System Reference

Hermes Agent — Local Config {#hermes}

⚙️ Config (`~/.hermes/config.yaml`) {#config}

Setting	Value
Default model	`nvidia/nemotron-3-ultra-550b-a55b:free`
Fallback model	`openrouter/owl-alpha`
Fallback chain	`openrouter/owl-alpha` → `openrouter/free`
Vision/media	`nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free`
Provider	OpenRouter
Goal turns	77
Agent max turns	60
Compression	enabled, threshold 0.33
Context engine	compressor
Gateway timeout	1800s

⌨️ CLI Essentials {#cli}

hermes config set <key> <val>  # Set config value
hermes config show             # Full config
hermes skills list             # List skills
hermes skills reload           # Reload after edits
hermes sessions list           # Active sessions
hermes --resume <id>           # Resume session
hermes gateway restart         # Restart gateway
hermes logs --follow           # Live log tail

📂 Key Paths {#paths}

~/.hermes/config.yaml          # Main config
~/.hermes/.env                 # API keys (NOT inherited by crons)
~/.hermes/skills/              # Installed skills
~/.hermes/cron/                # Cron job output
~/.hermes/logs/                # agent.log, errors.log, gateway.log
~/.hermes/dreams/              # Dream Swarm reports (dated)
~/.hermes/plugins/             # Hermes plugins (OMP Suite)
~/.hermes/kanban.db            # OMP Kanban SQLite
~/.hermes/state.db             # Session/usage stats SQLite

Hermes Agent — Full Documentation Condensed {#hermes-docs}

Condensed from hermes-agent.nousresearch.com/docs/llms.txt — the complete Hermes Agent documentation index. Covers all 10 doc categories, ~100 pages.

🚀 1. Getting Started

Topic	What
Installation	`curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh \| bash` — Linux, macOS, WSL2, Termux
Quickstart	Install → first conversation in <5 min
Learning Path	Choose path by experience level and goals
Updating	Update or uninstall to latest version
Termux (Android)	Run Hermes directly on Android phone
Nix Setup	`nix run` quick start → fully declarative NixOS module with container mode

Repo: github.com/NousResearch/hermes-agent

🖥️ 2. Using Hermes

Topic	What
CLI	Terminal interface — commands, keybindings, personalities
TUI	Ink terminal UI — mouse-friendly, rich overlays, non-blocking input
Configuration	`config.yaml`, providers, models, API keys
Configuring Models	Model setup and switching
Sessions	Persistence, resume, search, management, per-platform tracking
Profiles	Isolated Hermes instances with own config/skills/cron
Git Worktrees	Run multiple Hermes agents safely on same repo
Docker Backend	Run in Docker or use Docker as terminal backend
Security	Dangerous command approval, user authorization, container isolation
Checkpoints & Rollback	Shadow git repos + automatic snapshots for destructive ops

⚡ 3. Core Features

Topic	What
Tools	Built-in tools overview, toolsets, terminal backends
Skills System	On-demand knowledge docs — progressive disclosure, agent-managed, Skills Hub
Curator	Background maintenance for agent-created skills — usage tracking, staleness, archival
Memory	`MEMORY.md`, `USER.md`, session search — remembers across sessions
Memory Providers	External plugins: Honcho, OpenViking, Mem0, Hindsight, Holographic, RetainDB, ByteRover, Supermemory
Context Files	`.hermes.md`, `AGENTS.md`, `CLAUDE.md`, global `SOUL.md`, `.cursorrules` — auto-injected
Context References	Inline `@-syntax` for attaching files, folders, git diffs, URLs into messages
Personality & SOUL.md	Customize voice — global SOUL.md, built-in personalities, custom personas
Plugins	Extend with custom tools, hooks, integrations
Built-in Plugins	Shipped plugins via lifecycle hooks (disk-cleanup, etc.)

