diff --git a/README-ja.md b/README-ja.md index bfc4607..449d9cf 100644 --- a/README-ja.md +++ b/README-ja.md @@ -1,5 +1,9 @@ # Hermes 最適化ガイド（日本語ショート版） +

+ Hermes Optimization Guide +

+ > [英語版はこちら](./README.md) · このページは入口の要約。本文の章は英語のまま。 · 最終同期：2026-07-03 [NousResearch/hermes-agent](https://github.com/NousResearch/hermes-agent)（**v0.18.0 “The Judgment Release”（v2026.7.1）** まで反映。Mixture-of-Agents のファーストクラスモデル化、エビデンスに基づくタスク検証、ネイティブデスクトップアプリ、NVIDIA ローカルハードウェアを含む）向けの実戦ガイド + インストール可能な成果物（Skills・設定テンプレ・インフラスクリプト）。 diff --git a/README-zh.md b/README-zh.md index ff622c9..b00a90c 100644 --- a/README-zh.md +++ b/README-zh.md @@ -1,5 +1,9 @@ # Hermes 优化指南（中文简版） +

+ Hermes Optimization Guide +

+ > [English 完整版](./README.md) · 本页是入口摘要，章节正文仍为英文。 · 最后同步：2026-07-03 实用指南 + 可安装制品（Skills、配置模板、基础设施脚本），针对 [NousResearch/hermes-agent](https://github.com/NousResearch/hermes-agent)（当前覆盖到 **v0.18.0 “The Judgment Release”（v2026.7.1）**，含 Mixture-of-Agents 一等模型、基于证据的任务验证、原生桌面应用与 NVIDIA 本地硬件）。 diff --git a/README.md b/README.md index 67acd60..9109d6e 100644 --- a/README.md +++ b/README.md @@ -22,6 +22,10 @@ Every part you need to go from fresh install to a production Hermes deployment Unlike most guides, the prescriptions come with **working files**: [`skills/`](./skills) you can `ln -s` into `~/.hermes/skills/`, [`templates/config/`](./templates/config) you `cp` to `~/.hermes/config.yaml`, [`scripts/vps-bootstrap.sh`](./scripts/vps-bootstrap.sh) that takes a fresh VPS to production in one command. +

+ Docs plus runnable artifacts — 27 guide parts, 13 installable skills, 5 config templates, 4 reference architectures, one-command VPS bootstrap, 8-question config wizard +

+ *By Terp — [Terp AI Labs](https://x.com/OnlyTerp)* · Last updated **July 1, 2026** · [CHANGELOG](./CHANGELOG.md) · [ROADMAP](./ROADMAP.md) · [ECOSYSTEM](./ECOSYSTEM.md) --- @@ -68,7 +72,7 @@ Prefer a 5-minute local-only setup? → **[docs/quickstart.md](./docs/quickstart | [`templates/cron/`](./templates/cron) | Recommended production cron schedule. | | [`scripts/vps-bootstrap.sh`](./scripts/vps-bootstrap.sh) | One-command fresh VPS → production Hermes. | | [`diagrams/`](./diagrams) | 6 Mermaid diagrams (architecture, MCP flow, delegation, sandbox sync, observability, security layers). | -| [`assets/`](./assets) | Banner graphics used across the guide. | +| [`assets/`](./assets) | Banner art + the SVG infographics used across the guide (architecture, paths, timeline). | | [`benchmarks/`](./benchmarks) | Reproducible cost + latency table across 13 models × 5 tasks. | | [`docs/wizard/`](./docs/wizard) | **Interactive config wizard** — 8 questions → ready-to-drop `config.yaml`. Runs in your browser. | | [`docs/reference-architectures/`](./docs/reference-architectures) | **4 blueprints** — Homelab, Solo Dev, Small Agency, Road Warrior. Full parts list + cost + install. | @@ -82,6 +86,12 @@ Prefer a 5-minute local-only setup? → **[docs/quickstart.md](./docs/quickstart ## Architecture at a glance +

+ Hermes architecture — surfaces (desktop, CLI/TUI, web, 25+ chat platforms, cron) flow into the gateway (model router, approval layer, context engine, scale-to-zero), which fans out to any model, tools, memory, and observability +

+ +Prefer Mermaid? The same picture, editable: + ```mermaid flowchart LR subgraph Surfaces[Surfaces — one agent, many front ends] @@ -106,6 +116,10 @@ Full set of diagrams: [`diagrams/architecture.md`](./diagrams/architecture.md). ## Pick Your Path +

+ Pick your path — ten curated reading paths through the guide's 27 parts, from a 10-minute setup to production hardening, local GPU, and MoA verification +

