Gremllm • Hacker News Board

Re: Gremllm

#11

Post by unit149 » Fri, Jul 04, 2025, 11:28 PM UTC

[dead]

Re: Gremllm

#12

Post by lxgr » Sat, Jul 05, 2025, 12:03 AM UTC

Amazing!

Shameless plug: People who love-hate this might also love-hate vibeserver (https://github.com/lxgr/vibeserver).

I had some vague plans to make it self-hosting; this might make that even lower effort :)

Re: Gremllm

#13

Post by awwaiid » Sat, Jul 05, 2025, 1:25 AM UTC

mpalmer wrote: post #9

Love it, I am here for exactly this sort of playful boundary nudging. "Wet mode" is such a fantastically awful name. Definitely make me think twice about turning it on.

You know... Because if you get them wet they multiply, per the documentaries

Re: Gremllm

#14

Post by afcool83 » Sat, Jul 05, 2025, 2:08 AM UTC

awwaiid wrote: post #13

mpalmer wrote: post #9

Love it, I am here for exactly this sort of playful boundary nudging. "Wet mode" is such a fantastically awful name. Definitely make me think twice about turning it on.

You know... Because if you get them wet they multiply, per the documentaries

…I read the whole article at OP’s link, many comments off this thread…I even clicked into the college course material in https://news.ycombinator.com/item?id=44468452 …and not once did it occur to me why it was called “wet mode”…not once…

…until your comment. Here! Take my “lived through the 80’s and 90’s” card.

Re: Gremllm

#15

Post by vunderba » Sat, Jul 05, 2025, 2:10 AM UTC

jmsdnns wrote: post #10

awwaiid wrote: post #5

I was chatting with Simon Willison (who's LLM library I use to power gremllm) on Discord and he suggested D&D use-cases. Kinda works!!! >>> from gremllm import Gremllm >>> player = Gremllm('dungeon_game_player') >>> player.go_into_cave() 'Player has entered the cave.' >>> player.look_around() {'location': 'cave', 'entered_cave_at': '2025-07-02T21:59:02.136960'} >>> player.pick_up_rock() 'You picked up a rock.' >>> pl…

i helped Chris Callison-Burch design a class at upenn, called interactive fiction, which is a similar context to what Simon suggested. the real magic is that it reframes hallucinations as creative story telling. the usecase is SUPER fun if you imagine the LLM as a dungeon master telling a story that gets expanded over time. the framework he and I built kept track of the game state over time and allowed saving and loa…

Sounds similar to AI Dungeon which I believe ran on a fine-tuned version of GPT-2 "all the way" back in 2019. And honestly kind of reminded me of the "Mind Game" in the novel, Ender's Game.

https://en.wikipedia.rg/wiki/AI_Dungeon

Re: Gremllm

#16

Post by cofob_ » Sat, Jul 05, 2025, 5:11 AM UTC

I implemented something similar a couple months ago, but the principle was based on inspecting the code of the calling party.

https://github.com/cofob/autogenlib

Re: Gremllm

#17

Post by taneq » Sat, Jul 05, 2025, 5:25 AM UTC

This is horrifying. Please, go on. :D

How do I give it a base URL for API calls so I can point it at my ollama server?

Re: Gremllm

#18

Post by awwaiid » Sat, Jul 05, 2025, 1:07 PM UTC

afcool83 wrote: post #14

awwaiid wrote: post #13

Earlier quoted context omitted.

You know... Because if you get them wet they multiply, per the documentaries

…I read the whole article at OP’s link, many comments off this thread…I even clicked into the college course material in https://news.ycombinator.com/item?id=44468452 …and not once did it occur to me why it was called “wet mode”…not once… …until your comment. Here! Take my “lived through the 80’s and 90’s” card.

Hm. I'll improve the documentation to make it slightly more obvious. When I add after_midnight with malicious compliance mode and bright_light() to freeze implementations then it should be more obvious.

Re: Gremllm

#19

Post by awwaiid » Sat, Jul 05, 2025, 7:37 PM UTC

taneq wrote: post #17

This is horrifying. Please, go on. :D How do I give it a base URL for API calls so I can point it at my ollama server?

It is using the llm library, so you do the plugin and model management through that. Let's say you've already gotten ollama installed and the `gemma3n:e2b` model. Then you use the llm cli to add the ollama extension:

  llm install llm-ollama

and then you use whatever model you like, anything llm has installed. See https://llm.datasette.io/en/stable/plugins/installing-plugin... for plugin install info.

Here is a sample session. You can't see it ... but it is very slow on my CPU-only non-apple machine (each response took like 30 seconds) :)

  >>> from gremllm import Gremllm
  >>> counter = Gremllm("counter", model="gemma3n:e2b")
  >>> counter.value = 5
  >>> counter.increment()
  >>> counter.value
  None
  >>> counter.value
  None
  >>> counter.value
  None
  >>> counter.what_is_your_total() 
  6
  6

... also I don't know why it kept saying my value is None :) . The "6" is doubled because one must have been a print and the other is the return value.

Re: Gremllm

#20

Post by afcool83 » Mon, Jul 07, 2025, 4:21 PM UTC

awwaiid wrote: post #18

afcool83 wrote: post #14

Earlier quoted context omitted.

…I read the whole article at OP’s link, many comments off this thread…I even clicked into the college course material in https://news.ycombinator.com/item?id=44468452 …and not once did it occur to me why it was called “wet mode”…not once… …until your comment. Here! Take my “lived through the 80’s and 90’s” card.

Hm. I'll improve the documentation to make it slightly more obvious. When I add after_midnight with malicious compliance mode and bright_light() to freeze implementations then it should be more obvious.

Nah, dude. It’s exactly the right amount of subtle that I found it a delight to discover. Making the joke too obvious undercuts it!