KM 11 is so slow it's unusable for me (video inside)

Questions & Suggestions
, ,
21

As I noted earlier in this thread, I am also observing the slow down. I went to disable 3rd party integrations in 1Password following Peter's observation. There is no option to disable these integrations in 1Password 7 so I reached out to 1Password support. I was advised that these integrations are not used in version 8.

We made sure that version 7 wasn't installed, reinstalled version 8, then removed the data for version 7 that contains the 3rd part integrations from my system - just to be sure they weren't being used by any other apps.

So, in my case running 1Password 8 with no 3rd party integrations my macros are very slow compared to the situation running KM 10.

I would conclude that if 3rd party integrations with 1Password are responsible in some way for some people's slow downs, that doesn't explain my situation and there must be some other cause.

22

Thanks for the update, PaulF. That entirely coincides with my experience. The only variable that changed when I simultaneously and instantaneously took a performance hit on three different Macs was the upgrade to KM11. All of my machines were playing nicely with 1P8 until I changed KM versions.

23

I don't know what to say, other than to repeat the advice given by the founder of 1Password, "Don't use 1P8 with third party apps until we add support for that." The fact that 1P8 sometimes works with third party apps is not proof that the fault lies with third party apps.

25

Now that 1Password is discontinuing support for 1P7, I'm wondering whether there has been a resolution yet (and whether KM and 1P8 play nicely now).

26

I searched the web and I also used ChatGPT to ask your question. The threads on the web cited above appear to have been deleted from the 1Password website.

However ChatGPT said 1Pv8 could be used with KM but only if you do the following:

Question: Is the app 1Password version 8 now compatible with Keyboard Maestro?

Yes, 1Password version 8 is compatible with Keyboard Maestro, though the integration methods differ from previous versions. Direct AppleScript support has been reduced in 1Password 8, but you can still achieve integration through alternative approaches:

1. Using the 1Password Command-Line Interface (CLI)

The 1Password CLI (op) allows you to access your vaults and items via terminal commands, which can be incorporated into Keyboard Maestro macros. Here’s how you can set this up:

  1. Install the 1Password CLI:

• Download and install the CLI from the 1Password developer website.

  1. Authenticate the CLI:

• Open Terminal and sign in using:

op signin

• Follow the prompts to authenticate.

  1. Create a Keyboard Maestro Macro:

• In Keyboard Maestro, create a new macro that executes a shell script.

• Use the op commands to fetch the desired data. For example, to retrieve a password:

op item get "Item Name" --fields password

• Ensure the macro securely handles any sensitive information.

For a detailed walkthrough and example macros, refer to the Keyboard Maestro forum.

2. Simulating User Input

If you prefer not to use the CLI, you can create macros that simulate user input to interact with the 1Password 8 interface:

  1. Unlocking 1Password:

• Use Keyboard Maestro to simulate keystrokes that bring 1Password to the foreground and enter your master password.

• Example steps:

  1. Activate 1Password.

  2. Pause for a short duration (e.g., 0.5 seconds).

  3. Type your master password.

  4. Simulate pressing the Return key.

• Note: Ensure this method aligns with your security preferences, as it involves storing your master password within a macro.

  1. Accessing Specific Items:

• Simulate keyboard shortcuts to search for and copy credentials:

  1. Activate 1Password.

  2. Simulate the search shortcut (e.g., Command + F).

  3. Type the item’s name.

  4. Simulate the copy password shortcut (e.g., Command + Shift + C).

For more insights and user experiences, you can visit the Keyboard Maestro forum discussion.

Caution: When automating interactions with password managers, always prioritize security. Ensure that any scripts or macros do not expose your sensitive information.

In summary, while 1Password 8’s integration with Keyboard Maestro requires different methods than previous versions, utilizing the 1Password CLI or simulating user input are effective approaches to maintain workflow automation.

27

I appreciate the legwork. That suggests that nothing has really changed in the last year. I'm mainly concerned about whether the slowness problems involving KM and 1P8 persist.

1 Like
28

It may persist for you -- I've had no such problems with KM (v10 and v11) and 1Pass (v7 and v8). Which, unfortunately, isn't much help to you (though it may reassure others).

If you still have 1Pass v7 you can try the fix above. If not you may be able to track the file down and nuke-from-orbit.

2 Likes
29

Thanks. I'll see how things go.

30

Be careful – I don't think this forum should be used to pass on strings generated by an auto-complete system as if it had any theory whatsoever of what it was "saying".

If users want to search with LLMs or Google, they can very easily do that themselves.