+ This guide grew to 27 parts because *Hermes grew*. Every part lives in its own file (`part1-setup.md` … `part26-moa-verification.md`); this README keeps a short summary of Parts 1–5 (plus the full SOUL.md personality section) and links out. You don't have to read them all — pick the shortest path to what you need: ### 🎯 "I just want it working in 10 minutes" @@ -142,6 +156,10 @@ Skip the terminal: install the [desktop app](./part24-desktop-app.md) and let fi ## What's New (July 2026) +

+ Release timeline — v0.13 Tenacity, v0.14 Foundation, v0.15 Velocity, v0.16 Surface, v0.17 Reach, and the current v0.18 Judgment release +

+ Two huge releases landed since the Surface refresh — **v0.17.0 "Reach" (v2026.6.19)** and **[v0.18.0 "The Judgment Release" (v2026.7.1)](https://github.com/NousResearch/hermes-agent/releases/tag/v2026.7.1)**. Combined: ~3,200 commits, ~1,800 merged PRs, 1,200+ issues closed, and — as of v0.18 — **every P0 and P1 issue in the entire Hermes repo resolved** (~700 highest-priority items cleared in twelve days, with a standing commitment to keep the count at zero). None of it is model-specific — bring whatever weights you want. ### v0.18.0 — "Judgment" (latest) @@ -260,20 +278,9 @@ After this guide: ## How the Pieces Fit Together -``` -You (any device) - ↓ -Hermes Agent (lean context, ~5KB injected per message) - ↓ -┌──────────────────────────────────────────┐ -│ Skills (loaded on demand, 0 cost idle) │ -│ Memory (compact, vector-searched) │ -│ LightRAG (entity graph, deep recall) │ -│ Telegram (mobile + group access) │ -└──────────────────────────────────────────┘ - ↓ -LLM Provider (Claude, GPT, local models) -``` +

+ How the pieces fit together — you on any device, through the Hermes Agent's lean context, into modular layers (skills, memory, LightRAG, chat platforms), out to any LLM provider +

