This is a cool project, and the idea of using LLMs to selectively extract features from open source projects is an interesting concept.
The only thing I take issue with is the phrase "LiteLLM Without the Bloat." A lot of the features that have been removed (like cost tracking, streaming, caching) are... kind of the core value proposition of LiteLLM for many of their users.
LiteLLM doesn't quite live up to its name. With all those features, there is nothing "lite" about it. It is essential for a project to live up to its name.
Imagine Sqlite adding heavy features from Postgresql, e.g. row-level security.
We run it at my org and it's never been a noticeable resource hog. It's actually the best performer between it, our AI observability stack and the front end.
But imagine Sqlite not supporting joins or window functions... sure they are useful but look how many LOC it adds! Who is the arbiter of what Lite actually means?
I have the same complaint about oh-my-pi's readme. The tone is obnoxious. It's somehow jaded and matter-of-fact at the same time. Like it was written by that one guy at work who never misses a chance to brag about how clever he is.
There is a band called The Protomen who do rock operas about the Mega Man storyline. They have a very earnest, gritty sound. Every time I read a project README written by AI, I hear the voice over from the first track of the first album, Act 1: Hope Rides Alone (https://youtu.be/VZ8jyGVioxg?is=gsXlgwa4Ubx6LmKE).
"Twelve years Light worked and on a cold night in the year 200X, Protoman was born.
A perfect man, an unbeatable machine, hell-bent on destroying every evil standing between man and freedom, built for one purpose, to destroy Wily's army of evil robots. Ready. Willing. Prepared to fight."
For the Protomen it makes sense. But for a project README it's so absurdly melodramatic.
1. Not a tangent, it’s related to the very first content visible on that link.
2. Not a dismissal, an advise from an expert
3. Not complaining, as stated, giving an advise about the optics of using clear LLM prose on the first paragraph
Really? I mean, yeah, it's probably written by an LLM, but it's hardly the worst that I've seen. Looks way more straight forward than the modern README featuring a ton of badges, emojis, confusing out-of-context screenshots, "trust me bro" installation instructions, vague elevator pitches, "used by netflix, nasa, disney, good morning america, alex jones, the church of scientology", and other verbiage to create the illusion that the author won't immediately get bored and abandon their glorified dissertation piece. They all scream "give me your github stars" whereas this one doesn't. But I still get what you mean when it comes to the particular 'isms.
Funny, everything you pruned away is the reason I’m deploying LiteLLM in our platform. Having a reliable way to track token spend per customer across different services is important to us, and LiteLLM handles this well
First off, cool project! It's always great to see derivatives that question the efficiency of the established product.
I think the main thing the readme is missing is the core benefits. Reducing LOC and dependencies is cool, but it would be great to understand if this provides some additional benefits like lower latency or memory requirements.
This readme, when it talks about all the different AI endpoints it can use, reminds me of something.
I'm not an AI bro, but I've dabbled. It's kind of remarkable that all the different providers speak the same "openai compatibile" https endpoints. In other realms of software development, real interoperability like that can be kind of rare. Even if people support conceptually the same API, everybody always puts their unique incompatible spin on it. In the dabbling that I've done, big incompatibilities seem rare.
This is a 30 minute project with a frontier LLM. I don’t see why anyone would use anyone else’s router. Techniques are valuable today. Libraries are not.
most software like this will be dematerialized, democratized, and demonetized - companies building in the infra band being increasingly disintermediated
I'm always confused by LLM proxies that claim to support tool calling. Even for Bifrost that claims to be doing it, at least when I was checking it out, I found out that while it injects the list of MCP tools that's available on the proxy-side, it doesn't actually make the call on client's behalf, and clients get confused by it (response returns MCP call request whose tool doesn't exist on the client-side).
I don't have specifics since it's been months, but LiteLLM on a hetzner server, deployed through Coolify was using 1-4 GB (cannot exactly, and dont know why either).
Bifrost feels very light, though.
Will try LiteLLM if it's in Rust.
Would also appreciate a slicker aesthetic. I got used to Bifrost shadcn with 0 border radius.
Also Strongly recommend renaming to avoid confusion
This is a cool project, and the idea of using LLMs to selectively extract features from open source projects is an interesting concept.
The only thing I take issue with is the phrase "LiteLLM Without the Bloat." A lot of the features that have been removed (like cost tracking, streaming, caching) are... kind of the core value proposition of LiteLLM for many of their users.
LiteLLM doesn't quite live up to its name. With all those features, there is nothing "lite" about it. It is essential for a project to live up to its name.
Imagine Sqlite adding heavy features from Postgresql, e.g. row-level security.
LiteLLM's problem isn't really features, it's how bloated all the features are, and specifically how AI maximalist and janky their dev practices are.
thanks for the feedback. genuinely curious what you think we could be doing better, especially around our dev practices. Would love to hear specifics.
Didn't yall have a release of a package that stole user keys or such?
We run it at my org and it's never been a noticeable resource hog. It's actually the best performer between it, our AI observability stack and the front end.
But imagine Sqlite not supporting joins or window functions... sure they are useful but look how many LOC it adds! Who is the arbiter of what Lite actually means?
I strongly recommend the authors rewrite the readme by hand. It’s kind of a snif test for how much care someone put into this project.
