Main changes

• Bring Your Own Provider (@leseb) - use out-of-tree provider code to execute the distribution server
• OpenAI compatible inference API in progress (@bbrowning)
• Provider verifications (@ehhuang)
• Many updates and fixes to playground
• Several llama4 related fixes

What's Changed

• docs: llama4 getting started nb by @ehhuang in #1878
• feat: make multi-turn tool call tests work with llama4 by @hardikjshah in #1886
• docs: Fix typo in README.md by @franciscojavierarceo in #1880
• docs: update importing_as_library.md by @mattf in #1863
• refactor: move all llama code to models/llama out of meta reference by @ashwinb in #1887
• test: verification on provider's OAI endpoints by @ehhuang in #1893
• feat: Updated playground rag to use session id for persistent conversation by @MichaelClifford in #1870
• fix: type by @ehhuang in #1898
• docs: colorize Discord badge & add icon in README by @wescpy in #1865
• refactor: move missing tests to test directory by @leseb in #1892
• feat: adds test suite to verify provider's OAI compat endpoints by @ehhuang in #1901
• docs: Update remote-vllm.md with AMD GPU vLLM server supported. by @AlexHe99 in #1858
• chore: remove unused tempdir in agent by @leseb in #1896
• chore: fix hash for thollander/actions-comment-pull-request by @booxter in #1900
• feat: ability to execute external providers by @leseb in #1672
• fix: use ollama list to find models by @mattf in #1854
• fix: update getting started guide to use ollama pull by @mattf in #1855
• fix: solve unregister_toolgroup error by @pdettori in #1608
• fix: Playground Container Issue by @Jaland in #1868
• feat: Add tools page to playground by @MichaelClifford in #1904
• docs: Add recent release notes by @terrytangyuan in #1899
• fix: add tavily_search option to playground api by @MichaelClifford in #1909
• chore: Detect browser setting for dark/light mode and set default to light mode by @franciscojavierarceo in #1913
• chore: simplify running the demo UI by @leseb in #1907
• fix: Mirror llama4 rope scaling fixes, small model simplify by @ashwinb in #1917
• fix: on-the-fly int4 quantize parameter by @jiawenliu64 in #1920
• docs: Avoid bash script syntax highlighting for dark mode by @terrytangyuan in #1918
• docs: Redirect instructions for additional hardware accelerators for remote vLLM provider by @terrytangyuan in #1923
• docs: fix model name by @leseb in #1926
• docs: Updating background color for code in darkmode by @franciscojavierarceo in #1930
• feat(verification): various improvements by @ehhuang in #1921
• fix: Fix linter failures from #1921 by @franciscojavierarceo in #1932
• fix: Playground RAG page errors by @ilya-kolchinsky in #1928
• fix: use torchao 0.8.0 for inference by @leseb in #1925
• docs: Update quickstart page to structure things a little more for the novices by @franciscojavierarceo in #1873
• test(verification): add streaming tool calling test by @ehhuang in #1933
• test(verification): overwrite test result instead of creating new ones by @ehhuang in #1934
• docs: fix errors in kubernetes deployment guide by @Bobbins228 in #1914
• fix: ensure resource registration arguments are typed by @mattf in #1941
• feat: Add a direct (non-agentic) RAG option to the Playground RAG page by @ilya-kolchinsky in #1940
• fix: remove extra sft args in NvidiaPostTrainingAdapter by @bbrowning in #1939
• feat: Add unit tests for NVIDIA safety by @JashG in #1897
• docs: Updated docs to show minimal RAG example and some other minor changes by @franciscojavierarceo in #1935
• feat: OpenAI-Compatible models, completions, chat/completions by @bbrowning in #1894
• docs: fixes to quick start by @raghotham in #1943
• docs: Update docs and fix warning in start-stack.sh by @aireilly in #1937
• feat: make training config fields optional by @cdoern in #1861
• fix: misleading help text for 'llama stack build' and 'llama stack run' by @nathan-weinberg in #1910
• feat: add batch inference API to llama stack inference by @ashwinb in #1945
• feat: update default tool use prompt by @ehhuang in #1803
• feat: support '-' in tool names by @ehhuang in #1807

New Contributors

• @wescpy made their first contribution in #1865
• @AlexHe99 made their first contribution in #1858
• @pdettori made their first contribution in #1608
• @Jaland made their first contribution in #1868
• @jiawenliu64 made their first contribution in #1920
• @ilya-kolchinsky made their first contribution in #1928
• @JashG made their first contribution in #1897
• @aireilly made their first contribution in #1937

Full Changelog: v0.2.1...v0.2.2