# GGUF Runtime — full public inventory

Generated UTC: 2026-07-24T15:07:33Z

## Source-of-truth boundaries
- LMRuntime.com: canonical package, API, architecture, capability, evidence, and release documentation.
- NuGet: current package release metadata.
- GgufRuntime.com: C# host application, reference UI architecture, intake, lifecycle, evidence visualization, and application-level examples.
- UAIX.org: portable AI Memory, Project Handoff, exchange, and conformance boundaries.
- Site discovery layer: human-first pages, route inventory, answer cards, publication-readiness result, and public-safe discovery files.

## Public documentation
- Package Selection: https://ggufruntime.com/docs/package-selection/ — Choose the narrowest UAIX.LmRuntime package for a host task without inferring capability from package names.
- Verified GGUF Intake: https://ggufruntime.com/docs/verified-gguf-intake/ — Admit a local artifact through explicit source, license, path, size, byte-count, hash, and structural checks.
- GGUF Inspection: https://ggufruntime.com/docs/gguf-inspection/ — Inspect bounded metadata and tensor descriptors without starting inference.
- Runtime and Host Boundaries: https://ggufruntime.com/docs/runtime-and-host-boundaries/ — Keep acquisition, authority, UI, transport, persistence, prompt assembly, and policy in the host.
- Model and Session Lifecycle: https://ggufruntime.com/docs/model-and-session-lifecycle/ — Represent model ownership, isolated sessions, continuation state, cancellation, and disposal order explicitly.
- Prompt and Chat Template Boundary: https://ggufruntime.com/docs/prompt-and-chat-template-boundary/ — Keep model-specific conversation formatting and special-token choices in the host.
- Bounded Generation: https://ggufruntime.com/docs/bounded-generation/ — Apply explicit prompt, output, stop-token, memory, and concurrency ceilings.
- Cancellation and Disposal: https://ggufruntime.com/docs/cancellation-and-disposal/ — Propagate host cancellation, distinguish it from failure, and release session/model ownership deterministically.
- Runtime Telemetry: https://ggufruntime.com/docs/runtime-telemetry/ — Collect path-specific measurements locally with documented timing and memory methods.
- Backend Registration, Probing, and Fallback: https://ggufruntime.com/docs/backend-registration-probing-and-fallback/ — Display backend package, registration, declarations, probes, selection, fallback, and execution as separate states.
- CPU and GPU Evidence: https://ggufruntime.com/docs/cpu-and-gpu-evidence/ — Use executed-path records rather than package or hardware labels when presenting acceleration.
- Cross-platform File Access: https://ggufruntime.com/docs/cross-platform-file-access/ — Keep trusted roots, file pickers, scoped storage, bookmarks, and path privacy host-owned.
- Avalonia Host Architecture: https://ggufruntime.com/docs/avalonia-host-architecture/ — Structure a proposed cross-platform desktop shell around explicit services, view models, dispatch, cancellation, and evidence.
- ASP.NET Core Local Host: https://ggufruntime.com/docs/aspnet-core-local-host/ — Expose a narrowly bounded loopback service without turning model acquisition or filesystem authority into a request API.
- Error Handling: https://ggufruntime.com/docs/error-handling/ — Preserve technical failure classes without leaking paths, prompts, credentials, or private environment data.
- Security Model: https://ggufruntime.com/docs/security-model/ — Apply bring-your-own-authority rules to files, memory, network, commands, telemetry, and persistence.
- Evidence Records: https://ggufruntime.com/docs/evidence-records/ — Retain stable, path-free records that distinguish identities, backend states, load, generation, cancellation, measurements, environment, and limitations.
- Troubleshooting: https://ggufruntime.com/docs/troubleshooting/ — Diagnose intake, load, session, backend, generation, cancellation, and resource failures without weakening controls.

## Public examples
- Console Quick Start: https://ggufruntime.com/examples/console-quick-start/ — Load one verified GGUF model, create an isolated session, run bounded generation, and dispose in order.
- Verified Model Intake: https://ggufruntime.com/examples/verified-model-intake/ — Compute and compare the current GGUF identity before admitting an artifact to a host catalog.
- GGUF Metadata Inspector: https://ggufruntime.com/examples/gguf-metadata-inspector/ — Parse bounded container metadata and tensor descriptors without starting inference.
- Isolated Session: https://ggufruntime.com/examples/isolated-session/ — Continue one deliberately prepared model session while preserving explicit reset behavior.
- Cancellation: https://ggufruntime.com/examples/cancellation/ — Pass host cancellation into bounded generation and observe committed tokens synchronously.
- Runtime Evidence Record: https://ggufruntime.com/examples/runtime-evidence-record/ — Create a typed, path-free JSON observation record without calling it a certificate.
- Avalonia Desktop Shell: https://ggufruntime.com/examples/avalonia-desktop-shell/ — Marshal CPU-bound generation and token display through an Avalonia-oriented MVVM boundary.
- WPF Host: https://ggufruntime.com/examples/wpf-host/ — Keep generation off the WPF dispatcher and marshal bounded token evidence back to the UI.
- WinForms Host: https://ggufruntime.com/examples/winforms-host/ — Run generation on a worker and update WinForms controls through BeginInvoke.
- ASP.NET Core Loopback Service: https://ggufruntime.com/examples/aspnet-core-loopback-service/ — Expose a bounded loopback-only generation endpoint while keeping model acquisition and trust outside requests.