AI／ML ワークステーション・クラスタ
ローカル LLM・エージェント基盤・GPU インフラと、その周辺のデータパイプライン。
AI／ML プリセットは、ローカル LLM ランタイム（Ollama・vLLM・LM Studio）、エージェント基盤（OpenClaw・Hermes・DeerFlow・LangGraph・CrewAI）、ベクターデータベース、GPU インフラ、ML ワークベンチをまとめて監視。ガード付き AI 管理者を備え、秘密情報はローカルに保持します。
この用途で対応するサービス
この用途で FrontierStack がインストール・接続・監視できる 13 カテゴリ 168 サービス。アプリでこのプリセットを選ぶと、まとめて表示されます。
AI / LLMs27
- AnythingLLM — Self-hosted RAG chat app
- Cerebras — Very fast hosted inference on wafer-scale hardware
- ChatLLM Teams — Abacus.AI's multi-model team chat & agent workspace (cloud)
- CrewAI (open source) — Multi-agent crews — Python framework + CLI (local)
- Darkbloom — Private AI inference marketplace on idle Apple Silicon Macs
- Dyad — Local open-source AI app builder (desktop)
- Hugging Face — Model hub, Inference API & local tooling
- IBM watsonx.ai — IBM's enterprise AI platform — Granite + third-party models (cloud/on-prem)
- Jan — Private, offline AI desktop app (OpenAI-compatible)
- LibreChat — Self-hosted multi-provider AI chat UI
- LiteLLM — Proxy/gateway for 100+ LLM APIs (OpenAI-compatible)
- llama.cpp Server — Lightweight local LLM server (OpenAI-compatible)
- LocalAI — Self-hosted, OpenAI-compatible inference server
- Mem0 — Memory layer for AI agents — self-hosted (Docker) or cloud
- MLX — Apple-silicon model serving via MLX — CLI server or the oMLX menu-bar app
- MLX (Apple) — Apple's ML framework for Apple Silicon
- Modular MAX — Modular's inference server — OpenAI-compatible on :8000
- Needle 2 — 45M on-device model for tool calling and structured extraction
- NotebookLM — Google's source-grounded research notebook (cloud)
- Open Computer — Virtual OS for AI agents — QEMU VM per agent (Mintplex Labs)
- Open Notebook — Open-source, self-hosted NotebookLM alternative
- OpenCode — Open-source AI coding agent (terminal)
- PrivateGPT — Ask questions of your documents, 100% offline (RAG)
- Tetrate Agent Router — Hosted LLM router — one OpenAI-compatible API (cloud)
- TurboFieldfare — Gemma 4 26B-A4B on Apple Silicon in ~2 GB of RAM
- Vane (Perplexica) — Self-hosted AI answer engine — the open-source Perplexity alternative
- vLLM — High-throughput LLM inference server (OpenAI-compatible)
Knowledge & Memory24
- AFFiNE — Docs + whiteboard + database (Notion/Miro alt.)
- AppFlowy — Open-source Notion alternative (docs/boards/DBs)
- BookStack — Self-hosted docs/wiki organised as books (PHP)
- Docmost — Open-source collaborative wiki & docs (Confluence/Notion alt.)
- DokuWiki — Flat-file PHP wiki — the closest macOS Server “Wiki” replacement
- Graphify — Turn a codebase into a queryable knowledge graph (AI coding-assistant skill)
- GraphRAG — Graph-based retrieval-augmented generation (Microsoft)
- Joplin Server — Sync server for Joplin notes (E2EE)
- Karakeep — AI bookmark/read-it-later (formerly Hoarder)
- Khoj — Self-hosted AI second brain — chat with your notes, docs and the web
- Linkding — Minimal, fast self-hosted bookmark manager
- Logseq — Local-first outliner & PKM (Markdown/Org)
- Lucid (Lucidchart & Lucidspark) — Visual collaboration — diagrams & whiteboards (cloud)
- MediaWiki — The wiki engine behind Wikipedia (PHP/MySQL)
- Memos — Lightweight, privacy-first notes/memo hub
- Notion — Workspace — docs, wikis, databases (cloud)
- Outline — Self-hosted team knowledge base (real-time)
- Readwise — Highlights sync & read-later (cloud API)
- Shiori — Simple self-hosted bookmark manager (Go)
- SurfSense — Self-hosted NotebookLM/Perplexity alternative
- Trilium Notes — Hierarchical personal notes (self-hosted server + app)
- Wallabag — Self-hosted read-it-later (Pocket alternative)
- Wiki.js — Modern self-hosted wiki (Node.js)
- Zotero — Reference & research manager (app + API)
AI Governance & Safety16
- Arize Phoenix — Open-source LLM tracing & evaluation (self-hosted)
- Credo AI — AI governance, risk & compliance platform (cloud)
- Fiddler AI — AI observability & model monitoring with explainability (cloud)
