I have to say, after I played with it for a while, it's rather disappointing. The usage of Ollama makes the performance far worse that it could otherwise be and the default models never work correctly. Sometimes they work, sometimes they just decide to give you no letters or a letter or 2-3 words and then stop generating . Nothing but closing the app can save the performance at that point. The available models are also far from the best RP models out there, with many being neither RP nor Story models. The website is decent but the promise of a desktop application is not fulfilled by this app. It also doesn't help that the Ollama back end is slower both to first token and to generate than alternatives like KoboldCCP and LM Studio. Overall, a disappointing program that would probably have potential were it not so fundamentally based around Ollama.
But this is just after a week of random usage, I'm sure other people that have been using it for longer might be able to give both more constructive criticism and more pertinent reviews. Better than Sillytavern on account of being actually designed for desktop/laptop use as opposed to server usage, but that's about it.
Yes and no. On the one hand when the models when they run they do run better, on the other hand it's still impossible to run the model set as character model and I have to use the custom model option. The good news is that it does run faster and that at least on custom I haven't encountered the issues I had before where it just wound't run.
BTW, did you put the linux AMD ROCm support in just for me as your one known linux user or is HammerAI actually detecting that I have an AMD AI capable CPU with iGPU besides the Nvidia GPU it's using right now? Because if the latter that's actually impressive - ROCm support is so spotty on linux is might as well not be there. The 780m from AMD is a lot weaker than the 4060 so I don't think It will see much usage, but I might try bigger models just to see how it behaves if Hammer can actually use the AMD iGPU natively.
PS. Please add a few newer RP models. Some of your competitors have a few finetunes that are open source licenses. ArliAI, Latitudegames, Dreamgen. Please add a few newer Nemo finetunes. Also, and this is just up to you, consider IQ quants and do not offload KV cache to VRAM. IQ quants can make 8GB fully enough to 100% offload most non-gemma 12B models, as long as one doesn't also try to offload KV cache to the Vram. That's in case you're not doing this already.
Anyways, cheers and thanks for the new Ollama update, it did in fact help.
Glad to hear it's better! I really need to get this update out to users. But there is one bug I know about that before I can launch.
Linux AMD ROCm support was just in case I had any users, I wanted to make sure it was awesome for them! Glad to hear that day happened faster than I expected.
Will definitely add some more models, I'm pretty behind. Any specific suggestions? Would love to hear what the best stuff is nowadays.
I will learn more about IQ quants and the KV cache offloading. Is that suggestion for the local LLMs, or the cloud-hosted ones?
Anyways, happy it's better. If you want to chat more, I'm hammer_ai on Discord - would be fun to chat more about finetunes to add / any other suggestions you have.
For the desktop version. Basically, there are two way to use AI, one with slowly building KV cache over yeach new prompt and reply and one where the user sends the entire conversation back to the AI to be processed with each new prompt to get a new response. On desktop it's faster to use KV cache than it is to reprocess the entire conversation again and again. Thing is, the KV cache can be separate from from rest of the model. IF the offload to VRAM option is used and there is enough VRAM it's always faster BUT if there isn't enough VRAM for the desired KV cache size then part of the model and part of the KV cache are in VRAM and the rest are in RAM and this is always slower.
If you can fit everything in VRAM you're at say 21 tokens per second, with only the model in VRAM and KV cache in RAM you'd be at around 15, and with part of the model and part of the KV cache in VRAM and the rest in RAM you could go as low as 10 or even 5 tokens per second. So it's always preferable to only load the model in VRAM and let the KV cache in RAM if you can't fit everything in VRAM. For the website, since you're only using 4k context window as long as everything fits in VRAM, I wouldn't touch it - if it ain't broke don't fix it and whatnot. But on desktop, allowing us to keep KV cache only in RAM or offload it to VRAM, can significantly increase performance.
As for recommended models. I'd say move the Nous models to Hermes 3 (nonthinking) look into the ArliAI RPMax v.1.3 series of models (4 models at 4 sizes, based on 3 different bases, Llama 3.1, qwen 2.5 and Mistral Nemo), and the latest Latitudegames models. I'm using Wayfarer 12b for RP and Muse 12b for story writing (latitudegames models) but they have larger models and again, all open source and on Huggingface. Dreamgen is also doing interesting stuff, but their older stuff is, well, older, and the new model - Lucid - is still in beta and fairly bad at following instruction.
But yeah, try Wayfarer, at least for me it's significantly superior to the Drummer Rocinante you have as a default option. I get actual RP responses from it wile Rocinante 12b wants to just continue my own posts 90% of the time. Also, I'd probably remove the thinking models from the default options. Honestly, most people are not going to have the kind of hardware to run them at high enough speeds to make the thinking steps worth it - at least not on desktop. Especially for smaller models that even unquantize can still catch themselves in an infinite thinking loop.
Overall, I'd try to find finetunes and test them if I were you. What I recommended is what I tested and found to be an improvement over what came before. I'd stay away from mergers, ablated and uncensored models. Just try to find RP and story finetunes that are open source and on huggingface to test. Also, and you did not hear this from me, try IBM's Granite 3.3 8B model... for a model designed for office work and which was instruction trained to be harmless and safe boy does it follow nsfw instructions well. And I do mean NSFW. And it's Apache 2.0 :)
As for IQ quants, they can offer similar performance to KV cache at smaller sized - but are only similarly fast under ROCm and CUDA - significant slowdowns under Vulkan and CPU. I know Ollama supports them, though I don't think you can DL IQ quant from their site directly. An IQ4_XS should be very similar to a Q4_KS in output - within margin or error for RP and story purposes - but substantially smaller.
I wrote earlier that I couldn't get the characters to respond offline. I now have it working. I noticed on reddit some had the same problem. What I did was delete the default model and down the lemon one. The characters all now respond in chat offline. Awesome program. If there is a better payment plan or perhaps donation set I'd be interested in donating a little.
I can look into putting it on flathub, but I don't have a linux machine, so just haven't actually tested any of the Linux apps myself. Sorry about that. Anything I need to fix with them?
Not really, they work. I would probably prefer a more LM Studio like approach where we can select if we want to run local models via CUDA, Vulkan or CPU but no, the applications works.