🤖 4. Automation

Topic	What
Cron Jobs	Schedule tasks with natural language, one cron tool, attach skills
Delegation	`delegate_task` — spawn isolated child agents for parallel workstreams
Kanban Multi-Agent	Durable SQLite-backed task board for coordinating multiple profiles
Kanban Tutorial	Step-by-step kanban setup
Persistent Goals	Standing goal across turns until done (Ralph loop)
Code Execution	Python execution with RPC tool access — collapse multi-step workflows
Hooks	Custom code at lifecycle points — log, alert, webhook
Batch Processing	Generate agent trajectories at scale — parallel, checkpointing, toolset distribution

🎨 5. Media & Web

Topic	What
Voice Mode	Real-time voice — CLI, Telegram, Discord (DMs, text + voice channels)
Browser	Full browser automation — navigate, click, snapshot, vision
Vision	Image analysis with auxiliary vision model
Image Generation	Generate images via integrated providers
Text-to-Speech	TTS + voice message transcription across all platforms

💬 6. Messaging Platforms (17)

Platform	Setup
Telegram	Bot via BotFather
Discord	Bot via Discord Developer Portal
Slack	Bot via Socket Mode
WhatsApp	Bot via built-in Baileys bridge
Signal	Bot via signal-cli daemon
Email	IMAP/SMTP assistant
SMS	Chatbot via Twilio
Matrix	Bot via Matrix SDK
Mattermost	Bot integration
Home Assistant	Smart home integration
Webhooks	GitHub, GitLab triggers
DingTalk	Enterprise messaging
Yuanbao	Chat platform
Microsoft Teams	Via plugin
LINE	Messaging platform
IRC	Via plugin
API Server	OpenAI-compatible frontend

🔌 7. Integrations

Topic	What
Providers	Nous Portal, OpenRouter, OpenAI, Anthropic, Google, any OpenAI-compatible
MCP	Model Context Protocol — connect external tool servers, filter which tools Hermes loads
ACP	Agent Context Protocol — use Hermes inside VS Code, Zed, JetBrains
API Server	Expose as OpenAI-compatible API for any frontend
Honcho Memory	AI-native persistent memory — dialectic reasoning, multi-agent user modeling
Provider Routing	Route requests across providers
Fallback Providers	Automatic failover chain
Credential Pools	Rotate across multiple API keys

📖 8. Guides & Tutorials

Guide	What
Tips & Best Practices	Prompt tips, CLI shortcuts, context files, memory, cost optimization, security
Local LLMs on Mac	llama.cpp / MLX on Apple Silicon — model selection, memory optimization, benchmarks
Daily Briefing Bot	Research → summarize → deliver to Telegram/Discord every morning
Team Telegram Assistant	Bot for whole team — code help, research, sysadmin
Python Library	Embed AIAgent in your own scripts/apps — no CLI required
Use MCP with Hermes	Connect MCP servers, filter tools, safe real workflows
Use Voice Mode	Setup voice across CLI, Telegram, Discord voice channels
Use SOUL.md	Shape default voice — what belongs there vs AGENTS.md vs /personality
Build a Plugin	Complete plugin with tools, hooks, data files, skills
Automate with Cron	Real-world patterns — monitoring, reports, pipelines, multi-skill
Work with Skills	Find, install, use, create skills
Delegation Patterns	When/how to use subagents — parallel research, code review, multi-file
GitHub PR Review Agent	Auto code reviewer — monitor repos, review PRs, deliver feedback

🛠️ 9. Developer Guide

Topic	What
Contributing	Dev setup, code style, PR process
Architecture	Major subsystems, execution paths, data flow
Agent Loop	AIAgent execution, API modes, tools, callbacks, fallback
Prompt Assembly	System prompt building, cache stability, ephemeral layers
Context Compression & Caching	Compression strategies, prompt caching
Gateway Internals	Boot, authorization, session routing, message delivery
Session Storage	SQLite session store with FTS5 search
Provider Runtime	Provider resolution, credentials, API modes, auxiliary models
Adding Tools	Schemas, handlers, registration, toolsets
Adding Providers	Auth, runtime resolution, CLI flows, adapters, tests, docs
Adding Platform Adapters	New messaging platform integration
Creating Skills	SKILL.md format, guidelines, publishing
Extending the CLI	Wrapper CLIs with custom widgets, keybindings, layout

