lemonade-sdk 8.1.4__py3-none-any.whl → 8.2.2__py3-none-any.whl

{lemonade_sdk-8.1.4.dist-info → lemonade_sdk-8.2.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lemonade-sdk
-Version: 8.1.4
+Version: 8.2.2
 Summary: Lemonade SDK: Your LLM Aide for Validation and Deployment
 Author-email: lemonade@amd.com
 Requires-Python: >=3.10, <3.14
@@ -16,24 +16,26 @@ Requires-Dist: numpy
 Requires-Dist: fasteners
 Requires-Dist: GitPython>=3.1.40
 Requires-Dist: psutil>=6.1.1
-Requires-Dist: wmi
+Requires-Dist: wmi; platform_system == "Windows"
 Requires-Dist: py-cpuinfo
 Requires-Dist: pytz
 Requires-Dist: zstandard
 Requires-Dist: fastapi
 Requires-Dist: uvicorn[standard]
-Requires-Dist: openai<1.97.1,>=1.81.0
+Requires-Dist: openai<3.0.0,>=2.0.0
 Requires-Dist: transformers<=4.53.2
 Requires-Dist: jinja2
 Requires-Dist: tabulate
 Requires-Dist: sentencepiece
 Requires-Dist: huggingface-hub[hf_xet]==0.33.0
 Requires-Dist: python-dotenv
+Requires-Dist: python-multipart
+Requires-Dist: rumps>=0.4.0; sys_platform == "darwin"
 Provides-Extra: oga-ryzenai
-Requires-Dist: onnxruntime-genai-directml-ryzenai==0.7.0.2.1; extra == "oga-ryzenai"
+Requires-Dist: onnxruntime-genai-directml-ryzenai==0.9.2.1; extra == "oga-ryzenai"
 Requires-Dist: protobuf>=6.30.1; extra == "oga-ryzenai"
 Provides-Extra: oga-cpu
-Requires-Dist: onnxruntime-genai==0.8.2; extra == "oga-cpu"
+Requires-Dist: onnxruntime-genai==0.9.2; extra == "oga-cpu"
 Requires-Dist: onnxruntime>=1.22.0; extra == "oga-cpu"
 Provides-Extra: dev
 Requires-Dist: torch>=2.6.0; extra == "dev"
@@ -45,42 +47,6 @@ Requires-Dist: human-eval-windows==1.0.4; extra == "dev"
 Requires-Dist: lm-eval[api]; extra == "dev"
 Provides-Extra: model-generate
 Requires-Dist: model-generate==1.5.0; (platform_system == "Windows" and python_version == "3.10") and extra == "model-generate"
-Provides-Extra: oga-hybrid
-Requires-Dist: lemonade-sdk[oga-ryzenai]; extra == "oga-hybrid"
-Provides-Extra: oga-unified
-Requires-Dist: lemonade-sdk[oga-ryzenai]; extra == "oga-unified"
-Provides-Extra: oga-hybrid-minimal
-Requires-Dist: lemonade-sdk[oga-ryzenai]; extra == "oga-hybrid-minimal"
-Provides-Extra: oga-cpu-minimal
-Requires-Dist: lemonade-sdk[oga-cpu]; extra == "oga-cpu-minimal"
-Provides-Extra: oga-npu-minimal
-Requires-Dist: lemonade-sdk[oga-ryzenai]; extra == "oga-npu-minimal"
-Provides-Extra: llm
-Requires-Dist: lemonade-sdk[dev]; extra == "llm"
-Provides-Extra: llm-oga-cpu
-Requires-Dist: lemonade-sdk[dev,oga-cpu]; extra == "llm-oga-cpu"
-Provides-Extra: llm-oga-npu
-Requires-Dist: onnx==1.16.0; extra == "llm-oga-npu"
-Requires-Dist: onnxruntime==1.18.0; extra == "llm-oga-npu"
-Requires-Dist: numpy==1.26.4; extra == "llm-oga-npu"
-Requires-Dist: protobuf>=6.30.1; extra == "llm-oga-npu"
-Requires-Dist: lemonade-sdk[dev]; extra == "llm-oga-npu"
-Provides-Extra: llm-oga-hybrid
-Requires-Dist: onnx==1.16.1; extra == "llm-oga-hybrid"
-Requires-Dist: numpy==1.26.4; extra == "llm-oga-hybrid"
-Requires-Dist: protobuf>=6.30.1; extra == "llm-oga-hybrid"
-Provides-Extra: llm-oga-unified
-Requires-Dist: lemonade-sdk[dev,llm-oga-hybrid]; extra == "llm-oga-unified"
-Provides-Extra: llm-oga-igpu
-Requires-Dist: onnxruntime-genai-directml==0.6.0; extra == "llm-oga-igpu"
-Requires-Dist: onnxruntime-directml<1.22.0,>=1.19.0; extra == "llm-oga-igpu"
-Requires-Dist: transformers<4.45.0; extra == "llm-oga-igpu"
-Requires-Dist: lemonade-sdk[dev]; extra == "llm-oga-igpu"
-Provides-Extra: llm-oga-cuda
-Requires-Dist: onnxruntime-genai-cuda==0.8.2; extra == "llm-oga-cuda"
-Requires-Dist: onnxruntime-gpu>=1.22.0; extra == "llm-oga-cuda"
-Requires-Dist: transformers<=4.51.3; extra == "llm-oga-cuda"
-Requires-Dist: lemonade-sdk[dev]; extra == "llm-oga-cuda"
 Dynamic: author-email
 Dynamic: description
 Dynamic: description-content-type
@@ -94,41 +60,31 @@ Dynamic: summary
 