That being said, just out of curiosity, aren't you running the website on a linux servers? I mean it's probably a docker container but still, I'd be actually shocked if you told me you're running the website on a windows servers.
The website is linux servers + Runpod. But it's a different codepath than the desktop app. So doing the website doesn't really help with the Electron app.
"Installation has failed: There was an error while installing the application" Nothing in the log and subsequent attempts to install fail as well, as trying to install previous versions.
I've done that and more: terminated processes, uninstalled, reinstalled, nothing helps. Does your installer leave anything in the registry? It looks like a failed installation will leave something that doesn't get cleaned up. I can't seem to find it.
i got a small feedback. Could you also make it, that you can not only create a person as a character but also an RP? like with scenery and not only the one character based creation
so im using this offline but the ai text back really slow one word at a time it takes a few mins for it to say a line or to is there a way i can fix this and speed it up so im not w8 so long for a short reply ?
i some what fix it but after the up date im getting the ai randomly saying <[im end]> <[im start}> <[im done]> thin it just go crazy and go on about its own story and makes me and character go on a long thing making both sides talk when i dint type . what is going on and how can i fix ?
So I have it mostly working already, but not yet polished up. I do want it really badly, and will update you when it's ready. Sorry for the long delay.
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: File 'C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Accessibility\LiveCaptions.lnk' could not be converted into a valid ShellLink: System.IO.FileNotFoundException: The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
at Squirrel.Shell.ShellLink.IShellLinkW.Resolve(IntPtr hWnd, UInt32 fFlags)
at Squirrel.Shell.ShellLink.Open(String linkFile, IntPtr hWnd, EShellLinkResolveFlags resolveFlags, UInt16 timeOut)
at Squirrel.UpdateManager.ApplyReleasesImpl.<fixPinnedExecutables>b__11_0(FileInfo file)
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: File 'C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Accessibility\Narrator.lnk' could not be converted into a valid ShellLink: System.IO.FileNotFoundException: The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
at Squirrel.Shell.ShellLink.IShellLinkW.Resolve(IntPtr hWnd, UInt32 fFlags)
at Squirrel.Shell.ShellLink.Open(String linkFile, IntPtr hWnd, EShellLinkResolveFlags resolveFlags, UInt16 timeOut)
at Squirrel.UpdateManager.ApplyReleasesImpl.<fixPinnedExecutables>b__11_0(FileInfo file)
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: File 'C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Accessibility\On-Screen Keyboard.lnk' could not be converted into a valid ShellLink: System.IO.FileNotFoundException: The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
at Squirrel.Shell.ShellLink.IShellLinkW.Resolve(IntPtr hWnd, UInt32 fFlags)
at Squirrel.Shell.ShellLink.Open(String linkFile, IntPtr hWnd, EShellLinkResolveFlags resolveFlags, UInt16 timeOut)
at Squirrel.UpdateManager.ApplyReleasesImpl.<fixPinnedExecutables>b__11_0(FileInfo file)
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: File 'C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Accessibility\VoiceAccess.lnk' could not be converted into a valid ShellLink: System.IO.FileNotFoundException: The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
at Squirrel.Shell.ShellLink.IShellLinkW.Resolve(IntPtr hWnd, UInt32 fFlags)
at Squirrel.Shell.ShellLink.Open(String linkFile, IntPtr hWnd, EShellLinkResolveFlags resolveFlags, UInt16 timeOut)
at Squirrel.UpdateManager.ApplyReleasesImpl.<fixPinnedExecutables>b__11_0(FileInfo file)
Sorry about that! It's a weird issue people get into related to Discord interfering with our update. The solution is to either restart your computer or kill all "HammerAI" and "Ollama" processes under Task Manager. Sorry again.
everytime i say hi the ai breaks chraracter and be a another chatgpt clone can you fix that i dont like paying to just have a roleplay thank you besides it is a little paywall like
Hi! Sorry about that, usually that's because the character you're chatting with wasn't written very well. If you try with one of these is it any better? https://www.hammerai.com/characters?tag=Featured
In terms of paywall like, it is 100% free to chat with the cloud-hosted LLM Smart Lemon Cookie 7B, or with any local LLM! But I am a solo dev building this, so I made saving chats is a paid feature, sorry about that. If it makes you feel better, I spend all the money I make to pay other contractors to help build it with me. And I have a 100% no questions asked refund policy, if you end up not being happy with it. Again, sorry for the issues.
Yes, it is safe! But you don't need to take my word for it, you can also ask in the Discord, or maybe read this review from someone on Reddit? They said:
> All in all it is one of the best options for a locally installed AI chatbot to use privately. Using wireshark, iftop, and other tools I didn't notice any unnecessary calls or shady traffic. Which is awesome. However, please be aware you lose some of that privacy as you need to log to discord to access basic docs for the app.
Didn't get no response but my RAM and disk went both 100% and after a while I lost 20GB of my C disk space and i cannot figure out where from since the hammer ai folder is only 3 GB. How is that cause I downloaded a model in D disk?
Sorry! We moved over to Discord to chat, tl;dr for others is to check C:\Users\<yourname>\AppData\Roaming\HammerAI and C:\Users\<yourname>\AppData\Local\HammerAI
Hi, sorry! I need more information to help you though 😂 I.e. what exactly can't install? Would also be great if you want to join the Discord so we can chat more quickly!
Hello, when they will update again, I am waiting for an update that adds image generation and that is faster, since I have a problem, when I send the first message in the chat it freezes the screen and closes the App, I do not have a good processor but to use an AI it must be sufficient
Hi! We've started on image generation, though no guarantee on timing. Regarding freezing, can you use Ollama by itself (https://ollama.com/download)? If not, then the issue is just that you need a more powerful computer :( If so, would be awesome if you want to join the Discord where I can help debug and get it working for you!
