Sphene: The Autonomous, Web-First Knowledge Engine for Humans & AI

Most modern personal knowledge tools are trapped inside heavy desktop-only applications. If you have ever tried accessing your personal knowledge base on your phone without paying recurring monthly subscription fees for cloud syncing, or tried letting an autonomous AI agent headlessly read, connect, and update your notes in real time, you know where the friction begins.

Heavy runtimes, mobile synchronization headaches, and closed desktop-first paradigms hold modern knowledge workflows back.

Sphene is built to solve this: an ultra-fast, flat-file personal knowledge engine with native web accessibility, bidirectional wikilinking, and an interactive Force-Directed Knowledge Graph—engineered from day one for seamless collaboration between you and your autonomous AI agents.

The Mental Model: Your Agent's Physical Filing Cabinet

The simplest mental model for pairing an autonomous AI agent (such as Hermes Agent) with Sphene is:

Hermes is the executive assistant doing the work; Sphene is its structured filing cabinet, memory shelf, and visual dashboard on your desk.

Sphene doesn't change how your AI agent reasons—it gives the agent a clean, plain-English place to read your context and write down what it did, while giving you an instant visual dashboard in your browser (on desktop, tablet, or phone) to review, edit, search, and navigate what the agent knows in real time.

Why Sphene Solves the Three Core Agent Memory Flaws

Without a structured knowledge engine, an AI agent's memory is a black box—buried in internal database tables or ephemeral session logs where you can't easily audit or steer it.

Sphene fixes three fundamental problems:

  1. A Human-Auditable Long-Term Memory Window: AI agents maintain internal scratchpads for short-term reasoning, but Sphene acts as the curated long-term shelf. You can open Sphene in any browser, see exactly what the agent recorded about your projects or preferences, edit a paragraph, or add your own notes. The next time the agent runs, it reads your updated version immediately.
  2. Persistent Context Without Token Waste: AI models have strict token limits and forgetting curves. Instead of cramming thousands of pages into a single prompt, your agent navigates directly to the specific notes it needs, following bidirectional links across your knowledge graph.
  3. 100% Data Ownership & Zero Vendor Lock-in: Your knowledge vault is stored as clean, standard Markdown (.md) files on your computer inside ./vault/. If you switch AI models, upgrade your agent, or change hardware, every single note, research document, and cross-link remains yours forever on your local drive.

How You Actually Use It: Everyday Flow & Prompts

From your perspective, you prompt your agent naturally, treating your Sphene vault as a shared desk space:

Scenario A: Reading Your Context & Requirements

What you prompt: "Read my note on [[Home Office Setup]] and give me a checklist of remaining parts I need to buy."

What happens: The agent inspects your Sphene vault, reads Home Office Setup, extracts your budget and constraints, and presents your tailored checklist.

Scenario B: Autonomous Research & Graph Synthesis

What you prompt: "Analyze the top 5 outdoor IP cameras with local RTSP support. Write a summary note in my Projects folder and link each camera to its own file."

What happens: The agent browses the web, synthesizes specs, and creates interconnected Markdown notes directly in Sphene, complete with [[Wikilinks]]. You refresh your browser, and the new notes and interactive knowledge graph clusters are already there waiting for you.

Scenario C: Daily Logging & Task Handoff

What you prompt: "Summarize what we solved today regarding database migrations and log it into my Daily Note."

What happens: The agent locates today's date in your daily notes and cleanly appends bullet points under an ## Agent Activity section.

[!TIP] Best Practice: The Scoped Sandbox When connecting an AI agent to your Sphene vault, consider giving the agent a dedicated folder (such as /Agent-Hub or /Projects). You can move project notes and documentation into that folder whenever you want the agent to collaborate on them, keeping your personal journals completely private.

Sphene Aegis: Sovereign, Zero-Knowledge Vault Encryption

Most legacy personal knowledge tools leave your local Markdown files completely unencrypted on disk, or tie encryption to proprietary, closed-source cloud synchronization services where server keys or escrow backdoors can be compromised.

