18 projects
kaos-nlp-transformers
Dense embeddings and small-model inference for the Kelvin Agentic OS — Rust-native ONNX backend, model2vec static lookup, optional GPU
kaos-llm-core
LLM programming primitives — typed, composable, optimizable LLM functions for KAOS
kaos-agents
Agentic runtime with persistent memory for KAOS
kaos-source
KAOS-native source discovery and materialization — filesystem, archive, HTTP, browser, plus REST connectors for Federal Register / eCFR / EDGAR / GovInfo / GLEIF and forensic parsers for VCard / EML / MBOX / PACER / EXIF
kaos-llm-client
Thin, provider-native LLM client for direct model calls within the KAOS ecosystem
kaos-web
Web extraction, browser automation, search, and domain intelligence for KAOS
kaos-ml-core
Classical ML primitives for the Kelvin Agentic OS — corpus, clustering, LLM labeling, logistic regression on the kaos-content AST
kaos-ui
Project scaffolding for KAOS user-facing applications — TUI, desktop, web, dashboard
kaos-core
Core protocol, runtime, and execution infrastructure for KAOS modules
kaos-office
Office document extraction for KAOS — DOCX, XLSX, PPTX to structured AST with provenance
kaos-content
Abstract document AST for KAOS — Block/Inline content model, provenance, annotations, serialization
kaos-nlp-core
High-performance NLP primitives for the Kelvin Agentic OS — SIMD string ops, multi-pattern matching, FST dictionaries, sentence segmentation, BM25 retrieval, fuzzy hashing
kaos-graph
High-performance graph library for the Kelvin Agentic OS — petgraph-backed engine, 40+ algorithms, RDF/SPARQL, knowledge-graph tools
kaos-tabular
DuckDB-powered tabular data engine for KAOS — register CSV / Parquet / JSON / SQLite, run SQL, export typed results
kaos-mcp
FastMCP runtime and server bridge for KAOS — exposes any KaosRuntime over stdio or streamable HTTP
kaos-pdf
PDF extraction and document processing for KAOS — structured AST output with provenance
kaos-names
Legal-flavored friendly name generator for KAOS sessions and agents
kaos-citations
Structured, high-performance citation identification and extraction. Text in, typed Pydantic Citation records out with provenance + Bluebook modifiers. 60 kinds across legal, financial, accounting, and identifier domains. Pure-Python, fully offline, deterministic, Rust-backed matching.
