Hermes Dashboard

A standalone, read-only web dashboard for Hermes Agent. Browse your config, sessions, memory, skills, and cron jobs through a clean web UI — without touching the Hermes codebase.

What is this?

Hermes Agent stores all its state in ~/.hermes/ — config files, SQLite databases, markdown memory files, skill definitions, and cron job schedules. This dashboard reads those files and presents them in a navigable web interface.

Zero coupling to Hermes. This project doesn't import from or depend on the Hermes Agent codebase. It just reads the same files.

Status

🚧 In development — spec and UI design phase.

• Requirements document
• UI design (v0-generated Next.js prototype)
• Technical design document
• Implementation tasks
• FastAPI backend
• Frontend (vanilla HTML/CSS/JS SPA)

Planned Features

Section	What it shows
Overview	Model info, session count, memory usage bars, skill count, cron job count, data file health checks
Config	All config.yaml settings as styled cards with raw YAML toggle
Environment	API keys with redacted preview (first 4 chars + asterisks)
Sessions	Paginated session list with FTS5 search, click-through to full message thread
Memory	Agent memory and user profile side-by-side with usage progress bars
Soul	Agent persona/personality file rendered as markdown
Skills	Card grid with category filter, click to expand skill details
Cron Jobs	Scheduled tasks with state badges, schedule, delivery target

Architecture

hermes-dashboard/
├── app.py                  # FastAPI backend (reads ~/.hermes/ files)
├── static/                 # Vanilla HTML/CSS/JS frontend (no build step)
│   ├── index.html
│   ├── style.css
│   └── app.js
├── requirements.txt        # Python deps: fastapi, uvicorn, pyyaml
├── v0-generated/           # Next.js UI prototype (design reference only)
├── .kiro/specs/            # Spec documents (requirements, design, tasks)
└── README.md

Backend: FastAPI serving JSON API endpoints + static files. Reads all data from ~/.hermes/ in read-only mode. Binds to 127.0.0.1:8787 (localhost only).

Frontend: Vanilla HTML/CSS/JS SPA. No build step, no npm, no framework. Fetches from the JSON API and renders client-side. Dark/light theme with OS auto-detect + manual toggle.

Data Sources

All read from ~/.hermes/ (respects HERMES_HOME env var):

Source	File	Format
Config	config.yaml	YAML
API Keys	.env	Key=value (redacted on display)
Sessions & Messages	hermes_state.sqlite	SQLite (WAL, FTS5)
Agent Memory	MEMORY.md	Markdown (§-delimited entries)
User Profile	USER.md	Markdown (§-delimited entries)
Persona	SOUL.md	Markdown
Skills	skills/*/SKILL.md	YAML frontmatter + Markdown
Cron Jobs	cron/jobs.json	JSON

API Endpoints

Method	Path	Description
GET	/api/status	Health check, file existence, hermes home path
GET	/api/config	Parsed config.yaml contents
GET	/api/env	Redacted environment variables
GET	/api/sessions	Paginated session list (supports ?search= and ?limit=)
GET	/api/sessions/{id}/messages	Messages for a specific session
GET	/api/memory	Memory + user profile content with usage stats
GET	/api/soul	Soul/persona file content
GET	/api/skills	All skills with parsed frontmatter
GET	/api/cron	Cron job list

Design Principles

• Read-only — never writes to Hermes data files
• Standalone — no imports from or dependencies on the Hermes codebase
• Localhost only — binds to 127.0.0.1, rejects non-localhost origins
• No build step — frontend is vanilla HTML/CSS/JS, backend is a single Python file
• Secrets stay secret — API keys are redacted before leaving the backend, SQLite opened in read-only mode

License

MIT

Credits

Built as a companion tool for Hermes Agent by Nous Research.