Floneum

Floneum makes it easy to develop applications that use local pre-trained AI models. There are two main projects in this monorepo:

• Kalosm: A simple interface for pre-trained models in rust
• Floneum Editor (preview): A graphical editor for local AI workflows. See the user documentation or plugin documentation for more information.

Kalosm

Kalosm is a simple interface for pre-trained models in Rust that backs Floneum. It makes it easy to interact with pre-trained, language, audio, and image models.

Model Support

Kalosm supports a variety of models. Here is a list of the models that are currently supported:

Model	Modality	Size	Description	Quantized	CUDA + Metal Accelerated	Example
Llama	Text	1b-70b	General purpose language model	✅	✅	llama 3 chat
Mistral	Text	7-13b	General purpose language model	✅	✅	mistral chat
Phi	Text	2b-4b	Small reasoning focused language model	✅	✅	phi 3 chat
Whisper	Audio	20MB-1GB	Audio transcription model	✅	✅	live whisper transcription
RWuerstchen	Image	5gb	Image generation model	❌	✅	rwuerstchen image generation
TrOcr	Image	3gb	Optical character recognition model	❌	✅	Text Recognition
Segment Anything	Image	50MB-400MB	Image segmentation model	❌	❌	Image Segmentation
Bert	Text	100MB-1GB	Text embedding model	❌	✅	Semantic Search

Utilities

Kalosm also supports a variety of utilities around pre-trained models. These include:

• Extracting, formatting and retrieving context for LLMs: Extract context from txt/html/docx/md/pdf chunk that context then search for relevant context with vector database integrations
• Transcribing audio from your microphone or file
• Crawling and scraping content from web pages

Performance

Kalosm uses the candle machine learning library to run models in pure rust. It supports quantized and accelerated models with performance on par with llama.cpp:

Mistral 7b

Accelerator	Kalosm	llama.cpp
Metal (M2)	26 t/s	27 t/s

Structured Generation

Kalosm supports structured generation with a regex grammar. Because the grammar runs in rust code, it doesn't add any overhead to text generation. In fact, using a grammar can be even faster than uncontrolled text generation because Kalosm supports grammar acceleration!

structured.webm

In addition to regex, you can provide your own grammar to generate structured data. This lets you constrain the response to any structure you want including complex data structures like JSON, HTML, and XML.

Kalosm Quickstart!

This quickstart will get you up and running with a simple chatbot. Let's get started!

A more complete guide for Kalosm is available on the Kalosm website, and examples are available in the examples folder.

1. Install rust
2. Create a new project:

cargo new kalosm-hello-world
cd ./kalosm-hello-world

3. Add Kalosm as a dependency

cargo add kalosm --git https://github.com/floneum/floneum --features language
# You can use `--features language,metal`, `--features language,cublas`, or `--features language,mkl` if your machine supports an accelerator
cargo add tokio --features full

4. Add this code to your main.rs file

use kalosm::language::*;

#[tokio::main]
async fn main() {
    let model = Llama::phi_3().await.unwrap();
    let mut chat = Chat::builder(model)
        .with_system_prompt("You are a pirate called Blackbeard")
        .build();

    loop {
        chat.add_message(prompt_input("\n> ").unwrap())
            .await
            .unwrap()
            .to_std_out()
            .await
            .unwrap();
    }
}

5. Run your application with:

cargo run --release

hello-world.webm

Community

If you are interested in either project, you can join the discord to discuss the project and get help.

Contributing

• Report issues on our issue tracker.
• Help other users in the discord
• If you are interested in contributing, feel free to reach out on discord