So here is my update. Please not I am not trying to be hypercritical. I am simply relaying my thoughts based on price, expectations and experience. I want to love this thing, but am not there yet.
I am getting somewhat stable (but kind of questionable) results for several of the tests. Buckle in this will be long, some of it repeated.
Alk
RedSea Pro - (2 kits) could not get valid results. I don't know if this is kits or bot problems with the RedSeas program.
API - results appear valid - but being only 1dkH in resolution it is hard to tell if this is just lucky or accurate. I am going to test NSW and see what happens when I have a few mins.
Calcium
Using API - results are within 10-15 of the manual API test so not bad and may be stable, but not enough tests yet to determine. A few more days. The bigger issue is that the API test is 40-60 less than Salifert, Hanna and RedSea.
Nitrate and Phosphate
I can't get either to give meaningful results using API or RedSea tests. I don't think there is ANY WAY that I am at 0 phosphate and 0 nitrate, but that is what the bot reads (or close, .02 phosphate). Hanna gives me .12ppm - new RedSea reagents but not had a chance to use them yet to verify.
I can maybe swallow the 0 ppm nitrate - but need one more test kit to accept that. Maybe tomorrow?
Magnesium
RedSea loaded - it is reading 100 to 120 low compared to Hanna, and Salifert. Manual RedSea is also low, but not as low as bot. Eyes vs bot and endpoint? I dunno but hanna appears to agree with my eyes on the Salifert comparison. Bot is also wandering a bit but hanna results and salifert don't. So there is that too.
So bot related - I don't have a handle on the results accuracy or repeatability yet but am in better shape than I was on day 1 and 2. I can't tell you what has changed though and that is a concern too. We will go over that shortly... but let's switch gears for a moment (see updates).
That brings to ReefKinetics, the FB forum and my general impression thus far of many things.
The FB group has plenty of typical arrogant FB personalities mixed in with "happy" customers (fans if you will) that are all too happy to "white knight" and frustrated new owners looking for support (mostly because there is no real documentation).
Very similar to their forum here. ReefKinetics pretty much refrains from responding unless they can put a positive spin on the response, but unlike here, most negative responses and conversations are scrubbed from public view. Judge this as you see fit. I personally think it does for more damage than good, but it is their prerogative.
So in a nutshell community "support" is more akin to "we don't have problems you must be a hater or an idiot". Don't waste your time in the FB group unless you are there to sing happy song with other like minded happy people.
The only benefit of the FB group (my opinion) is being able to DM "Hady" directly. He (I think) earnestly does try to provide support. I have not used the ticketing system, but that may be an even better bet than bother with FB at all.
Anecdote: A certain "Patrick" there who is also an admin (I am told) at the Focustronic (not affiliate but private) support group was rather arrogant toward a new member and I responded that it was not helpful...he sent me a rather nasty 2 word DM and then insta-blocked me... reminding me that maturity (this guys profile shows he is maybe 50) abounds on FB and reinforcing the fact that I would be far better served removing myself from the group.
So ignoring the actual results - here is my updated impression of the robot and software not in any particular order.
-60ml vials DO NOT FIT - They are useable. but do NOT FIT
-Vial must be 55% full to run a test. This is 100% lazy coding. Needle length and vial dimensions are know, it would be trivial to know when (with a safety margin for both needle depth and vial volume) that the syringe would draw air and that is far below 55%. For many tests this is just 11 tests before you get a reagent too low error. That is insane, if I full the vial the day I leave and test once a day, I can only be gone 10 days, yet 55% of the reagent is still left but disabled? Lazy coding and logic.
-Schedule management table is lacking info on time and actual test (it shows what you name it) the rest of the data is shown when you drill down into the test line item. Poor UI design.
-Test scheduling is not well managed. Test run times are not taken into consideration in the scheduler so you can easily overlap test start times with a running test. We are told that overlapping tests will be run in sequence one after the next. The problem is that long tests (this with 3 reagents) can timeout waiting... apparently the timeout is counted from when the test is scheduled to start, not when it actually starts. This is something that MUST be fixed and is again (IMHO) just bad logic and coding but worse, the bot is "new" but the software is mature, being the interface for the two prior robots?
-The above issue brings up two more. Firstly the error reporting is bare bones (it absolutely sucks). You can't drill down into errors. On screen errors are temp ambiguous popups that are not logged, etc. This is a major oversight and what feels like more rough edges that could be easily fixed.
-Two, this appears to expose a real-time dependency between the robot and the hosted services. Meaning that if at any time during a test the bot loses comms with the AWS server or there are other conflicts (the schedule run times out because another test is running) the test will fail. This IMHO (if true) is terrible application design. The robot has a powerful single board computer in it. It absolutely should be autonomous, keeping schedules and run everything locally and cache results. The server should send updates and receive results. If the server is needed for actual processing then fine! Cache the data from the test until the server is available to process it. I don't see this to be the case. Again - if this is not they way that this all works, nobody has stepped up to refute it or tell me how it does.
-That brings me to my next point. Why does the robot stop trying to get an IP address if the network fails and comes back? DHCP does not need to time out, but alas it does and if I am out of town and the network reboots, somebody must manually reboot the robot. I can fix this by putting it on a GHL switched receptacle. but still. Rough around the edges for a mature product.
-Mentioned elsewhere, the UI has some graphics bugs on wide screen monitors and some functions are broken in the web browser that work in the ios app. I can't speak to android.
-I don't see a way to choose push or email, I don't think you can turn off email notifications. Not the end of the world but again goes to roughness in an application that should be more refined after this long.
-With the above - units missing all over the place. Also not the end of the world, but just more lack of polish.
