Skip to content

Quick Start

Get NOUZ running in 5 minutes.

Installation

bash
pip install nouz-mcp

Setup

1. Create config.yaml (Optional)

Without config.yaml, the server starts in LUCA mode (pure graph). For PRIZMA/SLOI, create a local config:

bash
cp config.template.yaml config.yaml

On Windows PowerShell:

powershell
Copy-Item config.template.yaml config.yaml

If your MCP client starts the server outside the project directory, pass the path explicitly through NOUZ_CONFIG.

Minimal config:

yaml
mode: prizma  # luca | prizma | sloi

For semantic classification, add domains (etalons) — full examples → Configuration.

2. Point to Your Vault

bash
export OBSIDIAN_ROOT=/path/to/your/obsidian-vault

3. Connect an Embedding Provider (For PRIZMA/SLOI)

Any OpenAI-compatible API works: LM Studio, Ollama, or a cloud provider.

bash
export EMBED_API_URL=http://127.0.0.1:1234/v1
export EMBED_MODEL=nomic-embed-text

4. Connect to Your AI Client

Add to your MCP configuration (Claude Desktop, Cursor, OpenCode, etc.):

json
{
  "mcpServers": {
    "nouz": {
      "command": "python",
      "args": ["-m", "nouz_mcp"],
      "env": {
        "OBSIDIAN_ROOT": "/path/to/vault",
        "EMBED_API_URL": "http://127.0.0.1:1234/v1"
      }
    }
  }
}

Launch

bash
nouz-mcp
[INFO] Indexing database on startup...
[INFO] Indexed: 42 files, errors: 0
[INFO] Core etalons loaded from DB: ['S', 'D', 'E']
[INFO] NOUZ MCP Server v2.5.6 started

First Steps

Once NOUZ is connected, your AI assistant uses the tools directly:

python
# Note position in the graph
format_entity_compact("My Module.md")
# → (2E)[E]{E}

# Classify a new note
suggest_metadata("New Note.md")
# → {sign: "σE", artifact_sign: "σ", level: 4, bridges: [...]}

# Recalculate signs across the entire vault
recalc_signs()
recalc_core_mix()

Simple Start

Begin with LUCA mode — no embeddings required:

yaml
mode: luca

You get the full DAG: hierarchical connections, entity formulas, navigation. Add semantics later by switching to PRIZMA or SLOI.

Requirements

  • Python 3.10+
  • Obsidian vault (any directory with .md files)
  • No SQLite server required: Python includes sqlite3, and the package installs aiosqlite
  • LM Studio, Ollama, or OpenAI-compatible API (optional, for PRIZMA/SLOI)

{Semiotronika}
Telegram · Volnaya Sreda · Email