Sphene Aegis introduces an open, sovereign cryptographic layer designed for complete operational zero-knowledge and total agent interoperability:

1. Global Vault vs. Scoped / Selective Encryption

  • Scoped Mode (Note-by-Note): Seal individual sensitive documents (credentials, infrastructure keys, confidential client notes) while keeping the remainder of your knowledge graph openly indexed and searchable. In the Sphene web UI, checking "🔒 Seal with Sphene Aegis" encrypts the payload in your browser's RAM before it ever hits the wire.
  • Global Vault Mode: Encrypt every .md file inside your vault. Notes are sealed in-place on disk with AES-256-CBC, rendering raw storage files completely unreadable to unauthorized parties or hostile host environments.

2. The Visual Vault Settings & Management Center (1-Click Simplicity)

For non-technical users who don't want to manage encryption through terminal commands or environment variables, Sphene provides a built-in graphical Vault Settings & Security Center:

  • Access Anywhere: Click "Vault Settings" in the sidebar footer or press Ctrl + , (or Cmd + ,) anywhere in the app.
  • Real-Time Vault Health: Instantly monitor how many notes are sealed with AES-256 vs. open in plaintext, and see your current vault mode (Global Seal, Partial, or Open).
  • Session Passphrase Cache: Enter your passphrase once in the settings panel. It is stored exclusively in temporary browser RAM—enabling seamless note unlocking without repeated password prompts, and automatically clearing when you close your browser.
  • 1-Click Global Encryption: Click "🔒 Seal Entire Vault" to encrypt all open notes in your repository in seconds. Click "🔓 Decrypt Entire Vault" to restore everything to open Markdown at any time.
  • Default Policy (Auto-Seal): Toggle "Default to Sealed" so every note you compose with Ctrl + N is automatically encrypted by default.
  • 1-Click Emergency Script Download: Download decrypt_vault.sh directly from the settings panel for offline recovery on any computer.

3. Hermes Agent Bidirectional Sealed Access

Autonomous agents running Hermes can both create and access encrypted notes natively.

Export your passphrase in your agent's environment:

export SPHENE_VAULT_PASSPHRASE="your-secure-vault-passphrase"

Hermes automatically leverages this key to seal and unseal notes on the fly:

# Autonomously create a sealed note with AES-256 encryption
python3 ~/.hermes/skills/sphene-knowledge-hub/sphene_client.py write "Classified Project" \
  --sealed \
  --body "Secret infrastructure access credentials and internal architecture notes."

# Read the note (automatically decrypted in memory)
python3 ~/.hermes/skills/sphene-knowledge-hub/sphene_client.py read "Classified Project"

# Seal or unseal notes directly on disk
python3 ~/.hermes/skills/sphene-knowledge-hub/sphene_client.py seal "Unencrypted Note"
python3 ~/.hermes/skills/sphene-knowledge-hub/sphene_client.py unseal "Classified Project"

When browsing in Sphene, sealed notes display an interactive Aegis Shield Card. Entering your passphrase decrypts the note instantly in browser memory via native WebCrypto (window.crypto.subtle) without refreshing the page or transmitting the secret.

4. Non-Destructive In-Place Decryption & Zero External Dependencies

Your data sovereignty is absolute. Sphene uses standard OpenSSL PBKDF2 key derivation (SHA-256, 100,000 iterations) with AES-256-CBC.

If Sphene is offline, shut down, or removed, your notes are never held hostage. You can decrypt your entire vault in place on any Linux, macOS, or Unix-compatible system with zero dependencies using standard system openssl:

# Decrypt all sealed .md files in-place across your entire vault
curl -fsSL https://cimpean.com/sphene/decrypt_vault.sh | bash -s -- ./vault "your-secure-vault-passphrase"

No proprietary unpackers. No vendor escrow. 100% sovereign data ownership.

What Makes Sphene Different?