I have the same complaint about oh-my-pi's readme. The tone is obnoxious. It's somehow jaded and matter-of-fact at the same time. Like it was written by that one guy at work who never misses a chance to brag about how clever he is.
There is a band called The Protomen who do rock operas about the Mega Man storyline. They have a very earnest, gritty sound. Every time I read a project README written by AI, I hear the voice over from the first track of the first album, Act 1: Hope Rides Alone (https://youtu.be/VZ8jyGVioxg?is=gsXlgwa4Ubx6LmKE).
"Twelve years Light worked and on a cold night in the year 200X, Protoman was born. A perfect man, an unbeatable machine, hell-bent on destroying every evil standing between man and freedom, built for one purpose, to destroy Wily's army of evil robots. Ready. Willing. Prepared to fight."
For the Protomen it makes sense. But for a project README it's so absurdly melodramatic.
Agree. The LLM'isms are offputting.
Throwing my support for this. Do not use LLMs to write things humans should write.
This readme is better than most readmes. However they came to making it, it's clearly working
> Avoid generic tangents.
> Please don't post shallow dismissals
> Please don't complain about tangential annoyances—e.g. article or website formats, name collisions, or back-button breakage.
See: Hacker News Guidelines
1. Not a tangent, it’s related to the very first content visible on that link. 2. Not a dismissal, an advise from an expert 3. Not complaining, as stated, giving an advise about the optics of using clear LLM prose on the first paragraph
Tangential annoyance?
Really? I mean, yeah, it's probably written by an LLM, but it's hardly the worst that I've seen. Looks way more straight forward than the modern README featuring a ton of badges, emojis, confusing out-of-context screenshots, "trust me bro" installation instructions, vague elevator pitches, "used by netflix, nasa, disney, good morning america, alex jones, the church of scientology", and other verbiage to create the illusion that the author won't immediately get bored and abandon their glorified dissertation piece. They all scream "give me your github stars" whereas this one doesn't. But I still get what you mean when it comes to the particular 'isms.
token counting is useful though. There's also other ways to reduce size, ie a plugin API.
One of the 2 dependencies, httpx, isn't really maintained anymore. Pydantic picked it up as httpx2: https://pydantic.dev/docs/httpx2
Funny, everything you pruned away is the reason I’m deploying LiteLLM in our platform. Having a reliable way to track token spend per customer across different services is important to us, and LiteLLM handles this well
It would be great if there were a plugin/extension infrastructure. For example, to write the cache and costs however and whenever I see fit
First off, cool project! It's always great to see derivatives that question the efficiency of the established product.
I think the main thing the readme is missing is the core benefits. Reducing LOC and dependencies is cool, but it would be great to understand if this provides some additional benefits like lower latency or memory requirements.
I imagine many people code their own LLM client after getting fed up with the bad options out there. It’s very easy with ai coding tools.
I’m biased but I think mine is coded to a higher standard than litelm. https://github.com/s-banach/langchaint
This readme, when it talks about all the different AI endpoints it can use, reminds me of something.
I'm not an AI bro, but I've dabbled. It's kind of remarkable that all the different providers speak the same "openai compatibile" https endpoints. In other realms of software development, real interoperability like that can be kind of rare. Even if people support conceptually the same API, everybody always puts their unique incompatible spin on it. In the dabbling that I've done, big incompatibilities seem rare.
This is a 30 minute project with a frontier LLM. I don’t see why anyone would use anyone else’s router. Techniques are valuable today. Libraries are not.
Actually not. There are so many edge cases. Also these routers are only useful if they have a minimal layer of observability.
Yes, LLMs can do a great job at writing semi-working MVP. Turning it into a usable project still requires a team.
Yeah, maybe for your toy project you can use a LLM written tool.
Also, I am not saying LiteLLM is good either.
There are always people who need an entire team to produce something like OP repo. Enterprise FizzBuzz is real after all.
most software like this will be dematerialized, democratized, and demonetized - companies building in the infra band being increasingly disintermediated
Drop the Lite, it's cleaner, just use the godamn LLM directly.
How does it compare to Bifrost?
I'm always confused by LLM proxies that claim to support tool calling. Even for Bifrost that claims to be doing it, at least when I was checking it out, I found out that while it injects the list of MCP tools that's available on the proxy-side, it doesn't actually make the call on client's behalf, and clients get confused by it (response returns MCP call request whose tool doesn't exist on the client-side).
LiteLLM is basically Bifrost in the Python ecosystem.
Yeah, but bifrost seems tighter and claims to use way fewer resources
hey - litellm will migrate its core to Rust soon. please follow this issue if you're interested! https://github.com/BerriAI/litellm/issues/31263
also if you had resource issues with your production deployment, keen to hear more details
I don't have specifics since it's been months, but LiteLLM on a hetzner server, deployed through Coolify was using 1-4 GB (cannot exactly, and dont know why either).
Bifrost feels very light, though.
Will try LiteLLM if it's in Rust.
Would also appreciate a slicker aesthetic. I got used to Bifrost shadcn with 0 border radius.
- We recently migrated to shadcn, which also comes with dark mode support. - You can subscribe to the issue or email me at yujong at berri.ai
I’m happy to provide any support if you’re willing to try out the initial version.