 <p align="center">
   <a href="https://discord.gg/5xXzkMu8Zk">
-    <img src="https://img.shields.io/badge/Discord-7289DA?logo=discord&logoColor=white" alt="Discord" />
-  </a>
+    <img src="https://img.shields.io/badge/Discord-7289DA?logo=discord&logoColor=white" alt="Discord" /></a>
   <a href="https://github.com/lemonade-sdk/lemonade/tree/main/test" title="Check out our tests">
-    <img src="https://github.com/lemonade-sdk/lemonade/actions/workflows/test_lemonade.yml/badge.svg" alt="Lemonade tests" />
-  </a>
+    <img src="https://github.com/lemonade-sdk/lemonade/actions/workflows/test_lemonade.yml/badge.svg" alt="Lemonade tests" /></a>
   <a href="docs/README.md#installation" title="Check out our instructions">
-    <img src="https://img.shields.io/badge/Windows-11-0078D6?logo=windows&logoColor=white" alt="Windows 11" />
-  </a>
+    <img src="https://img.shields.io/badge/Windows-11-0078D6?logo=windows&logoColor=white" alt="Windows 11" /></a>
   <a href="https://lemonade-server.ai/#linux" title="Ubuntu 24.04 & 25.04 Supported">
-    <img src="https://img.shields.io/badge/Ubuntu-24.04%20%7C%2025.04-E95420?logo=ubuntu&logoColor=white" alt="Ubuntu 24.04 | 25.04" />
-  </a>
+    <img src="https://img.shields.io/badge/Ubuntu-24.04%20%7C%2025.04-E95420?logo=ubuntu&logoColor=white" alt="Ubuntu 24.04 | 25.04" /></a>
+  <a href="https://lemonade-server.ai/" title="macOS 14+ with Apple Silicon">
+    <img src="https://img.shields.io/badge/macOS-14%2B-000000?logo=apple&logoColor=white" alt="macOS 14+" /></a>
   <a href="docs/README.md#installation" title="Check out our instructions">
-    <img src="https://img.shields.io/badge/Python-3.10--3.13-blue?logo=python&logoColor=white" alt="Made with Python" />
-  </a>
+    <img src="https://img.shields.io/badge/Python-3.10--3.13-blue?logo=python&logoColor=white" alt="Made with Python" /></a>
   <a href="https://github.com/lemonade-sdk/lemonade/blob/main/docs/contribute.md" title="Contribution Guide">
-    <img src="https://img.shields.io/badge/PRs-welcome-brightgreen.svg" alt="PRs Welcome" />
-  </a>
+    <img src="https://img.shields.io/badge/PRs-welcome-brightgreen.svg" alt="PRs Welcome" /></a>
   <a href="https://github.com/lemonade-sdk/lemonade/releases/latest" title="Download the latest release">
-    <img src="https://img.shields.io/github/v/release/lemonade-sdk/lemonade?include_prereleases" alt="Latest Release" />
-  </a>
+    <img src="https://img.shields.io/github/v/release/lemonade-sdk/lemonade?include_prereleases" alt="Latest Release" /></a>
   <a href="https://tooomm.github.io/github-release-stats/?username=lemonade-sdk&repository=lemonade">
-    <img src="https://img.shields.io/github/downloads/lemonade-sdk/lemonade/total.svg" alt="GitHub downloads" />
-  </a>
+    <img src="https://img.shields.io/github/downloads/lemonade-sdk/lemonade/total.svg" alt="GitHub downloads" /></a>
   <a href="https://github.com/lemonade-sdk/lemonade/issues">
-    <img src="https://img.shields.io/github/issues/lemonade-sdk/lemonade" alt="GitHub issues" />
-  </a>
+    <img src="https://img.shields.io/github/issues/lemonade-sdk/lemonade" alt="GitHub issues" /></a>
   <a href="https://github.com/lemonade-sdk/lemonade/blob/main/LICENSE">
-    <img src="https://img.shields.io/badge/License-Apache-yellow.svg" alt="License: Apache" />
-  </a>
+    <img src="https://img.shields.io/badge/License-Apache-yellow.svg" alt="License: Apache" /></a>
   <a href="https://github.com/psf/black">
-    <img src="https://img.shields.io/badge/code%20style-black-000000.svg" alt="Code style: black" />
-  </a>
+    <img src="https://img.shields.io/badge/code%20style-black-000000.svg" alt="Code style: black" /></a>
   <a href="https://star-history.com/#lemonade-sdk/lemonade">
-    <img src="https://img.shields.io/badge/Star%20History-View-brightgreen" alt="Star History Chart" />
-  </a>
+    <img src="https://img.shields.io/badge/Star%20History-View-brightgreen" alt="Star History Chart" /></a>
 </p>
 <p align="center">
   <img src="https://github.com/lemonade-sdk/assets/blob/main/docs/banner.png?raw=true" alt="Lemonade Banner" />
@@ -156,7 +112,7 @@ Startups such as [Styrk AI](https://styrk.ai/styrk-ai-and-amd-guardrails-for-you
 ### Use it with your favorite OpenAI-compatible app!
 