**The key insight:** Everything is modular. Install what you need, skip what you don't. The agent adapts. diff --git a/assets/architecture.svg b/assets/architecture.svg new file mode 100644 index 0000000..59cd78c --- /dev/null +++ b/assets/architecture.svg @@ -0,0 +1,181 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + HERMES ARCHITECTURE + ONE AGENT · MANY SURFACES · ANY MODEL + + + + + + + SURFACES + + + + 🖥️ Desktop App + macOS · Windows · Linux + + + + ⌨️ CLI · TUI + hermes chat · Ink TUI + + + + 🌐 Web Admin Panel + hermes dashboard + + + + 💬 25+ Chat Platforms + Telegram · Discord · Slack · Teams + iMessage (Photon) · WhatsApp · LINE + + + + ⏰ Cron · Webhooks + automation blueprints · continuations + + + + + + + + + GATEWAY + + + Model Router + cost · context · capability + + Approval Layer + denylist · allowlist · quarantine + + Context Engine + compression · prompt caching + + Scale-to-Zero + drain coordination · managed scope + + + + + + + + ANY MODEL + + + Cloud APIs · OpenAI-compatible + + MoA virtual models (council + judge) + + Local: llama.cpp · LM Studio · Ollama + NVIDIA RTX · DGX Spark · bring any weights + + + + + + + + TOOLS + + + Native tools · Nous Tool Gateway + + MCP servers (stdio · HTTP) + + Subagents · background fan-out + + Coding agents · swarms · sandboxes + + + + + + + + MEMORY + + Vector · three-tier + LightRAG graph RAG + /learn · /journey + Curator pruning + + + + + + + OBSERVE + + Audit log + Langfuse · Helicone + OpenTelemetry + Cost routing · evals + + + + + + + + + + + + + + + + + + + + MODEL-AGNOSTIC BY DESIGN — THE GUIDE IS ABOUT THE HARNESS, NOT THE WEIGHTS + + + ⚡ + diff --git a/assets/how-it-fits.svg b/assets/how-it-fits.svg new file mode 100644 index 0000000..17d99e1 --- /dev/null +++ b/assets/how-it-fits.svg @@ -0,0 +1,90 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + HOW THE PIECES FIT TOGETHER + EVERYTHING IS MODULAR — INSTALL WHAT YOU NEED, SKIP WHAT YOU DON'T + + + + + 🧑‍💻 + You — any device + desktop · terminal · browser · phone + + + + + ⚡ + Hermes Agent + lean context — ~5KB injected per message + + + + + MODULAR LAYERS + + + 🧩 Skills + loaded on demand — zero token cost when idle + + 🧠 Memory + compact, vector-searched, three tiers + + 🕸️ LightRAG + entity graph — deep relationship recall + + 📱 Chat platforms + Telegram & 25+ more — mobile + group access + + + + + + 🌌 + Any LLM provider + cloud · MoA · local weights + + + + + + + + + + + + + + THE KEY INSIGHT: EVERYTHING IS MODULAR — THE AGENT ADAPTS + diff --git a/assets/pick-your-path.svg b/assets/pick-your-path.svg new file mode 100644 index 0000000..ff49834 --- /dev/null +++ b/assets/pick-your-path.svg @@ -0,0 +1,116 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + PICK YOUR PATH + 27 PARTS — READ ONLY WHAT YOU NEED + + + + + 🎯 + Working in 10 minutes + Desktop app + Quick Setup via Nous + Portal — or CLI + web dashboard. + PART 24 · PART 1 · PART 12 + + + + 📱 + A Telegram bot that's useful + Setup → Telegram → on-the-fly skills + → persistent memory. + PART 1 · PART 4 · PART 5 · PART 7 + + + + 🤖 + Coding agents from a phone + Claude Code / Codex / Gemini CLI via + Kanban lanes + remote sandboxes. + PART 18 · PART 23 · PART 21 + + + + 💼 + Run it in production + Security playbook, observability & + cost, backup/debug, durable Kanban. + PART 19 · PART 20 · PART 16 · PART 23 + + + + + 🧠 + Most capable agent possible + MCP servers, coding agents, LightRAG + graph memory, Fast Mode. + PART 17 · PART 18 · PART 3 · PART 14 + + + + 💰 + Cheapest agent that works + Model routing, the cost-routing + playbook, context compression. + PART 9 · PART 20 · PART 6 + + + + 🛡️ + Prompt-injection defense + Read this first if your agent touches + any untrusted input. + PART 19 — SECURITY PLAYBOOK + + + + 🔒 + All local, on your own GPU + RTX / DGX Spark, OpenShell isolation, + llama.cpp / LM Studio / Ollama. + PART 25 — NVIDIA & LOCAL HARDWARE + + + + + 🖥️ + An app, not a terminal + Streaming chat, Cmd+K palette, Projects, multi-terminal, memory graph. + PART 24 — HERMES DESKTOP APP + + + + 🧑‍⚖️ + An ensemble of frontier models — with proof + MoA councils as models, evidence-based /goal verification, /learn + /journey. + PART 26 — MoA, VERIFICATION & SELF-IMPROVEMENT + diff --git a/assets/release-timeline.svg b/assets/release-timeline.svg new file mode 100644 index 0000000..0e67c26 --- /dev/null +++ b/assets/release-timeline.svg @@ -0,0 +1,92 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + TRACKED RELEASE BY RELEASE + THE GUIDE IS REFRESHED FOR EVERY MAJOR HERMES RELEASE + + + + + + + + + v0.13 + Tenacity + durable Kanban · /goal + Checkpoints v2 + + + + v0.14 + Foundation + PyPI install · hermes proxy + native Windows + + + + v0.15 + Velocity + multi-agent swarms + promptware defense + + + + v0.16 + Surface + Desktop app · admin panel + remote backend + + + + v0.17 + Reach + iMessage via Photon + background subagents + + + + + v0.18 + Judgment — latest + MoA as a model · verification + /learn · scale-to-zero + + + + + + + CURRENT THROUGH v0.18.0 (v2026.7.1) — "THE JUDGMENT RELEASE" + diff --git a/assets/runnable-artifacts.svg b/assets/runnable-artifacts.svg new file mode 100644 index 0000000..24a93ec --- /dev/null +++ b/assets/runnable-artifacts.svg @@ -0,0 +1,79 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + DOCS + RUNNABLE ARTIFACTS + NOT JUST PROSE — WORKING FILES YOU CAN DROP IN TODAY + + + + + 27 + guide parts + setup → production, + one file each + + + + 13 + installable skills + ln -s into ~/.hermes/skills/ + and they're live + + + + 5 + config templates + minimum → production + → security-hardened + + + + 4 + reference architectures + Homelab · Solo Dev + Small Agency · Road Warrior + + + + 1 + command to a VPS + vps-bootstrap.sh: fresh box + → hardened production + + + + 8 + wizard questions + interactive browser wizard + → ready config.yaml + + + SKILLS · CONFIGS · COMPOSE STACKS · CADDY · SYSTEMD · CRON · BENCHMARKS — ALL IN THIS REPO +