MoonshotAI
Featured
Kimi-K3
Kimi K3 is an open-weight, 2.8T-parameter native multimodal agentic model with a 1M-token context window, designed for frontier coding, knowledge work, and reasoning tasks.
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It provides a unified runtime for next-generation computing formats like UMC and LDP, and enables cross-format debugging and search via source maps and accessibility trees.
Who it is for
- Developers building document or file viewers
- WebAssembly plugin developers
- Researchers and engineers working on new file formats or codecs
- Developers needing a flexible runtime for fetching and rendering diverse content
Use cases
- Viewing content from archives (e.g., zip) without manual extraction
- Building custom document viewers that support multiple formats via plugins
- Debugging and searching across different file types within a single runtime
- Testing and prototyping new compression or multimedia formats (UMC, LDP)
Strengths
- Hot-swappable Wasm plugin architecture for protocols, compression, archives, transforms, and views
- Out-of-the-box support for common network schemes, compression algorithms, and file formats
- Planned cross-format debug server and multimodal search using source maps and accessibility tree
- Serves as the official runtime for innovative formats like UMC (ultra-efficient codec) and LDP (PDF replacement)
Considerations
- Still early-stage with roadmap items not yet implemented (cross-format debug/search)
- Built-in format support limited to a few types (zip, markdown, image formats) – others require community plugins
- Relies on WebAssembly plugin ecosystem; format integration effort may vary
README quick start
Omnishell 🌐
The programmable, format-agnostic browser runtime powered by WebAssembly.
Stop rebuilding document viewers from scratch. Omnishell treats network protocols, compression algorithms, archive structures, and file formats as simple, hot-swappable Wasm plugins.
🚀 Out-of-the-Box Launch Capabilities
Omnishell includes a robust baseline layer, allowing developers to immediately build on top of standard web protocols:
- Network Schemes:
file://(builtin) |http://|https:// - Compression:
zstd|gzip|deflate|raw-deflate - Archives:
zip - Transforms:
markdown➔html - Views:
html|svg|png|jpg|wasm(builtin)
🗺️ Roadmap
Planned features include cross format debugging and search capabilities by using the accessibility tree exported by view plugins and source maps exported by transform plugins, Omnishell delivers a built-in debug server and multimodal search natively across any file type.
🧪 The Innovation Sandbox
Omnishell was built to host next-generation computing formats. It serves as the official runtime and testing environment for:
- UMC (Universal Multimedia Codec): An ultra-efficient image/audio/video
compression format reaching
0.1bpp @30+dB PSNRon video with symmetric500MPsencoding/decoding throughput. - LDP (Lean Document Package): A modern, document format engineered to replace PDF in the 21st century.
🛠️ Contribute Your Format
Want to bring Typst, Iroh, BitTorrent, Pdf, Brotli, or DjVu to
Omnishell? Or integrate a js engine like Boa? Check out components/content/svg
to see how to compile your first plugin to Wasm in minutes.
Technical overview
All plugins either take or return a Content type:
struct Content {
url: String,
info: MimeInfo,
length: u64,
content: ContentReader,
}
struct MimeInfo {
/// corresponds to the http `content-type` header
mime: String,
/// corresponds to the http `content-encoding` header
encoding: Option,
}
There are several types of plugin:
-
network plugin: handles a url scheme, takes an url and returns a content record. examples: http
-
archive plugin: handles a mime type, takes a content record and a path and returns a new content record. examples: zip
-
compression plugin: handles content-encoding, takes a content record with a content
Description
The programmable, format-agnostic browser runtime powered by WebAssembly.
Related repositories
Similar projects matched by category, topics, and programming language.
jamesob
Featured
local-llm
A comprehensive guide for building and configuring a high-end local machine to run state-of-the-art LLMs, with detailed hardware choices, BIOS tuning, and Docker-based model serving.
Blaizzy
Featured
nativ
Nativ is a native macOS app that lets you run AI models locally on Apple silicon, offering chat, model management, performance analytics, and an OpenAI/Anthropic-compatible API server.