 <p align="center">
-  <a href="https://lemonade-server.ai/docs/server/apps/open-webui/" title="Open WebUI" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/openwebui.jpg" alt="Open WebUI" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/continue/" title="Continue" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/continue_dev.png" alt="Continue" width="60" /></a>&nbsp;&nbsp;<a href="https://github.com/amd/gaia" title="Gaia" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/gaia.ico" alt="Gaia" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/anythingLLM/" title="AnythingLLM" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/anything_llm.png" alt="AnythingLLM" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/ai-dev-gallery/" title="AI Dev Gallery" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/ai_dev_gallery.webp" alt="AI Dev Gallery" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/lm-eval/" title="LM-Eval" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/lm_eval.png" alt="LM-Eval" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/codeGPT/" title="CodeGPT" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/codegpt.jpg" alt="CodeGPT" width="60" /></a>&nbsp;&nbsp;<a href="https://github.com/lemonade-sdk/lemonade/blob/main/docs/server/apps/ai-toolkit.md" title="AI Toolkit" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/ai_toolkit.png" alt="AI Toolkit" width="60" /></a>
+  <a href="https://lemonade-server.ai/docs/server/apps/open-webui/" title="Open WebUI" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/openwebui.jpg" alt="Open WebUI" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/continue/" title="Continue" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/continue_dev.png" alt="Continue" width="60" /></a>&nbsp;&nbsp;<a href="https://github.com/amd/gaia" title="Gaia" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/gaia.ico" alt="Gaia" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/anythingLLM/" title="AnythingLLM" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/anything_llm.png" alt="AnythingLLM" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/ai-dev-gallery/" title="AI Dev Gallery" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/ai_dev_gallery.webp" alt="AI Dev Gallery" width="60" /></a>&nbsp;&nbsp;<a href="https://lemonade-server.ai/docs/server/apps/lm-eval/" title="LM-Eval" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/lm_eval.png" alt="LM-Eval" width="60" /></a>&nbsp;&nbsp;<a href="https://github.com/lemonade-sdk/lemonade-arcade" title="Lemonade Arcade" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/lemonade-arcade/refs/heads/main/docs/assets/favicon.ico" alt="Lemonade Arcade" width="60" /></a>&nbsp;&nbsp;<a href="https://github.com/lemonade-sdk/lemonade/blob/main/docs/server/apps/ai-toolkit.md" title="AI Toolkit" target="_blank"><img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/ai_toolkit.png" alt="AI Toolkit" width="60" /></a>
 </p>
 
 > [!TIP]
@@ -200,11 +156,11 @@ You can also import custom GGUF and ONNX models from Hugging Face by using our [
 
 Lemonade supports the following configurations, while also making it easy to switch between them at runtime. Find more information about it [here](./docs/README.md#software-and-hardware-overview).
 
-| Hardware | Engine: OGA | Engine: llamacpp | Engine: HF | Windows | Linux |
-|----------|-------------|------------------|------------|---------|-------|
-| **🧠 CPU** | All platforms | All platforms | All platforms | ✅ | ✅ |
-| **🎮 GPU** | — | Vulkan: All platforms<br>ROCm: Selected AMD platforms* | — | ✅ | ✅ |
-| **🤖 NPU** | AMD Ryzen™ AI 300 series | — | — | ✅ | — |
+| Hardware | Engine: OGA | Engine: llamacpp | Engine: FLM | Windows | Linux | macOS |
+|----------|-------------|------------------|------------|---------|-------|-------|
+| **🧠 CPU** | All platforms | All platforms | - | ✅ | ✅ | ✅ |
+| **🎮 GPU** | — | Vulkan: All platforms<br>ROCm: Selected AMD platforms*<br>Metal: Apple Silicon | — | ✅ | ✅ | ✅ |
+| **🤖 NPU** | AMD Ryzen™ AI 300 series | — | Ryzen™ AI 300 series | ✅ | — | — |
 
