Show HN: I built an LLM chat app because we shouldn't need 10 AI subscriptions

by maniknt28 • 1 year ago

58 65 1 year ago

I'm lost between ChatGPT vs Claude vs Gemini... which subscriptions to take? With Cursor and all these specific AI tools, I just wanted one simple chat app where I can use any model and pay only when I use it.

Couldn't find one, so I built one.

Pay only for what you use. Your prompts and docs, knowledge bases work with every model - no more copy-pasting between apps.

Started as a personal project, but thought someone else might benefit from this too.

https://prismharmony.com/chat

What do you think?

yellow_lead• 1 year ago

There's also https://t3.chat/

byyll3• 1 year ago

Is it pay per use?

Simulacra• 1 year ago

One improvement that would interesting, if not in there already, is to compare the answers from the different LLMs, and then combine them into the highest probability statement.

woodylondon• 1 year ago

Good idea, and I also explored this idea and a year ago and also started building one, recognising a gap in the market for a solution that supports multiple LLMs, but also provides small businesses with a centralised managed AI client - billing, monitoring, logging, company prompts, etc.

Ultimately, I discovered https://www.typingmind.com, which offers all of these features. I am sure there are others - I was amazed that not more of these came out. Might be worth to see what they have built. The more of these that come out the better - its a whole new market.

iLoveOncall• 1 year ago

You should integrate with AWS bedrock, you'd be able to offer 60 models instantly without having to spend any time integrating with separate APIs: https://docs.aws.amazon.com/bedrock/latest/userguide/models-...

shinycode• 1 year ago

I think this does the same as well https://mammouth.ai/

croes• 1 year ago

Or this

https://chatllm.abacus.ai/

warthog• 1 year ago

Just tried it but the messages keep disappearing after submission and streaming.

john_the_writer• 1 year ago

Reminds me of the xkcd https://xkcd.com/927/

Dilettante_• 1 year ago

This thread has Cunningham's Law¹ written all over it, too

¹https://en.wikipedia.org/wiki/Cunningham%27s_Law

supriyo-biswas• 1 year ago

I believe there's a couple of similar apps like https://msty.app and https://jan.ai that do the same and allow you to plug in your own API keys.

Alifatisk• 1 year ago

Isn't this what openrouter is for?

bob_theslob646• 1 year ago

What's openrouter?

v5v32• 1 year ago

https://openrouter.ai/

edited to remove statement saying API only, as per comments.

covercash• 1 year ago

Tangentially related, but is there any sort of chat bot that will look at the query and suggest which LLM might be the best for that particular task?

boomskats2• 1 year ago

Have a look at OpenRouter's Auto Router[0].

[0] https://openrouter.ai/openrouter/auto

lvl155• 1 year ago

I think in theory this type of approach is good but you know what’s going to happen eventually. Companies like OpenAI is already gatekeeping its best models. You need to pay and tier up on their platform to even use it. There’s no free lunch in AI landscape.

al_borland• 1 year ago

I like the idea of a single chat with many models. Pre-AI-everything, I was already a Kagi user, so already paying for that. I've started using the Kagi Assistant[0] to solve this for myself. I pay $10/month, as I always did, and that's my credit limit to the various LLMs. They added an AI cost tracker to create transparency to those costs. So far this month I've used $0.88 of my $10.00 allotment. So I don't feel like I'm in any danger of going over. If I wasn't already paying for this, I'd be pretty interested in an option that was pay-as-you-go.

Looking at your pricing I find the credit model a bit confusing. It feels like credit card points and I don't really have a concept of what that will get me. Tokens are a bit abstract, but that's the currency of AI, so it is what it is. Adding credits as an intermediary between tokens and dollars may have been done with the goal of simplify things, but it my head it makes it harder to understand and leaves more places for hidden fees to hide.

Giving some idea of how much usage someone could expect to get out of a 1,000 tokens or 100 credits or $1 would be useful. I can do the math and see I can do 20 web searches for $1, but does that include follow up questions? Is every question a web search? Kagi shows that I've used 15 searches so far today, and it's cost me less than 2¢ for the almost 19k tokens. So I'm a bit confused.

More generally on the chat-only umbrella tools, I do miss some of the nice-to-have options of going directly with the big players (interactive code editors, better images generation, etc), but not enough to be paying $20+/month/service.