It's weird that I cannot download the default model or other even though I downloaded the software in my computer (；へ：) and I can't even download the model in the web version (ಥ_ಥ)
I cannot add image here (I don't know why), if you need screenshots, please send me an email
Shoot! Well if you want to try the new beta, maybe it will work for you? It's a totally new way of using the LLM which should have WAY fewer errors! You can download it directly here: https://github.com/hammer-ai/hammerai/releases/tag/v0.0.97
Sorry for not replying to you for so long.ﾍ(;´Д｀ﾍ)
I spent a lot of times downloading the <HammerAI-0.0.97.Setup>, I can open it but it just shows "Installation has failed: There was an error while installing the application." When I want to open and check the setup log, it just disappeares (；д；).
I think I'm destined to be unusing your software (╥╯^╰╥), but I still appreciate what you have done for me and I really really hope you will succeed in the future!*\(๑• ₃ •๑)
Hope you have a wonderful day. (Sorry for my bad english, I'm from CN and I use the translator. (；へ：)
Hi, so sorry for this! But I have good news for you - I just rolled out a totally new LLM set up which is way more stable and should work on (almost) any computer. If you want to give it one more try I think it might work hopefully :) You can download it directly here: https://www.hammerai.com/desktop
probably wouldn't have bothered downloading it if i knew you had to subscribe for 9 bucks a month to be able to keep conversations. They just delete every time you leave the conversation.
Sorry about that :( If it helps you feel better, I use all the money I make to pay other devs to help work on the project with me. It was just getting too much for me to manage all alone with everything going on in my life. A top requested feature which I plan to add is that everyone can always have one conversation saved for free.
question will there be image generation in the future and if so will it be paid or free (what id do is have 2 versions a lesser quality one that is free and a higher quality one that is paid) also will there be a feature to save the chat on your device
Hi! No immediate plans for image generation, but definitely watching the space. If I see a nice library to help with this that is cross-platform I will add it. Not sure on pricing yet, but will definitely let you know!
For saving chats, we have that on Desktop currently! And are working on bringing it to Web as well. And when we roll out Mobile we'll have it.
Sometimes when I'm chatting, the answers starts to generate slower and then it just stops generating. Is there anyway to fix that? My GPU is not the best, so maybe that's it?
Mmm that could definitely be part of it, but the other is that our logic for long chats will lead to many tokens that need to get re-evaluated. Have been looking into fixing this, but no work around yet, sorry about that. One thing that helps is to use characters with less text in them (i.e. their personality is shorter). Will let you know when we have a better fix for this, sorry again!
Hi sorry! That is a known bug. The immediate fix is to turn off GPU acceleration under Settings. If that doesn't work, could you join the Discord so we can chat more and try to fix it for you?
Hi, that means we weren't able to detect your GPU correctly and so started using just CPU. We'll be rolling out updates in the future which should help here, sorry about that!
Hello, I bought the pro plan for a month and cancelled it right after I got the license (because with subscriptions I know that I'll forget to cancel at the end of the month). Now money is gone from my bank and it says that my pro license is cancelled. Can you help me out?
Oh, so sorry, that looks like a bug where we stop accepting licenses if the subscription is cancelled. Please DM me and I'll refund you for your order while fixing this. If you want to buy it again so it works for the month I'll also refund that (i.e. it will be free for you until I fix this).
Hello, this is a good tool, I wonder if you are going to translate it to Spanish or it is very difficult, I am not very good in English so I can't make the most use of it, this message is passed by translator
Hi, I really like this tool and I would like to have the "Chats saving" feature because it's a must have! However I can't really afford that and I think the price is a little expensive (108€ a year with taxes) and I would have preferred a one time payment choice (even if a little more expensive). So I really hesitate...
1. If one day I decide to buy the pro, how can I be sure that the servers where my characters and their profile images are stored will not shut down one day (for one reason or another). (Because If I understand correctly, even locally stored character share their profile picture and description to the servers, no?)
2. This error can happen sometimes (from my logs):
Then, the character no longer responds, I have to reload it. If I get the "Chats saving" feature, I wonder if the conversation will be able to resume after this kind of error.
That's a lot of question I know, thanks and sorry for my imperfect English!
Hi, glad you like it! Yeah I'm thinking a lot about a lifetime license because others have asked for it. What would you be able to pay for it? I'm just not quite sure how to price it.
For your questions:
1. When you save characters locally the description and images are all saved on your computer, so it will work forever! You can click here to see exactly what we save and where it is saved:
2. Ah, that is a bug with longer conversations. I don't yet have a fix, but I have also seen it and will try to fix. Sorry about that.
Anyways, thank you for the feedback and support. I will let you know when I fix this bug, and no worries if you do not want to pay for the pro version, I understand it is a lot of money.
I don't know about the price, but between 60 and 110 euro/dollar for a lifetime license could be a good compromise between your need for programmer funds and the limited means of certain users and also the limited size of chats even if saved (see bugs below). Fyi: I just purchased a short time licence today to test the PRO features and I like it! But I am a little bit uncomfortable knowing that there is automatic bank withdrawal each end of cycle. So, yes a lifetime license could be great (as long as there is no kind of Pro + version in the future) :p
TWO BUGS:
1. The dialogue generation can sometimes stop (especially on long sentences), it's not a really big deal because you just have to try again but it's annoying.
2. Sometimes during a totally normal conversation (long but not always), the character start rambling by mixing verbs and words from some old (or not) messages. IE: "Oh no, maybe it's then yes I said it I said, said it!, but now we are here to, i can it! I think it's alright too!" or unfinished smaller strange one like "(giggling*". After some verifications, it look like it's a problem with exceded tokens limitation. It could be great to have the possibility to set it to 16384 or even 32768, because conversation are short (even with 8192 size). :/
Ah, I see, I have been asked about a "continue generating feature", so that is on the roadmap. And yes, sorry about the bug with the rambling, that's known and I need to fix it still.
Also, just to offer, I have a 100% refund policy if you're unhappy with the Pro features, as the goal is not to make money, it's just to fund continued development. So if you're not happy just email / DM me your details on Discord and I will refund you. Would rather not make the money instead of making people unhappy.
No no, it's ok, thanks for your concern. However I would be interrested to join the discord, (I gave my profile name when ordering but I don't have invitation). :)
Ah sorry! I'm not sure how to directly invite people to the Discord. Instead I just assign people to a role who have purchased. If you join using the link here I'll then add you to the role! https://www.hammerai.com/
I get having to make money to fund development, but I feel that saying that "We will never charge for access to features" is a bit of a lie now that pro is a thing.