📚 10. Reference

Reference	What
CLI Commands	Authoritative terminal command reference
Slash Commands	Complete CLI + messaging slash command reference
Profile Commands	Profile management commands
Environment Variables	All env vars used by Hermes
Tools Reference	All built-in tools grouped by toolset
Toolsets Reference	Core, composite, platform, dynamic toolsets
MCP Config Reference	MCP config keys, filtering semantics, utility-tool policy
Model Catalog	Available models across providers
Bundled Skills Catalog	~90 skills shipped with Hermes
Optional Skills Catalog	~60 additional installable skills
FAQ & Troubleshooting	Common issues and solutions

Cron Jobs {#crons}

Job	Schedule	Skills	Toolsets	Purpose
`llm-wiki-evolve`	Daily 03:00	`hugo-llm-wiki`	terminal, file	Research → build → deploy → verify live site
`Nightly Dream Swarm`	Daily 04:44	`nightly-sentinel`	terminal, file, delegation	Ecosystem diagnostic → dream.md + Discord report

Dream Swarm Naming

Reports saved as ~/.hermes/dreams/YYYY-MM-DD/YYYY-MM-DD_{vivid-title}.md — poetic 2-3 word title derived from dominant signal of the day.

Examples: vivid-kanban-dream, marathon-embed-session, silent-cost-drift

Cron Env

Critical: Cron jobs spawn isolated shells — env vars from ~/.hermes/.env are NOT inherited. Always source first:

set -a; source ~/.hermes/.env; set +a

Local Projects {#projects}

Active / Primary {#primary}

Project	Path	Stack	Purpose
LLM-Wiki	`~/hugo-llm-wiki/`	Hugo + vanilla JS	Self-discovering docs site (this site)
Wiki	`~/wiki/`	Markdown	Personal knowledge base
macos-web	`~/macos-web/`	Svelte 5 + Vite	macOS dashboard app
Lyrix	`~/lyrix-bench/`	Svelte 5 + Vite	Lyrix benchmark app
LRX	`~/lrx/`	Svelte + Tailwind	Web app
Daisy	`~/daisyskull/`	Svelte + Vite	DaisyUI project
Skeleton	`~/skeleton/`	Monorepo (pnpm)	Skeleton UI framework
OMP Suite	`~/.hermes/plugins/omp-suite/`	Python	Routa kanban + wterm terminal plugin

Dev Experiments {#dev}

Project	Path
eleven3, flow2batch, flow2draw	`~/dev/`
generative-ai-docs, terracraft	`~/dev/`
todprod, udytor, vibesdk	`~/dev/`
vinki, z_uditor	`~/dev/`

Dashboards & CMS {#dashboards}

Project	Path	Stack	Purpose
Svelte Flexiboards	`~/svelte-flexiboards-dashboard/`	Svelte 5	FlexiBoard dashboard prototype
hugo-sveltia-lab	`~/hugo-sveltia-lab/`	Hugo + Sveltia	Hugo + Sveltia CMS experiment
sveltia-cms	`~/sveltia-cms/`	Svelte monorepo	Sveltia CMS core
svelteflow	`~/svelteflow/`	Svelte + XYFlow	Flow editor
hugging_chat-ui	`~/hugging_chat-ui/`	Svelte + Tailwind	HuggingFace chat UI
deep-chat	`~/deep-chat/`	Web component	Chat UI library
merms	`~/merms/`	—	Mermaid ecosystem tools
Dicebear	`~/dicebeardev/`	—	Avatar library dev

