gbnf

Atomic Agent is a local-first AI agent. Runs open-weight models on your own machine via llama.cpp.

Plug n Play GBNF Compiler for llama.cpp

llama.cpp gbml Menhir parser in OCaml

A composable Python DSL for building GBNF grammars compatible with llama.cpp.

Production-grade GBNF grammar-constrained decoding for LLMs. Zero dependencies, no unsafe code, pure Rust.

GBNF converter for llama.cpp Grammar directly from C# types

The Ultimate Local LLM Discord Bot!!!

Local-first agentic inference engine in C/C++. Multi-tier model routing, grammar-constrained output, MCP tool servers. Embeddable via C ABI.

GBNF (GGML BNF) grammar builder that exposes the GBNF notation system using a syntax that reflects the resulting grammar.

Convert a regular expression into a GBNF grammar for constrained decoding with llama.cpp and local LLMs. Zero dependencies.

End-to-end structured output for browser LLMs. Constrain Transformers.js/web-llm to valid JSON/SQL via GBNF grammars.

a collection of libraries for interacting with llamacpp

ServiceNow → GBNF grammar + iii worker generator | Open-source deterministic AI connector

A verification-first agentic engine that makes tiny local LLMs (Gemma E2B, Q4, CPU-only) reliably useful on non-prosumer hardware. Grammar-constrained decoding, deterministic orchestration, KV-cache-aware prompting, consent-based autonomy. The model stays small; the system becomes large. No cloud LLM APIs, ever.

MIA: Local LLM based Strength Training based on real trainers

Trilingual (ES/PT/EN) NLP optimisation study: one recommender, five interchangeable engines — a TF-IDF Naive Bayes trained on its own vocabulary against four local LLMs — scored on the same 140 golden cases. LangGraph, Chroma RAG, llama.cpp, grammar-constrained decoding, and a verifier that blocks any number the code did not compute.

Constrained decoding and enforced tool-call ordering for oh-my-pi: GBNF/regex/JSON-Schema output constraints plus user- or model-authored tool flows.

CPU-first local LLM inference for Rust: the llama.cpp decode loop as a library call, not a server.

Run llama.cpp inside a Unity game on iOS: static build with Metal, a C ABI IL2CPP can call, GBNF-constrained output, and the Xcode post-build wiring that makes it one click.