That is true, I'm sorry about that 😭. I should have qualified that statement when I made it, though I did really believe it at the time. The issue is that there are a lot of stability improvements and features to make, and I have a lot going on in my life. So my goal with collecting money is to be able to pay a developer to come work on the project. I'm sorry though, you're right that I went back on my word.
I really enjoy this Program, even if i have slight Issues on some Parts, but thats to be expected, i have no clue which Model i should use and which my Pc can handle xD.
maybe you have a Idea how to fix that the Bot starts looping the same Message and gets kind of "stuck"
Thank you! Yeah I've been talking to people in Discord about the looping. Right now my suggestion is to just use a bigger model or try prompting differently. Sorry about that!
Hi, i've downloaded the app, given it permission to access the network, but it is stuck with "loading model"... Am i missing something please, do i need to configure or download something else?
Thank you! Yes I develop it alone, though I am looking for a co-founder, preferably one with React Native experience who wants to bring this to mobile (and add some paid features, so we could make money there). No GitHub today, I have thought about open source but haven't decided to do it yet.
Why is the AI in this feel like it's robotic, and not really flowing, and learning like all limited memory AI? Literally I can say hi to someone and it literally says a set dialogue every time.
Hey! It can really depend on the model you're using, which are you using and have you tried a few? And have you tried modifying the temperature (try increasing it)? That may help make it more creative feeling:
If those don't help it may be the character. If the character prompt length (i.e. personality + scenario + example dialogs + first message) is too long there may not be enough context length left. So trying on some different characters may help things.
PS. If you want to join the Discord we can chat back and forth more easily, would love to help you get this working better.
This looks amazing! Thank you guys! I am curious, though, are you guys intending to integrate image generation into the chats? Such as scenes in a story while roleplaying?
Yeah that would definitely be awesome. I have been keeping my eye on it in case there is an approach I should take. Want to join the Discord and we can chat more about this?
← Return to app
Comments
Log in with itch.io to leave a comment.
I have to say, after I played with it for a while, it's rather disappointing. The usage of Ollama makes the performance far worse that it could otherwise be and the default models never work correctly. Sometimes they work, sometimes they just decide to give you no letters or a letter or 2-3 words and then stop generating . Nothing but closing the app can save the performance at that point. The available models are also far from the best RP models out there, with many being neither RP nor Story models. The website is decent but the promise of a desktop application is not fulfilled by this app. It also doesn't help that the Ollama back end is slower both to first token and to generate than alternatives like KoboldCCP and LM Studio. Overall, a disappointing program that would probably have potential were it not so fundamentally based around Ollama.
But this is just after a week of random usage, I'm sure other people that have been using it for longer might be able to give both more constructive criticism and more pertinent reviews. Better than Sillytavern on account of being actually designed for desktop/laptop use as opposed to server usage, but that's about it.
Thanks for the feedback! So right now our Ollama version is actually really old. Does it work better if you use this beta version? It updates Ollama and should be MUCH faster: https://github.com/hammer-ai/hammerai/releases/tag/v0.0.206
Yes and no. On the one hand when the models when they run they do run better, on the other hand it's still impossible to run the model set as character model and I have to use the custom model option. The good news is that it does run faster and that at least on custom I haven't encountered the issues I had before where it just wound't run.
BTW, did you put the linux AMD ROCm support in just for me as your one known linux user or is HammerAI actually detecting that I have an AMD AI capable CPU with iGPU besides the Nvidia GPU it's using right now? Because if the latter that's actually impressive - ROCm support is so spotty on linux is might as well not be there. The 780m from AMD is a lot weaker than the 4060 so I don't think It will see much usage, but I might try bigger models just to see how it behaves if Hammer can actually use the AMD iGPU natively.
PS. Please add a few newer RP models. Some of your competitors have a few finetunes that are open source licenses. ArliAI, Latitudegames, Dreamgen. Please add a few newer Nemo finetunes. Also, and this is just up to you, consider IQ quants and do not offload KV cache to VRAM. IQ quants can make 8GB fully enough to 100% offload most non-gemma 12B models, as long as one doesn't also try to offload KV cache to the Vram. That's in case you're not doing this already.
Anyways, cheers and thanks for the new Ollama update, it did in fact help.
Glad to hear it's better! I really need to get this update out to users. But there is one bug I know about that before I can launch.
Linux AMD ROCm support was just in case I had any users, I wanted to make sure it was awesome for them! Glad to hear that day happened faster than I expected.
Will definitely add some more models, I'm pretty behind. Any specific suggestions? Would love to hear what the best stuff is nowadays.
I will learn more about IQ quants and the KV cache offloading. Is that suggestion for the local LLMs, or the cloud-hosted ones?
Anyways, happy it's better. If you want to chat more, I'm hammer_ai on Discord - would be fun to chat more about finetunes to add / any other suggestions you have.
For the desktop version. Basically, there are two way to use AI, one with slowly building KV cache over yeach new prompt and reply and one where the user sends the entire conversation back to the AI to be processed with each new prompt to get a new response. On desktop it's faster to use KV cache than it is to reprocess the entire conversation again and again. Thing is, the KV cache can be separate from from rest of the model. IF the offload to VRAM option is used and there is enough VRAM it's always faster BUT if there isn't enough VRAM for the desired KV cache size then part of the model and part of the KV cache are in VRAM and the rest are in RAM and this is always slower.
If you can fit everything in VRAM you're at say 21 tokens per second, with only the model in VRAM and KV cache in RAM you'd be at around 15, and with part of the model and part of the KV cache in VRAM and the rest in RAM you could go as low as 10 or even 5 tokens per second. So it's always preferable to only load the model in VRAM and let the KV cache in RAM if you can't fit everything in VRAM. For the website, since you're only using 4k context window as long as everything fits in VRAM, I wouldn't touch it - if it ain't broke don't fix it and whatnot. But on desktop, allowing us to keep KV cache only in RAM or offload it to VRAM, can significantly increase performance.