[0] https://kagi.com/assistant

metrix2• 1 year ago

I've been using openrouter.ai to use "all llm's". No subscription, and can be tied to your editor of choice

pyman4• 1 year ago

For free? How's that possible when one AI prompt uses 10x more energy than a Google search [1]?

[1] Source: https://kanoppi.co/search-engines-vs-ai-energy-consumption-c...

kwamenum86• 1 year ago

Librechat seems perfect for your use case. It’s open source as well. Used by many of the big techcos to solve the problem you’re describing, so it’s battle tested https://www.librechat.ai

v5v32• 1 year ago

Only has some models but https://Duck.Ai is free

It's Duckduckgo's offering.

delduca• 1 year ago

There is an open source alternative

https://github.com/open-webui/open-webui

aae42• 1 year ago

anyone could run this themselves with open web UI and litellm, in fact that's the stack I've been using

apitman• 1 year ago

I'm not generally an open source purest but I believe their license is no longer OSI-compatible.

manx• 1 year ago

I also created my own. Frontend-only, no signup, based on openrouter, written in rust+leptos (wasm):

https://github.com/fdietze/tomatic

https://tomatic.app

fl0id• 1 year ago

this already existed with things like POE. like I mean for normal users, with a subscription for multiple models. same with frontends for api keys.

parsimo20102• 1 year ago

If you truly couldn’t find anything then you clearly don’t know how to use a web search. Poe.com brings many models under one subscription, and they have you completely beat on features and the number of models they offer. They are owned by Quora, so have a decently large group backing them. You.com lets you use multiple models, and also offers more features and models than you. Perplexity lets you use multiple models to chat. Merlin.ai lets you use multiple models. The list goes on, but there are a variety of established players in this space.

It looks like the only thing you offer over them is a “pay-as-you-go” option, where they are only subscription based. You kind of cheapen your differentiating factor by also offering a subscription. You need to show how you’re different from the competitors in this space, otherwise your growth will be very slow. You’re competing against OpenAI and Anthropic who are trying to sell their chat interfaces along with the other aggregator websites who have been around longer and have been developing features as integrating models from various providers this whole time. Do you think your pricing model will be enough, or do you have some killer features planned?

stoken• 1 year ago

If you truly couldn’t find anything then...

...you are clearly an engineer who has already decided to write their own so wasn't looking to hard, or did find something but went "I could do this better". Pretty much how many projects start.

parsimo20102• 1 year ago

OP said, in their post, they “Couldn't find one”

It’s totally fine to build your own implementation of something, especially if it’s for personal use or you’re not charging. OP is pitching a paid product. It’s not okay to ignore the sea of competitors and pitch your product on HN with a marketing blurb that isn’t true.

Anybody can slap together a chat UI and integrate a few LLM APIs. We need more than that if you’re charging money.

mritchie712• 1 year ago

Couldn't find one, so I built one.

there are no less than 100 of these.

exe34• 1 year ago

They couldn't find one of their own. All those 100 others were built by others!

PetrBrzyBrzek• 1 year ago

Couldn't find one, so I built one.

I’m sorry, but I find it hard to believe that you didn’t find any. I personally know at least 5 services that offer this.

swores• 1 year ago

Question to people talking about the various alternatives that already exist for this: does anyone know if there's something like OpenRouter that's open source and that, either as the only interface or preferably as an optional alternative to a web interface, lets you use a standard non-AI chat app (ideally Signal, more likely WhatsApp) as the interface?

Edit: I'd still be grateful for a reply with any recommendations or other options, but ChatGPT has given me a few things to look into when I'm at my PC - https://chatgpt.com/share/6873a9b5-ea8c-800c-b111-96b5f27a09...

deadbabe• 1 year ago

In the age of LLMs, why use something someone else built when an LLM could just build you your own!

17swagat• 1 year ago

Liked the concept.