Nginx-Served Domains {#domains}

Domain	Path	Purpose
`llm-wiki.0rk.de`	`~/hugo-llm-wiki/public/`	LLM Wiki (this site)
`mixbox-review.0rk.de`	—	Mixbox review
`n8n.0rk.de`	—	n8n automation
`opendaw-studio.0rk.de`	—	OpenDAW studio
`opendaw-yjs.0rk.de`	—	OpenDAW YJS sync

Archived {#archived}

Project	Path	Notes
daistory	`~/daistory/`	DaisyUI monorepo
scrapez	`~/scrapez/`	Python scraping
GURU	`~/GURU/`	Reference material
ASSETS	`~/ASSETS/`	Static assets
.antigravity-server	`~/.antigravity-server/`	Server project
mj-tools	`~/mj-tools/`	sref-panel tool
docs	`~/docs/`	General docs

Plugins {#plugins}

OMP Suite (`~/.hermes/plugins/omp-suite/`) {#omp}

Property	Value
LOC	1,626
Tables	5 SQLite
Endpoints	17
Hooks	7
Git	3 commits, no remote (private)

6 Kanban specialist skills (~/.hermes/skills/routas-rs/):

Skill	Role
`routas-backlog-refiner`	Clarify scope, write cards
`routas-blocked-resolver`	Classify blockers
`routas-dev-crafter`	Implement scoped changes
`routas-done-reporter`	Terminal lane, produce completion report
`routas-review-guard`	Independently verify
`routas-todo-orchestrator`	Re-validate backlog

Dashboard: port 9119. Kanban board: omp-suite-swarm.

Skills {#skills}

Skill Categories (30 dirs, 7 with SKILL.md) {#active-skills}

Category	Active Skill	Purpose
`hermes-essentials`	✅ SKILL.md	Core config, debugging, planning, TDD
`superpower-pipeline`	✅ SKILL.md	Research + web search pipeline
`superpower-research`	✅ SKILL.md	Deep research automation
`omp-suite`	✅ SKILL.md	OMP plugin agent
`hugo-llm-wiki`	✅ SKILL.md	Hugo docs site patterns (loaded by cron)
`thermo-nuclear-code-quality-review`	✅ SKILL.md	Strict maintainability review
`transcript-extract`	✅ SKILL.md	Extract transcripts from URLs
`discord-interview-dashboard`	✅ SKILL.md	Discord bot dashboard
`frontend/`	svelte-flexiboards-dashboard	Svelte 5 dashboard patterns
`devops/`	discord-send-embed, nightly-sentinel, server-disk-ops, puter, hermes-dashboard-plugins	DevOps tools
`routas-rs/`	6 specialist skills (see OMP above)	Kanban workflow
`mlops/`	evaluation, inference, models, research, training, vector-databases (subdirs)	ML ops reference
23 other dirs	DESCRIPTION.md only	Reference stubs — no active workflow

Skill Stats

30 skill directories total
7 with full SKILL.md (executable workflows)
~100 total files across all skill dirs
Pruned: svelte-dashboard-integration (replaced by vanilla dashboard)

Tools & Stack {#tools}

Core Infrastructure

Tool	Version	Purpose
Hugo	0.140+	Static site generator
Nginx	—	Reverse proxy + SSL
Let’s Encrypt (Certbot)	—	SSL certificates
Python 3	—	Research automation, scripts
Node.js	22	Svelte dev (not in deploy pipeline)
npm	10	Package management
OpenRouter	—	LLM API provider
DuckDuckGo	—	Research search (API often blocked)

Svelte Ecosystem

Framework	Usage
Svelte 5	macOS dashboard, Lyrix, Flexiboards
DaisyUI	Component framework
Skeleton UI	Component library
XYFlow	Flow editor (svelteflow)
FlexiBoard	Drag-and-drop dashboard panels