As for recommended models. I'd say move the Nous models to Hermes 3 (nonthinking) look into the ArliAI RPMax v.1.3 series of models (4 models at 4 sizes, based on 3 different bases, Llama 3.1, qwen 2.5 and Mistral Nemo), and the latest Latitudegames models. I'm using Wayfarer 12b for RP and Muse 12b for story writing (latitudegames models) but they have larger models and again, all open source and on Huggingface. Dreamgen is also doing interesting stuff, but their older stuff is, well, older, and the new model - Lucid - is still in beta and fairly bad at following instruction.
But yeah, try Wayfarer, at least for me it's significantly superior to the Drummer Rocinante you have as a default option. I get actual RP responses from it wile Rocinante 12b wants to just continue my own posts 90% of the time. Also, I'd probably remove the thinking models from the default options. Honestly, most people are not going to have the kind of hardware to run them at high enough speeds to make the thinking steps worth it - at least not on desktop. Especially for smaller models that even unquantize can still catch themselves in an infinite thinking loop.
Overall, I'd try to find finetunes and test them if I were you. What I recommended is what I tested and found to be an improvement over what came before. I'd stay away from mergers, ablated and uncensored models. Just try to find RP and story finetunes that are open source and on huggingface to test. Also, and you did not hear this from me, try IBM's Granite 3.3 8B model... for a model designed for office work and which was instruction trained to be harmless and safe boy does it follow nsfw instructions well. And I do mean NSFW. And it's Apache 2.0 :)
As for IQ quants, they can offer similar performance to KV cache at smaller sized - but are only similarly fast under ROCm and CUDA - significant slowdowns under Vulkan and CPU. I know Ollama supports them, though I don't think you can DL IQ quant from their site directly. An IQ4_XS should be very similar to a Q4_KS in output - within margin or error for RP and story purposes - but substantially smaller.
I wrote earlier that I couldn't get the characters to respond offline. I now have it working. I noticed on reddit some had the same problem. What I did was delete the default model and down the lemon one. The characters all now respond in chat offline. Awesome program. If there is a better payment plan or perhaps donation set I'd be interested in donating a little.
Nice! I do have this if you'd like, but no need, your nice words are enough! https://www.patreon.com/HammerAI
Stupid question: why provide a flatpack in place an appimage if you're not going use flathub?
Oh, so I just use the default Electron Forge makers: https://www.electronforge.io/config/makers/flatpak
I can look into putting it on flathub, but I don't have a linux machine, so just haven't actually tested any of the Linux apps myself. Sorry about that. Anything I need to fix with them?
Not really, they work. I would probably prefer a more LM Studio like approach where we can select if we want to run local models via CUDA, Vulkan or CPU but no, the applications works.
That being said, just out of curiosity, aren't you running the website on a linux servers? I mean it's probably a docker container but still, I'd be actually shocked if you told me you're running the website on a windows servers.
Okay interesting. I think the new update (which is in beta right now at https://github.com/hammer-ai/hammerai/releases/tag/v0.0.206) might help.
The website is linux servers + Runpod. But it's a different codepath than the desktop app. So doing the website doesn't really help with the Electron app.
"Installation has failed: There was an error while installing the application" Nothing in the log and subsequent attempts to install fail as well, as trying to install previous versions.
Can you try closing and trying again? Maybe turning on and off the computer? Sorry about that.
I've done that and more: terminated processes, uninstalled, reinstalled, nothing helps. Does your installer leave anything in the registry? It looks like a failed installation will leave something that doesn't get cleaned up. I can't seem to find it.
i got a small feedback. Could you also make it, that you can not only create a person as a character but also an RP? like with scenery and not only the one character based creation
Would love if you can join the Discord and post there! Then we can chat more about the feature.
so im using this offline but the ai text back really slow one word at a time it takes a few mins for it to say a line or to is there a way i can fix this and speed it up so im not w8 so long for a short reply ?
You probably don't have a powerful enough computer for the model you chose. Can you try a smaller one?
ant the 8b a small one ? im not sure how that works
i some what fix it but after the up date im getting the ai randomly saying <[im end]> <[im start}> <[im done]> thin it just go crazy and go on about its own story and makes me and character go on a long thing making both sides talk when i dint type . what is going on and how can i fix ?
Could you join the Discord to chat more? We can help you in there!
HI, I'm a developer and I want to know, how I can help you with this project?
Sure, DM me on Discord, hammer_ai is my username. Share your resume + Github + some projects you've worked on, please!
whats the extent of the NSFW content?
Uncensored! You can see the content policy here: https://www.hammerai.com/terms
Hello, could you at least roughly say when local image generation will be available?
So I have it mostly working already, but not yet polished up. I do want it really badly, and will update you when it's ready. Sorry for the long delay.
uninstalled an older version and went to install a newer verion got this
16900> 2024-12-08 05:31:10> Program: Starting Squirrel Updater: --install .
