chore: prepare v0.5.4 release

Andyyyy64 · Andyyyy64 · commit f66b8e3b0b5a · 2026-05-17T02:33:58.000+09:00
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -4,6 +4,16 @@ All notable changes to this project will be documented in this file.
 
 The format is based on [Keep a Changelog](https://keepachangelog.com/), and this project adheres to [Semantic Versioning](https://semver.org/).
 
+## [0.5.4] - 2026-05-17
+
+### Fixed
+
+- Strix Halo / Ryzen AI MAX systems are now modeled as AMD shared-memory APUs
+  instead of tiny-VRAM discrete GPUs. `STRXLGEN`, `Radeon 8050S`,
+  `Radeon 8060S`, and related names get a 256 GB/s bandwidth estimate and use
+  the system shared-memory pool for fit checks, avoiding false CPU-only,
+  99%-offload, and `0 tok/s` recommendations.
+
 ## [0.5.3] - 2026-05-17
 
 ### Added
diff --git a/pyproject.toml b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "whichllm"
-version = "0.5.3"
+version = "0.5.4"
 description = "Find the best LLM that runs on your hardware"
 authors = [{name = "Andyyyy64"}]
 readme = "README.md"
diff --git a/uv.lock b/uv.lock
