ZINC Zig API

Startup, argument parsing, and the top-level process path that wires model loading, tokenization, and generation together.

GGUF parsing, metadata normalization, and the runtime structures that move weights from disk into GPU-resident buffers.

Prompt and output text conversion between UTF-8 strings and token IDs used by the decode loop.

Static graph construction and dependency ordering for the per-token compute work that the runtime records and submits.

Decode state, pipeline ownership, command recording, and token sampling inside the active inference loop.

Logit post-processing, argmax helpers, and token-selection controls layered on top of the decode runtime.

Typed wrappers around the compute shaders that prepare push constants, descriptor layouts, and per-op dispatch dimensions.

Vendor and architecture heuristics that translate raw Vulkan properties into tuning defaults for AMD, NVIDIA, and Intel GPUs.

Low-level Vulkan setup, memory allocation, buffers, pipelines, and command submission primitives used throughout the engine.

Low-level Metal device discovery, buffers, pipelines, and command submission primitives used by the Apple Silicon backend.

Catalog metadata, cache management, model downloads, and active-selection helpers used by the CLI and server.

Continuous batching scheduler, paged KV cache management, and request lifecycle tracking for concurrent inference serving.

OpenAI-compatible HTTP server, route dispatch, SSE streaming, and session management for serving inference over the network.

Tool-use protocol helpers: chat-template-aware tool definitions, argument parsing, and response formatting for function-calling-capable models.

CUDA device discovery, context management, device buffers, NVRTC-compiled pipelines, and stream-based command submission for the NVIDIA backend.

Continuous-batching serving worker, request registry, per-token streaming channels, and GPU-owned decode loop for concurrent CUDA inference.