress/AP as it progresses throughout the WO.

In all, it is very inefficient in which actions it chooses in each stage. If there is anything we can do to help make this a bit more smooth please let us know. I think the bot is doing its best to complete the WO at 100% quality and progress. When it chooses its actions it is choosing the ones to get the job done the easiest. However, it doesn't take into account that doing a 50ap action 10 times creates a MUCH larger chance for a complication. I'd much rather spend 600AP in 3 actions to get the same effect as that 500AP because the chances of a complication popping and damaging the entire stage are much less.

Also, please add a checkbox that disables the "run path after resuming WO."

With the new changes in moving taskmasters I've been pausing the bot, grabbing WOs, walking back to the station, and starting it. It then does the WOs (filling up on fuel as necessary) and I repeat. However, every time I start the bot after walking back to the station it runs around and checks the path. Then it runs and buys supplies if it needs to. This wastes a bit of time that adds up throughout the day.

Thanks man, you're doing an awesome job btw =)