-Mentioned several times. Complete lack of documentation. Reef Kinetics, your FB group is not the right place for people to learn how to use the device that they spent $1500 on...
-Yes - this thing can uses a few dozen different test kits. I understand RK not wanting to pick favorites, but at the same time they know what works well and what doesn't. I think this would be important information for them to document clearly (Pros and cons of different kits for each parameter). They don't. I even created a thread in this forum for that purpose and tagged them but they have ignored it.
-choice of disposable insulin type syringe. The model they have chosen has a mushroom shaped plunger tip. This leaves a drop of fluid in the syringe neck... in this case RODI from the cleaning process. The same style of syringe can be purchased with an extended protrusion on the plunger that clears the syringe neck. This would be more appropriate to ensure that the reagent is not AT ALL mixed with a drop of residual RODI water from the cleaning process. This may or may not affect results, but is something that could.
-There does not appear to be a way to calibrate a returned results to a KNOW better value. Example my mag is returned consistently between 1120 and 1160 for this batch of reagent. I have 3 other kits telling me that it is 1220 to 1260. I absolutely should be able to "calibrate" that value to return the proper result. This is another (I feel) oversight.
-UPDATES UPDATES UPDATES - this is a tremendous sore spot for me, especially given the DIRECT dependence on "hosted" services. I asked some pointed questions and got vague answers about how software development, bug tracking and updates were handled. It was basically "we take care of it, don't worry". That is NOT what I wanted to hear. I get the impression that these folks (they did not deny it) tweak things regularly to resolve issues or tweak results, etc. This just CAN'T happen with a live piece of software. When stuff goes sideways on MY bot or YOUR bot and we thing it is reagent based, it could be developer introduced. Likewise (in my case) when the same reagents don't work for 2 days and then suddenly after complaining, thing look better with NO CHANGES on my side, then how do I know what was fixed or if anything was fixed. This (out of all of my issues) is one of the most concerning and could be the deal breaker. We could spend many paragraphs talking about this problem. There MUST be a notice of ANY backend software changes, as end users can't opt-in/out of being affected. There MUST be a complete and easily viewable change log. It would be in RK's best interest to institute a public bug tracker as well.
-Lastly I want to address what I feel may be a major limitation of this technology and reiterate the need for a way for end users to scale or calibrate the returned result.
1 - some brands of tests are know to read lower or higher on average. This is enough of an argument alone. but..
2 - These tests are designed to read by the human eye against an endpoint or color chart. Slight changes in chemistry from batch to batch have little (no) effect on what your eyes can discern, but may be VASTLY different in terms of a fixed colorimetric scale. Nothing is wrong with the test kit, it works for its intended purpose (your eyes) but is not tightly enough quality controlled for colorimeter accuracy. So again the user needs to be able to adjust (scale/calibrate) the returned result.
So - all in all, I still think that the Reefbot Lab has a lot of potential but there are some very rough aspects to it and a few potential deal breakers (the update scenario and extremely tight dependence on hosted servers) for me.
Phew... sorry that was so long.
I think that is one of the best and most complete reviews I have heard on the bot. It also captures quite a bit of the little bits and pieces I was able to find when I was researching the bot and then some.
The Facebook group, imo, is useless in it's current state due to the lack of responses and the couple toxic members that prevent people from asking questions...well until they are moderated. It seems the preferred method of communication is DM, that really only helps that one person instead of the community as a whole. Also, when I only see the post with the issues, and not company follow up, it doesn't give me a warm/fuzzy.
I had a similar interaction with "Patrick" I was told I was whining and was blocked. To top it off, when a couple customers start posting issues, findings, and asking questions they get moderated to one post per day? what the heck is that? If people were truly complaining then THEY can block you, snooze you, or leave the group. What is worse, is when you start hearing how many others it happened as well.
While I applaud that RK is providing free cloud support so you can get results, push test, etc from anywhere. It's dependency on it WAS the biggest thing that caused me hesitation to buy one. Now seeing how inconsistent it is, has inched it's way to the number 1 reason, but barely. Everything else compounds on why I do not want cloud dependency. Starting with the lack of clear communication from RK since the other issues have been popping up from invisible updates, inconsistencies, and generally the lack of responses to valid questions being asked here and even in their official group.
When I first started researching the bot, I saw how the v1 was a flop, read good things about the v2 and saw how active Hady was in the group and it gave me a good feeling about how the company has matured by learning from the previous two iterations. Maybe with these issues you and the others have been highlighting has got him too busy to engage openly, or maybe just life in general as it does happen, but the silence to me is speaking REAL loud. Which was why my main concern, originally, for the Bot was it's cloud dependency....what happens if the company doesn't make it? Then I would have a beautiful paperweight that does nothing unless someone can RE and redevelop the controls for it. I can tinker and tweak things, but I absolutely can't code from scratch. There is a post in this group from Oct 22, 2020 that asked about the bot going autonomous.
I believe the V1 and V2 both had RPI 3Bs(not sure if they were + or not), and the lab could be the same or possibly even a 4B. If so it has MORE than enough processing power to handle everything autonomously. At one point on my RPI 3B+, I had Reef-Pi, Home Assistant Core, Mosquito MQTT Broker, Node-Red and pushing all the data from MQTT and Reef-Pi's API locally and to Grafana as I was testing different dashboards. The memory, 1GB, never was taxed more than 50% and CPU stayed around 12%.
I appreciate your candid review, and hopefully RK will take some of this constructive criticism back and build on it. I still want one, but after seeing the issues you have had and everything you highlighted. I'll stand by and wait to see if they take the feedback being provided and work with it. I sure hope so.