16900> 2024-12-08 05:31:11> Program: Starting install, writing to C:\Users\jrski\AppData\Local\SquirrelTemp
16900> 2024-12-08 05:31:11> Program: About to install to: C:\Users\jrski\AppData\Local\Discord
16900> 2024-12-08 05:31:11> SingleGlobalInstance: Grabbing lockfile with timeout of 00:00:10
16900> 2024-12-08 05:31:11> CheckForUpdateImpl: Reading RELEASES file from C:\Users\jrski\AppData\Local\SquirrelTemp
16900> 2024-12-08 05:31:11> CheckForUpdateImpl: Remote version 1.0.9059 differs from local
16900> 2024-12-08 05:31:11> CheckForUpdateImpl: First run or local directory is corrupt, starting from scratch
16900> 2024-12-08 05:31:11> ApplyReleasesImpl: Writing files to app directory: C:\Users\jrski\AppData\Local\Discord\app-1.0.9059
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\chrome_200_percent.pak to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\chrome_200_percent.pak
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\chrome_100_percent.pak to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\chrome_100_percent.pak
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\app.ico to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\app.ico
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\d3dcompiler_47.dll to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\d3dcompiler_47.dll
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\Discord.exe.sig to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\Discord.exe.sig
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\Discord.exe to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\Discord.exe
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\ffmpeg.dll to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\ffmpeg.dll
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\icudtl.dat to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\icudtl.dat
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\installer.db to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\installer.db
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\libEGL.dll to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\libEGL.dll
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\libGLESv2.dll to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\libGLESv2.dll
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\resources.pak to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\resources.pak
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\snapshot_blob.bin to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\snapshot_blob.bin
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\v8_context_snapshot.bin to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\v8_context_snapshot.bin
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\vk_swiftshader_icd.json to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\vk_swiftshader_icd.json
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\vk_swiftshader.dll to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\vk_swiftshader.dll
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\updater.node to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\updater.node
16900> 2024-12-08 05:31:14> ApplyReleasesImpl: Moving file C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\lib\net45\vulkan-1.dll to C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\vulkan-1.dll
16900> 2024-12-08 05:31:15> ApplyReleasesImpl: Squirrel Enabled Apps: [C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\Discord.exe]
16900> 2024-12-08 05:31:19> Utility: Process Started: C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\Discord.exe --squirrel-install 1.0.9059, pid 9240
16900> 2024-12-08 05:31:30> Utility: Received exitcode 0 from process C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\Discord.exe
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: ran C:\Users\jrski\AppData\Local\Discord\app-1.0.9059\Discord.exe, pid 18364
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: Starting fixPinnedExecutables
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: fixPinnedExecutables: newCurrentFolder: app-1.0.9059
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: File 'C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Accessibility\LiveCaptions.lnk' could not be converted into a valid ShellLink: System.IO.FileNotFoundException: The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
at Squirrel.Shell.ShellLink.IShellLinkW.Resolve(IntPtr hWnd, UInt32 fFlags)
at Squirrel.Shell.ShellLink.Open(String linkFile, IntPtr hWnd, EShellLinkResolveFlags resolveFlags, UInt16 timeOut)
at Squirrel.UpdateManager.ApplyReleasesImpl.<fixPinnedExecutables>b__11_0(FileInfo file)
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: File 'C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Accessibility\Narrator.lnk' could not be converted into a valid ShellLink: System.IO.FileNotFoundException: The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
at Squirrel.Shell.ShellLink.IShellLinkW.Resolve(IntPtr hWnd, UInt32 fFlags)
at Squirrel.Shell.ShellLink.Open(String linkFile, IntPtr hWnd, EShellLinkResolveFlags resolveFlags, UInt16 timeOut)
at Squirrel.UpdateManager.ApplyReleasesImpl.<fixPinnedExecutables>b__11_0(FileInfo file)
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: File 'C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Accessibility\On-Screen Keyboard.lnk' could not be converted into a valid ShellLink: System.IO.FileNotFoundException: The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
at Squirrel.Shell.ShellLink.IShellLinkW.Resolve(IntPtr hWnd, UInt32 fFlags)
at Squirrel.Shell.ShellLink.Open(String linkFile, IntPtr hWnd, EShellLinkResolveFlags resolveFlags, UInt16 timeOut)
at Squirrel.UpdateManager.ApplyReleasesImpl.<fixPinnedExecutables>b__11_0(FileInfo file)
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: File 'C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Accessibility\VoiceAccess.lnk' could not be converted into a valid ShellLink: System.IO.FileNotFoundException: The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
at Squirrel.Shell.ShellLink.IShellLinkW.Resolve(IntPtr hWnd, UInt32 fFlags)
at Squirrel.Shell.ShellLink.Open(String linkFile, IntPtr hWnd, EShellLinkResolveFlags resolveFlags, UInt16 timeOut)
at Squirrel.UpdateManager.ApplyReleasesImpl.<fixPinnedExecutables>b__11_0(FileInfo file)
16900> 2024-12-08 05:31:30> ApplyReleasesImpl: Updating shortcut C:\Users\jrski\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Discord Inc\Discord.lnk
16900> 2024-12-08 05:31:31> ApplyReleasesImpl: Updating shortcut C:\Users\jrski\OneDrive\Desktop\Discord.lnk
16900> 2024-12-08 05:31:31> ApplyReleasesImpl: Fixing up tray icons
16900> 2024-12-08 05:31:31> ApplyReleasesImpl: cleanDeadVersions: for version 1.0.9059
16900> 2024-12-08 05:31:31> ApplyReleasesImpl: cleanDeadVersions: exclude current version folder app-1.0.9059
any help would be appreciated so i can install a newer version
Sorry about that! It's a weird issue people get into related to Discord interfering with our update. The solution is to either restart your computer or kill all "HammerAI" and "Ollama" processes under Task Manager. Sorry again.
You need to start hammer Ai as admin. then the downloade works for me. otherwise same issue here
When is it going to be available for phones? Just wanna know, not stress you out or something.
No specific timeframe, sorry! As a one-person project, I can only do so much 😭
PS. If I could find someone to work with me, I'd definitely go faster, so if anyone reading here is a dev, please reach out!
can you make everything free pls
everytime i say hi the ai breaks chraracter and be a another chatgpt clone can you fix that i dont like paying to just have a roleplay thank you besides it is a little paywall like
Hi! Sorry about that, usually that's because the character you're chatting with wasn't written very well. If you try with one of these is it any better? https://www.hammerai.com/characters?tag=Featured
In terms of paywall like, it is 100% free to chat with the cloud-hosted LLM Smart Lemon Cookie 7B, or with any local LLM! But I am a solo dev building this, so I made saving chats is a paid feature, sorry about that. If it makes you feel better, I spend all the money I make to pay other contractors to help build it with me. And I have a 100% no questions asked refund policy, if you end up not being happy with it. Again, sorry for the issues.
Hello, what is the memory system like on this? Does it forget things or does it remember everything?
It depends! The lower parameter (i.e. 7B and 8B) models forget more, and the higher (20B and 70B) forget less.
Can I trust it?
Yes, it is safe! But you don't need to take my word for it, you can also ask in the Discord, or maybe read this review from someone on Reddit? They said:
> All in all it is one of the best options for a locally installed AI chatbot to use privately. Using wireshark, iftop, and other tools I didn't notice any unnecessary calls or shady traffic. Which is awesome. However, please be aware you lose some of that privacy as you need to log to discord to access basic docs for the app.
https://www.reddit.com/r/HammerAI/comments/1i2a9tp/60ish_day_review/
PS. I'm working on adding docs to the site to help address their privacy concerns (they don't like Discord)
Can you have multiple ai bots chatting to you ie:
you say something then both bots type in their separate responses
Not yet, sorry! But on the roadmap.