- garak — LLM vulnerability scanner / red-teaming (self-hosted CLI)
- Guardrails AI — Input/output validation guardrails for LLMs (Python)
- Helicone — LLM observability & gateway — logs, costs, caching (self-hosted / cloud)
- Holistic AI — AI governance, risk & audit platform (cloud + OSS)
- Lakera Guard — Real-time GenAI security — prompt-injection firewall (cloud / self-host)
- Langfuse — Open-source LLM observability & tracing (self-hosted / cloud)
- NeMo Guardrails — Programmable guardrails for LLM conversations (NVIDIA, Python)
- OpenLLMetry — OpenTelemetry instrumentation for LLM apps (SDK)
- Promptfoo — Prompt/RAG testing, evals & LLM red-teaming (self-hosted CLI)
- Protect AI — AI/ML security — model scanning & ML supply chain (cloud + OSS)
- Ragas — Evaluation framework for RAG pipelines (Python)
- TruLens — Evaluation & tracking for LLM apps — feedback functions (Python)
- WhyLabs — AI observability & data/LLM monitoring (cloud + whylogs)
Databases16
- Apache Cassandra — Distributed wide-column database
- Apache CouchDB — Document database with HTTP API and sync
- CockroachDB — Distributed SQL database
- Couchbase Server — Distributed document, key-value and search database
- Firebase — Google's backend-as-a-service — Firestore, Auth, Storage, Functions (cloud)
- Microsoft SQL Server — Microsoft relational database for Windows and Linux
- MongoDB — Document (NoSQL) database
- Neo4j — Graph database with Cypher query language
- Neon — Serverless Postgres — autoscaling & branching (cloud API)
- Oracle Database — Enterprise relational database
- Patroni — HA PostgreSQL — automatic failover
- PlanetScale — Hosted MySQL with database branching (cloud API)
- PocketBase — Open-source backend in one file (SQLite + auth + realtime)
- PostgREST — Instant REST API over a PostgreSQL database
- ScyllaDB — High-performance Cassandra-compatible database
- Turso — Distributed SQLite (libSQL) at the edge (cloud API)
Agent Platforms15
- Browserbase — Headless browser infrastructure for AI agents (cloud sessions)
- ChatGPT Work Sites — OpenAI's work agent + published Sites/web apps (alpha)
- Devin — Cognition's AI software engineer (cloud API)
- E2B — Secure cloud sandboxes for AI agents (spin up / clone / kill)
- GitHub Copilot — AI pair programmer — live org seat/usage monitor
- Herdr — Agent multiplexer — tmux for AI coding agents (terminal)
- Hyperagent — Airtable's fleet-of-agents platform (cloud)
- Lakebed — Agent-native runtime for full-stack TypeScript capsules (alpha)
- Lumina — Local-first desktop AI agent with persistent memory & voice
- Manus — Autonomous general AI agent (cloud API)
- OpenAI Operator — OpenAI's browser-using agent (cloud)
- OpenHands (OpenDevin) — Open-source autonomous AI software engineer (self-hosted)
- Sandcastle — Orchestrate sandboxed coding agents (isolated containers)
- SuperAGI — Open-source autonomous-agent framework (self-hosted)
- T3 Code — Open-source control surface for the coding agents on your Mac
Runtimes15
- .NET (ASP.NET Core) — Run ASP.NET Core & Blazor apps (Kestrel)
- Astro — Content-first framework — zero JS by default; popular headless-WordPress front end
- code-server — VS Code in the browser (self-hosted)
- Fermyon Spin — Serverless WebAssembly apps & HTTP microservices
- Mojo — Modular's AI systems language (Python-family syntax)
- Next.js — React framework served by a Node process (SSR/SSG)
- Nuxt (Vue) — Vue framework served by a Node process (SSR/SSG)
- Solid (SolidJS) — Fine-grained reactive UI framework — SolidStart serves it as a Node process
- Supabase Edge Functions — Deno serverless functions on the edge
- SvelteKit (Svelte) — Svelte app framework served by a Node process (SSR/SSG)
- Tailwind CSS — Utility-first CSS — standalone build CLI (no Node)
- wasmCloud — Distributed WebAssembly platform (CNCF)
- WasmEdge — Lightweight WASM runtime for cloud/edge & AI (CNCF)
