How matrixsmaster/ligguf works
LiGGUF is a compact local GGUF inference engine with two single-file reference editions and a modular FAST runtime. The main user workflow is to load a model, tokenize a prompt, run transformer inference, sample and print generated tokens; FAST also supports interactive chat and optional retrieval-augmented memory. The graph includes documented model-family support, CPU acceleration, visualization, and the experimental distributed example. Some FAST internals were only partially excerpted, so uncertain links are omitted or marked unverified.
Main entry points: ligguf_fast_qk.cpp FAST Runtime · ligguf_distrib.cpp SARA Example · vdb.cpp BERT Embeddings · lil_math.cpp Tensor Math
Components 17
Reference Engines
- C Reference
c/ligguf.c - C++ Reference
cpp/ligguf.cpp - SARA Example
cpp/ligguf_distrib.cpp
FAST Runtime
- FAST Runtime
fast/ligguf_fast_qk.cpp - GGUF Loader
fast/lil_gguf.cpp - Tokenizer
fast/tokenize.cpp - Chat and State
fast/ligguf_fast_qk.cpp
Retrieval Memory
- BERT Embeddings
fast/vdb.cpp - Vector Retrieval
fast/vdb.cpp
Inference Compute
- Tensor Math
fast/lil_math.cpp - AVX2 Kernels
fast/arch_avx2.h - NEON Kernels
fast/arch_neon.h
Runtime Extensions
- Hotpath Visualizer
fast/viz_sdl.cpp
Other components
- CLI User
- GGUF Model
- SDL2
- Peer Workers
Connections 20
- CLI User → FAST Runtime: runs
README.md - CLI User → C Reference: runs
README.md - CLI User → C++ Reference: runs
README.md - FAST Runtime → GGUF Loader: loads model
fast/ligguf_fast_qk.cpp - GGUF Model → GGUF Loader: provides weights
README.md - FAST Runtime → Tokenizer: tokenizes prompt
fast/ligguf_fast_qk.cpp - FAST Runtime → Tensor Math: runs inference
fast/ligguf_fast_qk.cpp - Tensor Math → AVX2 Kernels: uses kernels
fast/lil_math.cpp - Tensor Math → NEON Kernels: uses kernels
fast/lil_math.cpp - FAST Runtime → Chat and State: handles chat
README.md - Chat and State → FAST Runtime: feeds turns
fast/ligguf_fast_qk.cpp - FAST Runtime → BERT Embeddings: requests retrieval
README.md - BERT Embeddings → GGUF Loader: loads BERT model
fast/vdb.cpp - BERT Embeddings → Vector Retrieval: embeds and queries
fast/vdb.cpp - Vector Retrieval → FAST Runtime: supplies prompt memory
README.md - FAST Runtime → Hotpath Visualizer: traces activations
fast/ligguf_fast_qk.cpp - Hotpath Visualizer → SDL2: renders with
README.md - CLI User → SARA Example: runs workers
README.md - SARA Example → Peer Workers: exchanges activations
README.md - SARA Example → GGUF Model: loads shared model
README.md