3sparks Chat

The developer seems to be constantly improving the app and for me it meets many of my used case needs with a back in connection to LM studio. I very much appreciate the image enablement.

After a week still can’t get it to work, while others chat apps have no problem with my server, waste of time and money.

I just got charged by this. I cannot cancel a subscription and there is no way to do this in the app either.

Great job overall. However, a few things are really annoying. The default token limit is too small. I’d like to set the default per connection. Another annoying one is that Gemini doesn’t use grounding. Please enable Search grounding for Gemini models. The last annoying thing is that there’s no iOS Live Activity corresponding to running queries. It’s really annoying to have to keep the app open for the query to not be interrupted.

So far works very well and was able to provide a picture to a local model that understands pictures. Also was able to provide a PDF via web URL link. But choosing Text file cannot provide local PDFs. Adding an option to provide local PDF files and drag and drop of files into the chat would be a great option!

I don't use this app, I wasted my money on it. Buy LATENT CHAT if you want backgrounding for your LLM. This developer acts like her cares by adding useless functionality like web to RAG, but u can just download a pdf and rag that in latent chat so it does the same thing, BUT as soon as you leave this app your prompt and the response WILL ERASE WHILE STREAMING. This is not ideal for large contexts and slow LLM responses, and this guy thought he could jam in vision enabled (analyze image input) without any repercussions. So that image adding function is basically useless if you expect a long response for an LLM that utilizes your full GPU offload, and if you accidentally leave the app you can kiss your maximum effort prompt you worked on prompt bye bye! 👋 TL;DR: I see right through this gimmicky update, its not worth it until he adds backgrounding while streaming responses. Otherwise it's ok if you just need very basic text response where you are stuck inside the app and cannot leave. 😘yw for the feedback😘

You can’t set default instructions, skip the setup and use the server-side system prompt, or choose a default model. There’s no way to lock in default settings like temperature, context window, or streaming mode. Way too limited for power users.

I had problems at first, developer responded and fixed it asap. Impressive! The app is exactly what I wanted. Highly recommend. Thanks!! Update April 17, 2025: this app is wonderful. I use it to connect to LLM’s on my Mac mini. Dev is putting out updates weekly, adding a lot of features. Wow. Very impressed!

I use this with a Llama 70b model with Ollama server, which can respond slowly but that's fine for my purposes... after a couple of initial successful LLM text responses I start receiving "LLM is running slowly..." errors. It never recovers. The problem is persistent. The app will get better and better, but for now the timeout issue makes it unusable.

I use this app daily for the integration with LM Studio in which I can host my own LLM's in LM Studio and then access them via this app on my cell or iPad for chat capabilities, which works perfectly. Also I have given several suggestions to the Dev and he always is very responsive and dedicated to making this app the most useful it can be by adding additional features.