Python Ecosystem

Domain	Tools
Scraping	Scrapy (scrapez)
ML/AI	Generative AI docs, vibesdk
Data	flow2batch, flow2draw

Infrastructure {#infra}

Hugo Site {#hugo}

hugo-llm-wiki/
├── config.toml              # Site config, baseURL, outputs
├── content/
│   ├── _index.md            # Homepage
│   ├── llm-wiki-reference.md  # This page
│   ├── topics/              # AI/Hugo/Svelte research topics (6 pages)
│   ├── research/            # Research automation docs
│   └── evolving/            # Continuously updated changelog
├── themes/lotusdocs/
│   ├── layouts/
│   │   ├── _default/
│   │   │   ├── baseof.html      # HTML shell + sidebar + nav
│   │   │   ├── single.html      # Article layout
│   │   │   ├── list.html        # Section index
│   │   │   └── list.json.json   # JSON feed per section
│   │   ├── index.html           # Homepage layout
│   │   └── topic/list.json.json # Topic JSON
│   └── static/css/main.css      # Dark theme
├── scripts/
│   ├── deploy.sh            # hugo --minify → nginx reload
│   └── research-automation.py # DuckDuckGo → Hugo markdown
└── public/                  # Built output → Nginx

Deploy Pipeline {#deploy}

# One command
cd ~/hugo-llm-wiki && hugo --minify && sudo nginx -t && sudo systemctl reload nginx
 
# Script
bash ~/hugo-llm-wiki/scripts/deploy.sh

Nginx {#nginx}

server {
    listen 443 ssl http2;
    server_name llm-wiki.0rk.de;
    root /home/dv/hugo-llm-wiki/public;
    index index.html;
    ssl_certificate /etc/letsencrypt/live/llm-wiki.0rk.de/fullchain.pem;
    ssl_certificate_key /etc/letsencrypt/live/llm-wiki.0rk.de/privkey.pem;
    location / {
        add_header Access-Control-Allow-Origin * always;
        try_files $uri $uri/ =404;
    }
    location /dashboard {
        try_files $uri $uri/ /dashboard/index.html;
    }
    location ~* \.(css|js|png|jpg|jpeg|gif|ico|svg|woff|woff2|ttf|eot)$ {
        expires 7d;
        add_header Cache-Control "public, immutable";
    }
}

Topic Knowledge Base {#topics}

Managed in ~/hugo-llm-wiki/content/topics/. Fed by JSON API at /topics/index.json.

Topic	File	Status
Hugo documentation automation	`hugo-documentation-automation-patterns.md`	✅ Complete
LLM-powered knowledge bases	`llm-powered-knowledge-bases.md`	✅ Complete
Self-discovering doc systems	`self-discovering-documentation-systems.md`	✅ Complete
AI content evolution	`ai-content-evolution-strategies.md`	✅ Complete
Svelte 5 best practices	`svelte-5-best-practices.md`	✅ Complete
Svelte 5 migration guide	`svelte-5-migration-guide.md`	✅ Complete

Data flow: research-automation.py → content/topics/*.md → hugo build → /topics/index.json → dashboard SPA

Quick Ref {#quick}

Task	Command
Hermes model switch	`hermes config set model.default "prov/model:free"`
New Hugo content	`hugo new content/topics/my-topic.md`
Hugo dev server	`hugo server -D`
Hugo build	`hugo --minify`
Deploy	`bash ~/hugo-llm-wiki/scripts/deploy.sh`
Check JSON feed	`curl -s llm-wiki.0rk.de/topics/index.json \| python3 -m json.tool`
Run research	`python3 ~/hugo-llm-wiki/scripts/research-automation.py`
View Hermes logs	`tail -f ~/.hermes/logs/agent.log`
Nginx test+reload	`sudo nginx -t && sudo systemctl reload nginx`
List cron jobs	`hermes cron list`
Source env (cron)	`set -a; source ~/.hermes/.env; set +a`

🧭 Explore

Graph View