 <details>
 <summary><small><i>* See supported AMD ROCm platforms</i></small></summary>
@@ -227,7 +183,7 @@ Lemonade supports the following configurations, while also making it easy to swi
     </tr>
     <tr>
       <td><b>gfx120X</b> (RDNA4)</td>
-      <td>Windows only</td>
+      <td>Windows, Ubuntu</td>
       <td>Radeon AI PRO R9700, RX 9070 XT/GRE/9070, RX 9060 XT</td>
     </tr>
     <tr>
@@ -309,6 +265,7 @@ This project is:
   - [OnnxRuntime GenAI](https://github.com/microsoft/onnxruntime-genai)
   - [Hugging Face Hub](https://github.com/huggingface/huggingface_hub)
   - [OpenAI API](https://github.com/openai/openai-python)
+  - [IRON/MLIR-AIE](https://github.com/Xilinx/mlir-aie)
   - and more...
 - Accelerated by mentorship from the OCV Catalyst program.
 - Licensed under the [Apache 2.0 License](https://github.com/lemonade-sdk/lemonade/blob/main/LICENSE).

lemonade_sdk-8.2.2.dist-info/RECORD

@@ -0,0 +1,83 @@
+lemonade/__init__.py,sha256=W1Qk7r0rnQqFhPNHp6BIBT_q-OH3s-8Q_POoVfAmKW0,117
+lemonade/api.py,sha256=Oc4yBA3LZg8FrTsbuDq1p9-XE74pqNnIEUhXyKa7qg8,5786
+lemonade/cache.py,sha256=fUjtHYkRdHTULjNMrDNOFDGW_QMVUg54ZE1NukBP2oM,3314
+lemonade/cli.py,sha256=qU5bW7RQAUKNSpvrhVyzn68NMxyi-336Ke_JU4bsv1Q,5708
+lemonade/sequence.py,sha256=KSH7BPsiyDKsOsg_ziQKEGsDwMmuO_YbgPRBxkZd0pw,13267
+lemonade/state.py,sha256=sdSezla7Cd7KYL90xY3p9kcNV4ndSyN6UvNLOr3vBMA,5261
+lemonade/version.py,sha256=qHoUgbIt6pjWzUYt9wXlO8qcrd9dEl09FyPsRTkJshI,22
+lemonade/common/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+lemonade/common/build.py,sha256=zTb0m1-kuUx6zw5QHp2SNnVuN6jOTMQ2FCdj9iH374U,6140
+lemonade/common/cli_helpers.py,sha256=hjBfXrTtFl8gmCFlL-ksviXR0mOcdPtTWVNKoEp3PG4,4993
+lemonade/common/exceptions.py,sha256=w83sVKmL1QXoJlGjj_bRyjIBMhlMqdVQy_FEOTu2YQI,2050
+lemonade/common/filesystem.py,sha256=QV3cHhKNu-7W2rr8wZ4JQfD2rP_5T2Js7jiDQBYWHVQ,12142
+lemonade/common/inference_engines.py,sha256=3bUGQe9wtfTiwt8kvI_ry077uyc9lid2G1fJX95kN1A,12969
+lemonade/common/network.py,sha256=qXpUjDYQEYM_gH3JwTtU-pu_yCKcaa1IeohJRPy91-A,2903
+lemonade/common/printing.py,sha256=GFFzrXIineIOMa9yu0lo5sL4j6A5BBg_T9aUCdP-juw,3229
+lemonade/common/status.py,sha256=dxAahChPGVmfT91DJW949Xjgm9r5E-Y9KOLPEw7BMh8,16562
+lemonade/common/system_info.py,sha256=PWpEtOz6CqvnUYL8_3lX61_GNbrbp0O2x4ZIwRh6RJg,48780
+lemonade/common/test_helpers.py,sha256=Gwk-pa_6xYAo2oro-2EJNfuouAfw8k_brCbcMC-E-r0,758
+lemonade/profilers/__init__.py,sha256=JKVonvJ4XZ9_6sKXPWsiMLQCNyzQOxhQw5BEHR1qOfU,31
+lemonade/profilers/agt_power.py,sha256=t_37VEg8LPapjSKSjJln-jFznZtTIf5UpzlAXcVGOrc,16771
+lemonade/profilers/hwinfo_power.py,sha256=UQr-EHq7B4T-IvzmErCRK0-QxcFnho4ftCaWy5p8Qvo,15819
+lemonade/profilers/memory_tracker.py,sha256=1iuKt0FmNVYLDnOc-oZM8dX9TUksvoxO0m2EoYWjhYQ,9367