Didn't get no response but my RAM and disk went both 100% and after a while I lost 20GB of my C disk space and i cannot figure out where from since the hammer ai folder is only 3 GB. How is that cause I downloaded a model in D disk?
Sorry! We moved over to Discord to chat, tl;dr for others is to check
C:\Users\<yourname>\AppData\Roaming\HammerAIandC:\Users\<yourname>\AppData\Local\HammerAII can't install last version, any help?
Hi, sorry! I need more information to help you though 😂 I.e. what exactly can't install? Would also be great if you want to join the Discord so we can chat more quickly!
I fixed it, don't worry. I had to install a uninstaller because the previous ver was giving those error
Hello, when they will update again, I am waiting for an update that adds image generation and that is faster, since I have a problem, when I send the first message in the chat it freezes the screen and closes the App, I do not have a good processor but to use an AI it must be sufficient
Hi! We've started on image generation, though no guarantee on timing. Regarding freezing, can you use Ollama by itself (https://ollama.com/download)? If not, then the issue is just that you need a more powerful computer :( If so, would be awesome if you want to join the Discord where I can help debug and get it working for you!
Wow, I like this service you respond quickly, let me see how to solve the problem and I get into the discord
Hi, image generation is now live! Right now it's only with cloud models (and is paid), but local is coming (and will be free).
It's weird that I cannot download the default model or other even though I downloaded the software in my computer (；へ：) and I can't even download the model in the web version (ಥ_ಥ)
I cannot add image here (I don't know why), if you need screenshots, please send me an email
( • ̀ω•́ )✧ my email: nihilismnil@163.com
Hi, sorry, that is unexpected! Could you join the Discord and we can chat there?
I'm sorry I can't (；д；) I can't open the discord in my country (I think it's because of the firebreak in my country.)
But I have an account in github, can we use that? ヾ(•ω•`。)
Shoot! Well if you want to try the new beta, maybe it will work for you? It's a totally new way of using the LLM which should have WAY fewer errors! You can download it directly here: https://github.com/hammer-ai/hammerai/releases/tag/v0.0.97
Sorry for not replying to you for so long.ﾍ(;´Д｀ﾍ)
I spent a lot of times downloading the <HammerAI-0.0.97.Setup>, I can open it but it just shows "Installation has failed: There was an error while installing the application." When I want to open and check the setup log, it just disappeares (；д；).
I think I'm destined to be unusing your software (╥╯^╰╥), but I still appreciate what you have done for me and I really really hope you will succeed in the future!*\(๑• ₃ •๑)
Hope you have a wonderful day. (Sorry for my bad english, I'm from CN and I use the translator. (；へ：)
Hi, so sorry for this! But I have good news for you - I just rolled out a totally new LLM set up which is way more stable and should work on (almost) any computer. If you want to give it one more try I think it might work hopefully :) You can download it directly here: https://www.hammerai.com/desktop
probably wouldn't have bothered downloading it if i knew you had to subscribe for 9 bucks a month to be able to keep conversations. They just delete every time you leave the conversation.
Sorry about that :( If it helps you feel better, I use all the money I make to pay other devs to help work on the project with me. It was just getting too much for me to manage all alone with everything going on in my life. A top requested feature which I plan to add is that everyone can always have one conversation saved for free.
question will there be image generation in the future and if so will it be paid or free (what id do is have 2 versions a lesser quality one that is free and a higher quality one that is paid) also will there be a feature to save the chat on your device
Hi! No immediate plans for image generation, but definitely watching the space. If I see a nice library to help with this that is cross-platform I will add it. Not sure on pricing yet, but will definitely let you know!
For saving chats, we have that on Desktop currently! And are working on bringing it to Web as well. And when we roll out Mobile we'll have it.
Hi, image generation is now live! Right now it's only with cloud models (and is paid), but local is coming (and will be free).
Sometimes when I'm chatting, the answers starts to generate slower and then it just stops generating. Is there anyway to fix that? My GPU is not the best, so maybe that's it?
Mmm that could definitely be part of it, but the other is that our logic for long chats will lead to many tokens that need to get re-evaluated. Have been looking into fixing this, but no work around yet, sorry about that. One thing that helps is to use characters with less text in them (i.e. their personality is shorter). Will let you know when we have a better fix for this, sorry again!
Hi sorry! That is a known bug. The immediate fix is to turn off GPU acceleration under Settings. If that doesn't work, could you join the Discord so we can chat more and try to fix it for you?
Hi, I released a new update which uses Ollama under the hood and is much more stable! Would love to hear if it works for you now. Sorry again.
nice, thank god theres no membership/credits ass thing. need to blow off steam if u catch my drift ;)
Will there be an android version?
Hi, yes! We are currently working on a mobile app. No timeline yet, but will let you know when it's ready to test.
I get this message, and the AI don't do anything.
"Server started without GPU acceleration using no AVX support."
Hi, that means we weren't able to detect your GPU correctly and so started using just CPU. We'll be rolling out updates in the future which should help here, sorry about that!
Hello, I bought the pro plan for a month and cancelled it right after I got the license (because with subscriptions I know that I'll forget to cancel at the end of the month).
Now money is gone from my bank and it says that my pro license is cancelled. Can you help me out?
Best regards,
Poro
Oh, so sorry, that looks like a bug where we stop accepting licenses if the subscription is cancelled. Please DM me and I'll refund you for your order while fixing this. If you want to buy it again so it works for the month I'll also refund that (i.e. it will be free for you until I fix this).
All good, I just want the license to work for the month :)
So far I like the program a lot.
Where do I DM you?
Discord is best! You can see the link to join at the top here: https://www.hammerai.com/
Hello, this is a good tool, I wonder if you are going to translate it to Spanish or it is very difficult, I am not very good in English so I can't make the most use of it, this message is passed by translator
Hi, that's a great idea! I will look into translating the app, I had not thought too much about it previously.
Hi, I really like this tool and I would like to have the "Chats saving" feature because it's a must have! However I can't really afford that and I think the price is a little expensive (108€ a year with taxes) and I would have preferred a one time payment choice (even if a little more expensive). So I really hesitate...
