v0.3.3

What's Changed

• feat: add a Qwen3.5 9B NVFP4 MTP recipe for the DGX Spark by @james-shore in #30
• feat: add a Qwen3.6 35B A3B MTP recipe for the M5 Pro by @james-shore in #31
• feat: make Qwen3.8 27B the default and list featured models first by @james-shore in #32
• chore: bump version to 0.3.3 by @james-shore in #33

Full Changelog: v0.3.2...v0.3.3

v0.3.2

What's Changed

• feat: clearer TUI model list and run screen by @james-shore in #20
• chore: bump version to 0.3.2 by @kartikey-vyas in #29

Full Changelog: v0.3.1...v0.3.2

v0.3.1

What's Changed

• test: keep the fake server's output events in separate reads on Windows by @kartikey-vyas in #25
• feat: add a Qwen3.5 9B Q4_K_M MTP recipe for the M5 Pro by @kartikey-vyas in #26
• fix: ask vLLM and SGLang to report cached prompt tokens by @kartikey-vyas in #27
• chore: bump version to 0.3.1 by @kartikey-vyas in #28

Full Changelog: v0.3.0...v0.3.1

v0.3.0

What's Changed

• docs: explain server requirement, reduced-context runs, and the Rust client by @james-shore in #18
• chore: use US spelling throughout by @james-shore in #17
• feat!: send the typed submission contract by @kartikey-vyas in #24

New Contributors

• @james-shore made their first contribution in #18

Full Changelog: v0.2.0...v0.3.0

v0.2.0

What's Changed

• feat: move managed-run recipes to recipes/// by @kartikey-vyas in #4
• feat: support Windows by @kartikey-vyas in #2
• feat: migrate dataclasses to Pydantic models by @kartikey-vyas in #3
• feat: keep the API key a SecretStr until the request header by @kartikey-vyas in #5
• refactor: move one-field range and emptiness checks into Pydantic types by @kartikey-vyas in #6
• fix: let attached TUI runs send tool_choice none by @kartikey-vyas in #8
• refactor: read JSON records through strict Pydantic validation by @kartikey-vyas in #7
• chore: remove the Qwen3.8 Flash-Next Strix Halo recipe from main by @kartikey-vyas in #9
• feat: add Ling 3.0 flash llama.cpp recipes with DSpark by @kartikey-vyas in #10
• feat: make the parallel_tools probe advisory in runtime qualification by @kartikey-vyas in #13
• chore: bump version to 0.2.0 by @kartikey-vyas in #14
• fix: run the release version check in the locked project environment by @kartikey-vyas in #15

Full Changelog: v0.1.0...v0.2.0

v0.1.0

What's Changed

• fix: publish rustcore from its own PyPI environment by @kartikey-vyas in #1

New Contributors

• @kartikey-vyas made their first contribution in #1

Full Changelog: https://github.com/ArtificialAnalysis/aa-agentperf-local/commits/v0.1.0