But I've a question: Are you using the official APIs from each of these AI models directly into your app, or is there some other approach you're using to make them accessible?

lemming• 1 year ago

I'll bite: what is the web search story like? This is the killer feature of e.g. chatgpt that none of the alternative or OSS options offer. Having the search be fast (i.e. not round tripping to the client on every search) and integrated into the thinking process is unbeatable, I use it constantly. The big API providers all provide search options in their APIs now, but they're very quirky - openai doesn't allow search via their API with thinking models. Gemini doesn't allow you to use search with any other tools available. Claude's just doesn't seem to work that well, even in their own web UI. I even paid for typing mind, which is nice, but I never use it and always just end up paying for chatgpt again because of this.

gcanyon• 1 year ago

"If you're on the free plan, there's an 19.8% platform fee"

I think you mean the "pay as you go" plan? If not, that's pretty confusing, and 19.8% of "free" should still be "free" :-)

fredwu• 1 year ago

Couldn't find one, so I built one.
What do you think?

You were lost between all the AI stuff... but have you not tried to simply use Google to find a bunch of similar services?

intellectronica• 1 year ago

Cool and hats off to you for building an app ... but how come you couldn't find one?! There are so many. I use Raycast, which has access to all the models, there's Poe (from Quora) which is popular, Perplexity offers access to many models, and there are many more...

byyll3• 1 year ago

Is Poe pay per use or a subscription? I don't want to sign up just to find out.

stavros• 1 year ago

I use Librechat and OpenWebUI for that, but lately I've gotten a Claude subscription because A) deep research is amazing, B) their "canvas" UI is great, and C) Claude Code.

vinhnx• 1 year ago

I also built VT (https://vtchat.io.vn), a secure, privacy-first AI chat platform focused on data sovereignty. It supports BYOK (Bring Your Own Key) models, OpenRouter integration, and local models through LM Studio and Ollama (currently in beta).

bearjaws• 1 year ago

Should have used ChatGPT deep research to prevent you wasting time building the 41st version of this same product.

Here's all the main competitors:

1. You.com

2. Poe

3. Mammouth

4. Magai

5. TeamAI

6. TypingHive

7. WritingMate

8. ChatHub

9. Monica

KMnO4• 1 year ago

The pricing doesn’t render properly on mobile. The responsive design puts what is presumably a table into a vertical stack, which means I have no idea which price corresponds to which feature.

gtech1• 1 year ago

off-topic: anyone aware of a service where I could plugin my api keys for openai/gemini/claude and it asks the same question to all 3 and refines the answer by using one of them as the final arbiter ?

byyll3• 1 year ago

https://msty.app (cross-platform) and https://chorus.sh (Mac only) do that though they are both a desktop app rather than a service. Arguably better than putting your API key somewhere online in my opinion.

edmundsauto• 1 year ago

AnythingLLM is great for this, it even enables you to setup RAG from your own set of docs, which it then retrieves via embedding and adds as context.

I use that + OpenRouter which gives me API access to more models as well. Huge fan of this approach.

k9294• 1 year ago

I'm a tech co-founder at loqus.ai, me and my team thought that it’s a great idea to make a one app “to rule them all”, but it appeared to be a bad idea.

We decided to stop active development when realized that most people don't want different models, custom instructions, tools, etc..

People want their work done, and no one cares how exactly it will be done - the less they have to think about the setup, the better (and chatgpt just works in most of the cases).

replwoacause2• 1 year ago

So the loqus.ai service is defunct even though I can still sign up and be a paying customer?

abdibrokhim• 1 year ago

Couldn't find one, so I built one.

LMAO. You should learn how to use Google search

shekharupadhaya• 1 year ago

Check out https://chorus.sh/ they are really good. Multimodal support and super clean mac app.

replwoacause2• 1 year ago

Does using something like this mean you forfeit niceties like Projects, Memory, etc, that vendors like OpenAI and Anthropic give you out of the box? For me it’s more than just chat…having that context is hugely important across chats.

tensorblock• 1 year ago

We've built something sort of similar to this. Its functions are a bit different but will ultimately solve the same problem. It combines AI models into a single API.

Check it out if you would like (it's open source): https://github.com/TensorBlock/forge https://tensorblock.co/forge

filipeisho• 1 year ago

It's completely not the same but the first thing that crossed my mind was the comic of "How standards proliferate"

[0] https://imgs.xkcd.com/comics/standards_2x.png

Show HN: I built an LLM chat app because we shouldn't need 10 AI subscriptions

Related Stories