+lemonade/profilers/profiler.py,sha256=Y5FSbc386bMlTVbqCuya9pYrso5aTthxahR1V_ZKQ9E,1902
+lemonade/tools/__init__.py,sha256=_6xRc-FHxmujoLjLjWtpYrWYEXtCSneSy-5ya01kyPk,53
+lemonade/tools/accuracy.py,sha256=X2OLzx3VDhpOZO9aHKH3vrjeeKI1lTYXZFeZF9Qpslg,15188
+lemonade/tools/adapter.py,sha256=fFRSnLiQk20WYQmYFD0JK1t-fDLKEWhD6u9qqB_BECk,3496
+lemonade/tools/bench.py,sha256=jyL85yqBPzyT2pr9OWzihvyoN0v3JQg_BHtIHy9d2zg,10881
+lemonade/tools/humaneval.py,sha256=JbxuoOzvR4iyxZv4R6MI7a3gUt5ef_Jj6Ie-9VP2wzY,9531
+lemonade/tools/management_tools.py,sha256=HQBcr7LYuMqVRYQtvnkNpfutBTA7lblszyoAjjVGu1Y,10201
+lemonade/tools/mmlu.py,sha256=c2QaIMDzjqxCvgHlMXmy_dP1sAFkwkDxL7RO2nogI6s,11071
+lemonade/tools/perplexity.py,sha256=eiaTZ3yhqF2pfwOffVbKKJLwjSri7Im2pC-tBJr7LLU,5638
+lemonade/tools/prompt.py,sha256=PyLksp1k8jsZsU7XBRK61k1DUHhbdLa20h-AP8Noh3w,9011
+lemonade/tools/tool.py,sha256=UsxVYukfm_iM3BfeGYPZxQlTK5UfDfDOl3RIyLr8A1Y,13256
+lemonade/tools/flm/__init__.py,sha256=NQ4CEzJZGS_VvxPMlfrK4Dcx48bQSoUR4iG8e7yZjas,46
+lemonade/tools/flm/utils.py,sha256=5KS0IND_8E3lgKNdIFeJqMcdiCsPpYCjgL1FzH8YvGk,9933
+lemonade/tools/huggingface/bench.py,sha256=OMkIm_6o4znu8xZo4MzvHULGu0WPQl529kpAngJut-g,6907
+lemonade/tools/huggingface/load.py,sha256=KsSGOBBD-tNEIfYC8mCWV_jpnkjHMhN3juVmC1Ln4uQ,7745
+lemonade/tools/huggingface/utils.py,sha256=j1S-IgjDsznUIVwkHSqqChmFyqIx9f3WcEelzohWwvU,13955
+lemonade/tools/llamacpp/bench.py,sha256=JfsKboJCDdkZ5ZJ9Xl-ty5cCBj6pAlUrXZiEQ7Tt60c,9689
+lemonade/tools/llamacpp/load.py,sha256=YvDHqXe4tKGuT5aG_wP4xH3Px0RmG92u73kVOTHSEgc,7671
+lemonade/tools/llamacpp/utils.py,sha256=kHL2xaCnQ7hLY15FoQJLYGf29KddxoEyczXmN1TWfQY,46182
+lemonade/tools/oga/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+lemonade/tools/oga/bench.py,sha256=V-qc8mbFHvOv13QTrgFOXSmUAhzzGTK8S7UpVCl-KIw,4329
+lemonade/tools/oga/load.py,sha256=yxeRPUUrDOgrjv9d_R8MGLInunvI7HId_Q2cN8K_8h4,31113
+lemonade/tools/oga/migration.py,sha256=zo34chEbtkWk4GaDwgW326o09eF2wAsoOiTzqQ6DMLU,13505
+lemonade/tools/oga/utils.py,sha256=F8UVLKlfYcLa2SUqlehar8-jaX2Aw4u58DjHNNvLdOA,17675
+lemonade/tools/report/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+lemonade/tools/report/llm_report.py,sha256=bVHhwCINA-Ok2EdSwAsLubsc83N3KWOVuwTguw7jDcE,6676
+lemonade/tools/report/table.py,sha256=7hsJNr9x9yxvpTTVp64FpTAfxyUDvXHcDvGk5L84j1g,30187
+lemonade/tools/server/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+lemonade/tools/server/flm.py,sha256=NVpjFzcmX6YzW6wCWyMEvQEe_QpRK-Jlcaajxy3kWuo,4025
+lemonade/tools/server/llamacpp.py,sha256=0scaDxJXne-PfUGh-r25UyqDSIOJdFkD1Js1WPFPzv0,11676
+lemonade/tools/server/serve.py,sha256=WYsKZTtPFp6JAp5K-N734wTsD3sltiTYrwV6oCs7ZWk,83487