1. If one day I decide to buy the pro, how can I be sure that the servers where my characters and their profile images are stored will not shut down one day (for one reason or another). (Because If I understand correctly, even locally stored character share their profile picture and description to the servers, no?)
2. This error can happen sometimes (from my logs):
[2024-04-14 19:15:36.242] [error] [llama-cpp-server@generateResponse] Error TypeError: terminated
Then, the character no longer responds, I have to reload it. If I get the "Chats saving" feature, I wonder if the conversation will be able to resume after this kind of error.
That's a lot of question I know, thanks and sorry for my imperfect English!
Hi, glad you like it! Yeah I'm thinking a lot about a lifetime license because others have asked for it. What would you be able to pay for it? I'm just not quite sure how to price it.
For your questions:
1. When you save characters locally the description and images are all saved on your computer, so it will work forever! You can click here to see exactly what we save and where it is saved:
2. Ah, that is a bug with longer conversations. I don't yet have a fix, but I have also seen it and will try to fix. Sorry about that.
Anyways, thank you for the feedback and support. I will let you know when I fix this bug, and no worries if you do not want to pay for the pro version, I understand it is a lot of money.
Thanks for your clear answer! :)
I don't know about the price, but between 60 and 110 euro/dollar for a lifetime license could be a good compromise between your need for programmer funds and the limited means of certain users and also the limited size of chats even if saved (see bugs below). Fyi: I just purchased a short time licence today to test the PRO features and I like it! But I am a little bit uncomfortable knowing that there is automatic bank withdrawal each end of cycle. So, yes a lifetime license could be great (as long as there is no kind of Pro + version in the future) :p
TWO BUGS:
1. The dialogue generation can sometimes stop (especially on long sentences), it's not a really big deal because you just have to try again but it's annoying.
2. Sometimes during a totally normal conversation (long but not always), the character start rambling by mixing verbs and words from some old (or not) messages. IE: "Oh no, maybe it's then yes I said it I said, said it!, but now we are here to, i can it! I think it's alright too!" or unfinished smaller strange one like "(giggling*". After some verifications, it look like it's a problem with exceded tokens limitation. It could be great to have the possibility to set it to 16384 or even 32768, because conversation are short (even with 8192 size). :/
Thank you! Okay that is very useful.
Ah, I see, I have been asked about a "continue generating feature", so that is on the roadmap. And yes, sorry about the bug with the rambling, that's known and I need to fix it still.
Also, just to offer, I have a 100% refund policy if you're unhappy with the Pro features, as the goal is not to make money, it's just to fund continued development. So if you're not happy just email / DM me your details on Discord and I will refund you. Would rather not make the money instead of making people unhappy.
No no, it's ok, thanks for your concern. However I would be interrested to join the discord, (I gave my profile name when ordering but I don't have invitation). :)
Ah sorry! I'm not sure how to directly invite people to the Discord. Instead I just assign people to a role who have purchased. If you join using the link here I'll then add you to the role! https://www.hammerai.com/
I get having to make money to fund development, but I feel that saying that "We will never charge for access to features" is a bit of a lie now that pro is a thing.
That is true, I'm sorry about that 😭. I should have qualified that statement when I made it, though I did really believe it at the time. The issue is that there are a lot of stability improvements and features to make, and I have a lot going on in my life. So my goal with collecting money is to be able to pay a developer to come work on the project. I'm sorry though, you're right that I went back on my word.
Hi, thank you for the kind words! What model do you want? I can add some other bigger ones.
Thank you 😁
Will there be an Option do save the Chats, so you dont have to start over everytime you use the a Model?
Yes that is the next feature I will add, definitely top of mind!
Perfect ^^.
I really enjoy this Program, even if i have slight Issues on some Parts, but thats to be expected, i have no clue which Model i should use and which my Pc can handle xD.
maybe you have a Idea how to fix that the Bot starts looping the same Message and gets kind of "stuck"
Thank you! Yeah I've been talking to people in Discord about the looping. Right now my suggestion is to just use a bigger model or try prompting differently. Sorry about that!
Hi itt66, saving chats is now supported!
Yeah, i saw it. But sadly i wont be able to use it ^^°
I really dont like subscribtions, i would rather do a one time Payment.
But it is how it is, i still like the Programm.
Hi, i've downloaded the app, given it permission to access the network, but it is stuck with "loading model"... Am i missing something please, do i need to configure or download something else?
Hi! Could you come and ask in Discord? Then we can debug more easily. Thanks!
It run quite well on Windows. Nice work :) you have develop the sw alone ? Any github repo avaiable ?
Thank you! Yes I develop it alone, though I am looking for a co-founder, preferably one with React Native experience who wants to bring this to mobile (and add some paid features, so we could make money there). No GitHub today, I have thought about open source but haven't decided to do it yet.
Does anyone know how to regenerate/refresh a single message? I hate having to reload the entire chat, just because one message was bad.
Sorry, no way to do that currently. But that's a nice feature to add and is possible, I will look into that when I next have time!
Thanks for adding a local option. This app is great.
Glad you like it!
Does it generate images or it's only text?
Only text! But I have heard people ask for images, maybe some day.
Hi, image generation is now live! Right now it's only with cloud models (and is paid), but local is coming (and will be free).
Why is the AI in this feel like it's robotic, and not really flowing, and learning like all limited memory AI? Literally I can say hi to someone and it literally says a set dialogue every time.
Hey! It can really depend on the model you're using, which are you using and have you tried a few? And have you tried modifying the temperature (try increasing it)? That may help make it more creative feeling:
If those don't help it may be the character. If the character prompt length (i.e. personality + scenario + example dialogs + first message) is too long there may not be enough context length left. So trying on some different characters may help things.
PS. If you want to join the Discord we can chat back and forth more easily, would love to help you get this working better.
This looks amazing! Thank you guys! I am curious, though, are you guys intending to integrate image generation into the chats? Such as scenes in a story while roleplaying?
Yeah that would definitely be awesome. I have been keeping my eye on it in case there is an approach I should take. Want to join the Discord and we can chat more about this?