1 Like
31

We've discussed this before. As long as this forum doesn't prohibit quoting LLMs, I will continue to do so, despite your opposition to AI. But if the rules change, I will 100% comply. Have you been advocating for a change to the rules?

32

Even before avoiding the problem that any auto-generated LLM strings placed on a public website becomes input to LLM building, degrading the web to autotrophic mush, we need to consider the quality of the kind of advice that we are likely to be giving unfortunate users when we preface anything with ChatGPT "said"

I just asked it

How confident are you that you are not misleading users with these strings sourced from an LLM ?

Does that seem like a trivial question, when something like a password manager is involved.

Let's "ask" chatGPT for the names of the parts of a bicycle, and see what it "says":

Create a line-drawn image of a bicycle, labelling each of its parts clearly.

Result (this is the kind of thing that you are passing on when you say ChatGPT "says"):

DALL·E 2025-02-19 18.04.06 - A detailed line-drawn illustration of a bicycle, with each part clearly labeled. The drawing should include key components such as the frame, handleba
DALL·E 2025-02-19 18.04.06 - A detailed line-drawn illustration of a bicycle, with each part clearly labeled. The drawing should include key components such as the frame, handleba1024×1024 418 KB

3 Likes
33

As long as I warn people that the information comes out of an AI, and I did so very clearly in this case, I'm not misleading anyone. Not only that, but part of my post was my own research. I wasn't just parroting AI.

As I've said, we discussed this before, and I'm following the rules of this website. I don't need to follow your personal rules. I keep encouraging you to try to change the rules of this website to meet your personal standards, but you seem to be failing to do that, and you seem to want to persuade people on an individual basis to stop quoting AI. Why make your life so difficult? Just get the rules changed, and everyone will follow the rules and you will be happy and your life will be easier.

34

Perhaps a bicycle is too advanced a concept to be reliable about ?

Let's try this:

Create an image of six basic geometric shapes, labelling each of them clearly.

DALL·E 2025-02-19 18.14.24 - A clean, line-drawn illustration of six basic geometric shapes, each clearly labeled. The shapes should include a circle, square, triangle, rectangle,
DALL·E 2025-02-19 18.14.24 - A clean, line-drawn illustration of six basic geometric shapes, each clearly labeled. The shapes should include a circle, square, triangle, rectangle,1024×1024 265 KB

A token distribution statistics model doesn't even know what a token stream is. That doesn't stop suppliers from packaging it with expressions of misleading confidence. Under the nonsense above, it helpfully emitted a follow-up string:

"Here is a line-drawn illustration of six basic geometric shapes, 
each clearly labeled."

"ChatGPT said" ?

You have to be joking.

Is passing these strings on any more than thoughtless pollution of the commons ?

4 Likes
35

Now I see the difference between a recangle and a renatgle.

2 Likes
36

( Very hard to discern. Fortunately, we are assisted by Artificial Intelligence tm )

( Simply indispensible. Puzzling that those who have invested in it so heavily are reporting no benefits, so far ... )

37

Perhaps you might appreciate this blog post (or perhaps you've already read it). I personally found it somewhat carthartic to read someone else's rage on the topic.
https://ludic.mataroa.blog/blog/i-will-fucking-piledrive-you-if-you-mention-ai-again/

1 Like
38

"AI-generated answer. Please verify critical facts." – Brave search.

I am not as sceptical as @ComplexPoint about what this particular field in AI might offer, but I have to say that, with the current state of play, I would not treat automatically compiled answers as anything other than a mash-up of sources, useful for guesses, inspiration or hints about what the undistorted facts may be.

By chance and algorithm, I saw this just a little earlier. It's not a wholly original point, but the wording amused me.

The Google AI results at the top of searches are so pointless. We have a resource (the internet!) that knows everything, and instead of going to the source, it's like you asked your stupid cousin who is trying to recall the information from a podcast he heard last week.

39

and at least your unfortunate and over-stretched cousin has some concepts ...

( LLMs only have token distribution statistics. They don't even have a clue about what tokens and token streams might be – let alone about anything else )

40

Yes, the cousin's mistakes would have been of a different nature than those of generative AI! The question of concepts evokes Searle's Chinese room argument, a philosophical position on strong AI. It's a big question as to whether we could ever expect weak AI algorithms ever to to yield trustworthy summaries on the basis of LLMs and the like.

Such techniques can yield somewhat useful results, but they are useful only after human scrutiny and evaluation.

"Auto-generated based on listed sources. May contain inaccuracies." – DuckDuckGo.

41

This has gone way off topic. Who has the permission to branch it?

1 Like