+lemonade/tools/server/tool_calls.py,sha256=xrAlQwKG-nv2xLlf8f9CDSaUbyMn8ZtHkds9iZLG9K8,5230
+lemonade/tools/server/tray.py,sha256=EFnSc2Ra4owiHVz6ykoMhxi2fYqZAK1g21AynAYBiyk,24426
+lemonade/tools/server/webapp.py,sha256=GGSVIzN19C2ZaadOEPBg_D7Lt0PuF339NuWwjMPfZu8,1225
+lemonade/tools/server/wrapped_server.py,sha256=uh7ifrRX1Hx0IuRwZRCGPyQOukitE7kKQipCCz0bSGA,19844
+lemonade/tools/server/static/favicon.ico,sha256=hMmP9qGJNeZ0mFS86JIqPbZstXMZn0Z76_HfHQpREAU,126745
+lemonade/tools/server/static/logs.html,sha256=b9BtKWQwXJgr8xc2lWLDTx1zwYgM0Nsv5ZHRT_EdBM8,6157
+lemonade/tools/server/static/styles.css,sha256=XOi3-Aw_W0WOBEGBnyYZcniwjeHx7dhRBU2Lt4ePN00,52815
+lemonade/tools/server/static/webapp.html,sha256=QgUcUo7coLNDpy2HuF7IHu1t6wySSTNwcZQ2-Q8WqlU,21506
+lemonade/tools/server/static/js/chat.js,sha256=xil7PrnGohiNGtlqBcZ3upBr9hw5Y0P6KYqjbHpUYh0,39812
+lemonade/tools/server/static/js/model-settings.js,sha256=MZZkflb7cRaRqSagyknR_Pgii6pbDSlVKiAbgYqrGsQ,6737
+lemonade/tools/server/static/js/models.js,sha256=OirBQ7dEflazOaPwr4XtCjqcuaW_SUZY6vMkAAhctFg,53178
+lemonade/tools/server/static/js/shared.js,sha256=QPekwJQ9BLD2c0D6-CkoB0VjsehVy05TF8vTQmWGf-c,19312
+lemonade/tools/server/utils/macos_tray.py,sha256=xwHW44ZN5hDVlJcwIpHHfqn4VRXWxXHuDACaT-ZqdO8,7095
+lemonade/tools/server/utils/port.py,sha256=J7-g-Aqygb50jNoHLhhRfBZVM-uhGlcB5-oYBAehvgw,2263
+lemonade/tools/server/utils/thread.py,sha256=Z-PDzGcpgfN2qxTmtlROWqrUN0B2fXdPrqo_J10fR_w,2772
+lemonade/tools/server/utils/windows_tray.py,sha256=2z5aTmUPlkT-QfkcfwHsyA6dv6nSNBT0gXUErarhac8,13170
+lemonade_install/__init__.py,sha256=26zohKg2jgr_5y7tObduWMYQg8zCTWMZHL8lfi2zZVQ,40
+lemonade_install/install.py,sha256=3NtQ4JayFM9VW7IOHommaV4s9RgdightyfhhxQ190O4,7753
+lemonade_sdk-8.2.2.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+lemonade_sdk-8.2.2.dist-info/licenses/NOTICE.md,sha256=RSca9LE5e6pvdWA_LXAUCcACIHPmINKqkRX-AVRqBGo,3499
+lemonade_server/cli.py,sha256=BItlVpuykHwTnCZJs-7v1qWRPBCnEx-aeP43a7S9uFk,23918
+lemonade_server/model_manager.py,sha256=dboE-7wF1Y9iLGUcBkaA-lpF8XJKSIN_y0mUYzgCm2w,32657
+lemonade_server/pydantic_models.py,sha256=QpsQrBjplv0IlwwUEt259o3eGv5YUuhq_sSiX2xBAyM,4074
+lemonade_server/server_models.json,sha256=cZAxlDkxQ5neb57jQ8j2fvIeuAfyk4Ga1Yele58OeTY,15062
+lemonade_server/settings.py,sha256=JOlZmirUXO9rA6BCODVFwyXrrHtYoH_LiKYm49lGm_c,1260
+lemonade_sdk-8.2.2.dist-info/METADATA,sha256=e_usINpNwof9VkWrKlpWNqyLH3UD0j6D6zX_Bimynww,15417
+lemonade_sdk-8.2.2.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+lemonade_sdk-8.2.2.dist-info/entry_points.txt,sha256=7sRvpNhi1E7amnM7RZo57e8yFF9iA5uuRaIeJ1Xre6w,193
+lemonade_sdk-8.2.2.dist-info/top_level.txt,sha256=10ap5GNiPhalO4V50LRoxA1FqRT9g3Xkia6BITu880k,42
+lemonade_sdk-8.2.2.dist-info/RECORD,,

