经 AI Skill Hub 精选评估,AMD Strix Halo 获评「推荐使用」。这款AI工具在功能完整性、社区活跃度和易用性方面表现出色,AI 评分 7.5 分,适合有一定技术背景的用户使用。
AMD Strix Halo 是一款基于 Python 开发的开源工具,专注于 installable、ai、amd-ryzen-ai 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
AMD Strix Halo 是一款基于 Python 开发的开源工具,专注于 installable、ai、amd-ryzen-ai 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
# 方式一:pip 安装(推荐)
pip install hal0
# 方式二:虚拟环境安装(推荐生产环境)
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install hal0
# 方式三:从源码安装(获取最新功能)
git clone https://github.com/Hal0ai/hal0
cd hal0
pip install -e .
# 验证安装
python -c "import hal0; print('安装成功')"
# 命令行使用
hal0 --help
# 基本用法
hal0 input_file -o output_file
# Python 代码中调用
import hal0
# 示例
result = hal0.process("input")
print(result)
# hal0 配置文件示例(config.yml) app: name: "hal0" debug: false log_level: "INFO" # 运行时指定配置文件 hal0 --config config.yml # 或通过环境变量配置 export HAL0_API_KEY="your-key" export HAL0_OUTPUT_DIR="./output"
<picture> <source media="(prefers-color-scheme: dark)" srcset="./ui/public/brand/logo-halo-dark.svg"> <img src="./ui/public/brand/logo-halo-light.svg" alt="hal0" width="220"> </picture>
- OpenAI-compatible /v1/* API — chat, completions, embeddings, rerank, audio transcriptions, audio speech, image generations, image edits, models. Drop-in for any OpenAI SDK; point your client at http://localhost:8080/v1 and go. - Slots — each named target in capabilities.toml carries a Lemonade-vocab type (`llm | embedding | reranking | transcription | tts | image), a device (gpu-rocm | gpu-vulkan | cpu | npu`), a model, plus enabled and optional default. Six seeded slots (primary, embed, rerank, stt, tts, img) plus three NPU slots (agent, stt-npu, embed-npu) when FastFlowLM is installed. User-added slots via hal0 slot add NAME --type TYPE --model MODEL. - Lemonade as the unified runtime — one lemond process, loopback-only on :13305, supervised by hal0-lemonade.service. Cache + config at /var/lib/hal0/lemonade/. The hal0 capability layer is the only user-facing inference surface; Lemonade is treated as an internal runtime, never exposed off-box. - Bundle picker on first run — capabilities.toml ships empty by design. The dashboard's first load surfaces four hardware-anchored tiers (hal0-Lite ≥16 GB / Default ≥32 GB / Pro ≥64 GB / Max ≥100 GB) plus the vendor-blessed LMX-Omni-52B-Halo kit. Tiers that don't fit the detected unified RAM grey out with a tooltip. See ADR-0010 for the no-silent-default rationale. - Hardware-aware probe — detects GPU / NPU / unified memory, writes /etc/hal0/hardware.json, surfaces VRAM/RAM fit warnings inline in the slot form and the bundle picker. - Dispatcher — registry-aware routing, cold-cache prefetch, upstream fallback (OpenRouter, Anthropic, OpenAI, custom OpenAI-shaped endpoints). Mix local + remote per-model in one config. - Dashboard — Vue 3 + Tailwind 4 UI for slot/model management, hardware-aware configuration, live logs, and system health. SSE-backed status + log tail. Lemonade /logs/stream folded into the Journal panel. Settings → Lemonade admin panel surfaces the daemon's /internal/config for inspection. Dark by default. - OpenWebUI prewired — chat at :3001, zero config. The installer writes openwebui.env pointing at the local hal0 API. - OmniRouter (8 tools) — generate_image, edit_image, text_to_speech, transcribe_audio, analyze_image (vision), embed_text, rerank_documents, route_to_chat. Dispatched client-side from chat slots; dynamically filtered per request. - Image generation, day one — POST /v1/images/generations via sd-cpp (Lemonade-bundled). Bundle manifests pre-pick SDXL Turbo / Flux-2-Klein-9B as fits the tier. - First-run wizard + bundle picker — bundle pick (or "Skip — configure manually") → models download in background. - Atomic self-update with rollback — `hal0 update --channel stable|nightly`. Cosign-verified tarballs swap a /usr/lib/hal0/current symlink; --rollback reverts. - One-line install — curl -fsSL https://hal0.dev/install.sh | bash (--models-dir=PATH or HAL0_MODELS_DIR=PATH redirects model pulls off /var/lib/hal0/models). The bootstrap fetches the release manifest, sha256-verifies the tarball, cosign-verifies the signature against the workflow OIDC identity, then hands off to installer/install.sh. v0.1.x installs are detected and refused with explicit backup/wipe instructions (see docs/v0.2-upgrade.md).
```sh
If hal0 runs inside a Proxmox LXC, the container only sees its own cgroup slice of memory — other tenants, ZFS ARC, and the host kernel draw from the same physical DIMMs as GPU GTT but are invisible from inside. To surface that, drop a read-only PVEAuditor API token into the dashboard's Settings → "Proxmox integration" panel. Once saved, the unified-memory bar swaps to the physical host's DIMM total and adds a muted "Proxmox host" segment for other-tenant + kernel pressure. Token is sensitive and stored 0600 at /etc/hal0/proxmox.json; the API never echoes it back. Bare-metal and VM installs leave the panel off and the dashboard stays quiet.
该项目提供了一个开源的自主式家用AI推理平台,支持AMD Strix Halo多卡环境,快速部署和高效运算,但其社区和文档可能需要进一步完善。
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ Apache 2.0 — 宽松开源协议,可商用,需保留版权声明和 NOTICE 文件,含专利授权条款。
AI Skill Hub 点评:AMD Strix Halo 的核心功能完整,质量良好。对于AI 技术爱好者来说,这是一个值得纳入个人工具库的选择。建议先在非生产环境试用,再逐步推广。
| 原始名称 | hal0 |
| 原始描述 | 开源AI工具:Open-source self-hosted home AI inference platform for AMD Strix Halo — multi-ba。⭐6 · Python |
| Topics | installableaiamd-ryzen-aifastapihomelabigpupython |
| GitHub | https://github.com/Hal0ai/hal0 |
| License | Apache-2.0 |
| 语言 | Python |
收录时间:2026-05-23 · 更新时间:2026-05-30 · License:Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。