  • Native Web-First (Zero Mobile Sync Friction): Access your complete second brain from any phone, tablet, or laptop. Expose it via Cloudflare Tunnels, Tailscale, WireGuard, or local LAN with zero recurring sync fees.
  • Interactive Force-Directed Knowledge Graph: Explore your knowledge topology with an interactive node-and-edge visualizer rendered entirely in your browser with zero external CDN dependencies.
  • Bidirectional [[Wikilinks]] & Live Backlinks: Connect ideas with [[Note Name]] or [[Target|Custom Label]]. Sphene automatically calculates incoming backlinks and displays them in a dedicated collapsible drawer for every document.
  • Ludicrously Fast & Light: Runs on an optimized flat-file engine with Nginx and PHP 8.3 FPM. It uses under 50MB of RAM, boots in seconds, and requires zero database administration.
  • Instant In-Browser Note Creation & Settings: Press Ctrl + N to compose a note, or Ctrl + , to open the Vault Settings and Security Center.

Instant One-Line Cross-Platform Installation

Sphene runs on Linux, macOS, and Windows (via WSL2) with Docker in one single command—no manual git configuration or build steps required:

curl -fsSL https://cimpean.com/sphene/install.sh | bash

The installer automatically:

  1. Verifies Docker & Docker Compose.
  2. Downloads and loads the pre-compiled, encrypted Sphene engine image directly into Docker.
  3. Initializes your local ./vault/ folder for your Markdown notes.
  4. Boots the container on port 8743 and verifies system health.

Once booted, open your browser and navigate to: 👉 http://localhost:8743 (or your server's LAN IP).

1-Click Seamless Migration from Desktop Markdown Vaults

Already have an existing personal knowledge base built with desktop Markdown apps like Obsidian, Logseq, or Foam? Migrating to Sphene takes one single step:

  • From the Web UI: Open "Vault Settings" (Ctrl + ,) -> "1-Click Vault Migration", and drag & drop a .zip archive of your vault. Sphene immediately extracts your nested folders, converts image transclusions, indexes all bidirectional [[Wikilinks]], and renders your interactive knowledge graph.
  • From the Command Line: 
    curl -fsSL https://cimpean.com/sphene/migrate_vault.sh | bash -s -- /path/to/my-vault

What Gets Preserved?

  • Folder Hierarchies: Multi-tier directory trees (including 00_System/, 10_SecondBrain_Karpathy/, 20_Human_Thinking/) are mapped with 100% fidelity.
  • Bidirectional Links: [[Note]] and [[Target|Alias]] links hook into the Sphene graph and backlinks engine automatically.
  • Task Checklists: Interactive - [ ] and - [x] checkboxes become live, clickable toggles that update your markdown files on disk.
  • Attachments & Canvas: Images (png, svg, jpg, pdf) and visual .canvas maps are automatically imported and served.

Sealed Runtime & Manual Setup

Sphene is distributed as a sealed, encrypted runtime container. All proprietary engine logic executes in memory, while your knowledge files stay 100% transparent and editable in Markdown on your host machine:

Manual Docker Setup

# 1. Create your workspace
mkdir sphene && cd sphene && mkdir vault

# 2. Load the encrypted engine into Docker
curl -fsSL https://cimpean.com/sphene/sphene-image.tar.gz | docker load

# 3. Create docker-compose.yml
cat << "EOF_COMPOSE" > docker-compose.yml
services:
  sphene:
    image: sphene:1.0.0
    container_name: sphene-engine
    restart: unless-stopped
    ports:
      - "8743:80"
    volumes:
      - ./vault:/var/www/html/user/pages/02.knowledge:rw
EOF_COMPOSE

# 4. Start Sphene
docker compose up -d

# 5. Open in your browser
open http://localhost:8743

Connecting Hermes Agent

To connect your Hermes Agent deployment to Sphene, extract the included skill into your agent's skills catalog:

mkdir -p ~/.hermes/skills
curl -fsSL https://cimpean.com/sphene/hermes-skill.tar.gz | tar -xz -C ~/.hermes/skills/

Hermes will now automatically detect your Sphene knowledge vault, allowing you to ask it to read, create, seal, and cross-link notes seamlessly during your conversations.