lemonade_server/cli.py

@@ -1,6 +1,7 @@
 import argparse
 import sys
 import os
+import platform
 from typing import Tuple, Optional
 import psutil
 from typing import List
@@ -104,12 +105,34 @@ def serve(
         max_wait_time = 30
         wait_interval = 0.5
         waited = 0
-        while waited < max_wait_time:
-            time.sleep(wait_interval)
-            _, running_port = get_server_info()
-            if running_port is not None:
-                break
-            waited += wait_interval
+
+        if platform.system() == "Darwin":
+            # On macOS, use direct HTTP health check instead of process scanning for better
+            # performance
+            import requests
+
+            while waited < max_wait_time:
+                time.sleep(wait_interval)
+                try:
+                    response = requests.get(
+                        f"http://{host}:{port}/api/v1/health", timeout=1
+                    )
+                    if response.status_code == 200:
+                        break
+                except (
+                    requests.exceptions.ConnectionError,
+                    requests.exceptions.Timeout,
+                ):
+                    pass  # Server not ready yet
+                waited += wait_interval
+        else:
+            # On other platforms, use the existing approach
+            while waited < max_wait_time:
+                time.sleep(wait_interval)
+                _, running_port = get_server_info()
+                if running_port is not None:
+                    break
+                waited += wait_interval
 
         return port, server_thread
 
@@ -144,8 +167,12 @@ def stop():
                 except psutil.NoSuchProcess:
                     pass  # Child already terminated
 
-        # Wait for main process
-        process.wait(timeout=10)
+        # Wait for main process to terminate gracefully
+        # kill if it doesn't terminate gracefully
+        try:
+            process.wait(timeout=5)
+        except psutil.TimeoutExpired:
+            process.kill()
 
         # Kill llama-server child process if it didn't terminate gracefully
         for child in children:
@@ -172,6 +199,7 @@ def pull(
     checkpoint: Optional[str] = None,
     recipe: Optional[str] = None,
     reasoning: bool = False,
+    vision: bool = False,
     mmproj: str = "",
 ):
     """
@@ -193,15 +221,16 @@ def pull(
         for model_name in model_names:
             payload = {"model_name": model_name}
 
-            if checkpoint and recipe:
-                # Add the parameters for registering a new model
-                payload["checkpoint"] = checkpoint
-                payload["recipe"] = recipe
-
-                if reasoning:
-                    payload["reasoning"] = reasoning
-                if mmproj:
-                    payload["mmproj"] = mmproj
+            # Add the parameters to the payload
+            for key, value in [
+                ("checkpoint", checkpoint),
+                ("recipe", recipe),
+                ("reasoning", reasoning),
+                ("vision", vision),
+                ("mmproj", mmproj),
+            ]:
+                if value:
+                    payload[key] = value
 
             # Install the model
             pull_response = requests.post(f"{base_url}/pull", json=payload)
@@ -220,7 +249,11 @@ def pull(
             checkpoint=checkpoint,
             recipe=recipe,
             reasoning=reasoning,
+            vision=vision,
             mmproj=mmproj,
+            # The pull command will download an upgraded model if available, even
+            # if we already have a local copy of the model
+            do_not_upgrade=False,
         )
 
 
@@ -273,6 +306,11 @@ def run(
     """
     import webbrowser
     import time
+    import os
+
+    # Disable tray on macOS for run command due to threading issues
+    if platform.system() == "Darwin":
+        tray = False
 
     # Start the server if not running
     _, running_port = get_server_info()
@@ -288,6 +326,16 @@ def run(
             ctx_size=ctx_size,
         )
     else:
+        # macOS: Check for port conflicts when server is already running
+        if platform.system() == "Darwin":
+            requested_port = port if port is not None else DEFAULT_PORT
+            if running_port != requested_port:
+                print(
+                    f"Lemonade Server is already running on port {running_port}\n"
+                    f"You requested port {requested_port}. Please stop the existing server first "
+                )
+                sys.exit(ExitCodes.SERVER_ALREADY_RUNNING)
+
         port = running_port
 
     # Pull model
@@ -299,7 +347,10 @@ def run(
     # Open the webapp with the specified model
     url = f"http://{host}:{port}/?model={model_name}#llm-chat"
     print(f"You can now chat with {model_name} at {url}")
-    webbrowser.open(url)
+
+    # Only open browser if not disabled via environment variable
+    if not os.environ.get("LEMONADE_DISABLE_BROWSER"):
+        webbrowser.open(url)
 
     # Keep the server running if we started it
     if not server_previously_running:
@@ -356,6 +407,23 @@ def is_lemonade_server(pid):
     """
     Check whether or not a given PID corresponds to a Lemonade server
     """
+    # macOS only: Self-exclusion to prevent blocking server startup
+    if platform.system() == "Darwin":
+        current_pid = os.getpid()
+        if pid == current_pid:
+            return False
+
+        # Exclude children of current process to avoid detecting status commands
+        try:
+            current_process = psutil.Process(current_pid)
+            child_pids = [
+                child.pid for child in current_process.children(recursive=True)
+            ]
+            if pid in child_pids:
+                return False
+        except (psutil.NoSuchProcess, psutil.AccessDenied):
+            pass
+
     try:
         process = psutil.Process(pid)
 