- Wasmer — WebAssembly runtime with a package registry
- Wasmtime — Bytecode Alliance WebAssembly runtime (WASI)
Speech AI12
- Coqui TTS — Open-source text-to-speech & voice cloning
- Faster Whisper — CTranslate2 Whisper — up to 4× faster
- Handy — Free local push-to-talk speech-to-text (Whisper)
- Kokoro TTS — Small, fast, high-quality open-weight TTS (82M)
- Meetily — Local AI meeting notetaker (desktop app)
- Piper TTS — Fast, local neural text-to-speech
- Scriberr — Self-hosted audio transcription app (WhisperX) with speaker diarization
- VibeVoice — Microsoft's long-form, multi-speaker TTS
- Voicebox — Open-source local voice-to-text for macOS
- Whisper — OpenAI's speech-to-text model (Python)
- Whisper.cpp — Fast C/C++ Whisper inference (CPU/Metal)
- Wispr Flow — AI voice dictation that types into any app
ML Workbench11
- ClearML — Experiment tracking, orchestration & MLOps (self-hostable)
- DVC Studio — Data/model version control + experiment dashboard
- Kaggle — Datasets, notebooks & competitions (CLI)
- Kubeflow — ML toolkit & pipelines on Kubernetes
- MLflow — ML experiment tracking & model registry (self-hosted)
- pandas — The standard Python DataFrame library
- Polars — Fast multicore DataFrame library (Rust core)
- PyTorch — Deep-learning framework (GPU/MPS accelerated)
- scikit-learn — Classic machine-learning library for Python
- Weights & Biases (Local) — Self-hosted experiment tracking & dashboards
- XGBoost — Gradient-boosted decision trees (high-accuracy tabular ML)
AI Clusters9
- Apache Spark — Distributed big-data processing engine
- Dask — Parallel computing for Python (scales pandas/NumPy)
- Distributed Llama — Tensor-parallel Llama across cheap nodes (root + workers)
- JupyterLab — Interactive notebooks for data & AI (Python)
- NVIDIA Base Command Manager — GPU/HPC cluster provisioning & management (ex-Bright)
- Petals — BitTorrent-style distributed inference of big models
- Ray — Distributed compute for AI (training/serving/tuning)
- Run:ai — GPU orchestration & fractional GPUs on Kubernetes (NVIDIA)
- Slurm — HPC workload manager / job scheduler
Data Pipelines & ETL8
- Airbyte — Open-source ELT with 300+ connectors (self-hosted / cloud)
- Dagster — Asset-oriented data orchestrator (self-hosted / cloud)
- dbt — SQL-based data transformation & modeling (self-hosted / cloud)
- Kestra — Event-driven orchestration & scheduling, YAML flows (self-hosted)
- Meltano — Open-source ELT built on Singer taps & targets (CLI)
- NATS JetStream Pipelines — Persistent streams & consumers for event pipelines (self-hosted)
- Prefect — Modern Python workflow orchestration (self-hosted / cloud)
- Redpanda Connect (Benthos) — Declarative stream-processing & connectors (self-hosted CLI)
Image & Video AI7
- AUTOMATIC1111 WebUI — Stable Diffusion web UI (txt2img/img2img)
- AutoShorts — Local video → 9:16 short-clip finder (desktop)
- ComfyUI — Node-graph Stable Diffusion / video workflows
- Fooocus — Simplest Stable Diffusion — type a prompt, get art
- InvokeAI — Pro Stable Diffusion studio (Unified Canvas)
- Opus Clip — AI viral-clip generator (cloud SaaS + API)
- Tesseract OCR — Open-source OCR engine — images/PDF scans to text (CLI)
Vector Databases6
- Chroma — Developer-friendly embedding database (self-hosted)
- FAISS — Fast in-process vector similarity search (library)
- Milvus — Scalable, distributed vector database (self-hosted)
- pgvector — Vector search inside PostgreSQL (extension)
- Qdrant — Fast vector DB for embeddings & RAG (self-hosted)
- Weaviate — Vector DB with built-in vectorizers & hybrid search (self-hosted)
GPU Infrastructure2
- DCGM Exporter — Export NVIDIA GPU metrics to Prometheus
- NVIDIA DCGM — NVIDIA Data Center GPU Manager — telemetry & health
すべてをひとつの Mac アプリで。
FrontierStack は、ローカルでもフリート全体でも、スタック全体のインストール・監視・保護をひとつのネイティブ macOS アプリで行います。
FrontierStack をダウンロードApple 公証済み · 安全・安心 · macOS 13 Ventura 以降