@@ -364,11 +432,29 @@ def is_lemonade_server(pid):
             if process_name in [  # Windows
                 "lemonade-server-dev.exe",
                 "lemonade-server.exe",
+                "lsdev.exe",
             ] or process_name in [  # Linux
                 "lemonade-server-dev",
                 "lemonade-server",
+                "lsdev",
             ]:
                 return True
+            # macOS only: Python scripts appear as "python3.x", check command line
+            elif process_name.startswith("python") and platform.system() == "Darwin":
+                try:
+                    cmdline = process.cmdline()
+                    if len(cmdline) >= 2:
+                        script_path = cmdline[1]
+                        # Check for various lemonade server command patterns (macOS only)
+                        lemonade_patterns = [
+                            "lemonade-server-dev",
+                            "lemonade-server",
+                            "lsdev",  # Short alias for lemonade-server-dev
+                        ]
+                        if any(pattern in script_path for pattern in lemonade_patterns):
+                            return True
+                except (psutil.AccessDenied, psutil.NoSuchProcess):
+                    pass
             elif "llama-server" in process_name:
                 return False
             if not process.parent():
@@ -386,23 +472,43 @@ def get_server_info() -> Tuple[int | None, int | None]:
     2. The port that Lemonade Server is running on
     """
 
-    # Get all network connections and filter for localhost IPv4 listening ports
+    # Try the global approach first (works on Windows/Linux without permissions)
     try:
         connections = psutil.net_connections(kind="tcp4")
-
         for conn in connections:
-            if (
-                conn.status == "LISTEN"
-                and conn.laddr
-                and conn.laddr.ip in ["localhost"]
-                and conn.pid is not None
-            ):
+            if conn.status == "LISTEN" and conn.laddr and conn.pid is not None:
                 if is_lemonade_server(conn.pid):
                     return conn.pid, conn.laddr.port
-
-    except Exception:
+    except (psutil.AccessDenied, PermissionError):
+        # Global approach needs elevated permissions on macOS, fall back to per-process approach
+        pass
+    except Exception:  # pylint: disable=broad-exception-caught
         pass
 
+    # Per-process approach (macOS only - needs this due to permission requirements)
+    if platform.system() == "Darwin":
+        try:
+            for proc in psutil.process_iter(["pid", "name"]):
+                try:
+                    pid = proc.info["pid"]
+                    if is_lemonade_server(pid):
+                        # Found a lemonade server, check its listening ports
+                        connections = proc.net_connections(kind="inet")
+                        for conn in connections:
+                            if conn.status == "LISTEN" and conn.laddr:
+                                return pid, conn.laddr.port
+                        # If no listening connections found, this process is not actually serving
+                        # Continue looking for other processes
+                except (
+                    psutil.NoSuchProcess,
+                    psutil.AccessDenied,
+                    psutil.ZombieProcess,
+                ):
+                    # Some processes may be inaccessible, continue to next
+                    continue
+        except Exception:  # pylint: disable=broad-exception-caught
+            pass
+
     return None, None
 
 
@@ -417,12 +523,13 @@ def list_models():
 
     # Get all supported models and downloaded models
     supported_models = model_manager.supported_models
+    filtered_models = model_manager.filter_models_by_backend(supported_models)
     downloaded_models = model_manager.downloaded_models
 
     # Filter to only show recommended models
     recommended_models = {
         model_name: model_info
-        for model_name, model_info in supported_models.items()
+        for model_name, model_info in filtered_models.items()
         if model_info.get("suggested", False)
     }
 
@@ -499,7 +606,7 @@ def _add_server_arguments(parser):
         "--llamacpp",
         type=str,
         help="LlamaCpp backend to use",
-        choices=["vulkan", "rocm"],
+        choices=["vulkan", "rocm", "metal"],
         default=DEFAULT_LLAMACPP_BACKEND,
     )
     parser.add_argument(
@@ -512,6 +619,13 @@ def _add_server_arguments(parser):
         default=DEFAULT_CTX_SIZE,
     )
 
+    if os.name == "nt" or platform.system() == "Darwin":
+        parser.add_argument(
+            "--no-tray",
+            action="store_true",
+            help="Do not show a tray icon when the server is running",
+        )
+
 
 def main():
     parser = argparse.ArgumentParser(
@@ -532,12 +646,6 @@ def main():
     # Serve command
     serve_parser = subparsers.add_parser("serve", help="Start server")
     _add_server_arguments(serve_parser)
-    if os.name == "nt":
-        serve_parser.add_argument(
-            "--no-tray",
-            action="store_true",
-            help="Do not show a tray icon when the server is running",
-        )
 
     # Status command
     status_parser = subparsers.add_parser("status", help="Check if server is running")
@@ -611,7 +719,7 @@ def main():
 
     args = parser.parse_args()
 
-    if os.name != "nt":
+    if os.name != "nt" and platform.system() != "Darwin":
         args.no_tray = True
 
     if args.version: