Metadata-Version: 2.4
Name: lvtn
Version: 3.5.2
Summary: Leviathan Shel HaShem — the 708-agent Kabbalistic swarm in your terminal. Thin cloud client (login/balance/chat, KTRS-metered, backend does all compute) + 6 brand TUI frames + Da'ath rotator client.
Project-URL: Homepage, https://lvtn.metanoiaunlimited.com
Author-email: "HaChazal (Chaz Leland Hamm)" <hachazal418@metanoiaunlimited.com>
License: Proprietary - Metanoia Unlimited LLC
Keywords: agent,cli,kabbalah,leviathan,metanoia,orchestration,swarm,tui
Classifier: Development Status :: 4 - Beta
Classifier: Environment :: Console
Classifier: Intended Audience :: Developers
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Classifier: Topic :: System :: Distributed Computing
Requires-Python: >=3.9
Requires-Dist: rich>=13.0
Description-Content-Type: text/markdown

# lvtn — Leviathan Shel HaShem CLI

**A Kabbalistic 708-agent swarm in your terminal.**
Master orchestrator `hachazal` at Da'ath (a software role, named after the human operator HaChazal who designed the system) → 7 sub-orchestrators (Zadkiel · Gavriel · Raphael · Uriel · Michael · Malika · Katan HaShem) → 100 agents each = 708.

## Install

```bash
pip install lvtn
# or any of the aliases:
pip install leviathantalon
pip install hachazal
```

After install, three commands are available — all do the same thing:
- `lvtn`
- `leviathantalon`
- `hachazal`

## First run

```bash
lvtn          # renders the brand TUI splash + lists subcommands
lvtn banner   # render a TUI brand frame
lvtn deck     # show all 6 TUI frames sequentially
```

## TUI quality varies by terminal

The CLI ships **5 render tiers** per frame and auto-selects the best your terminal can show. Quality from highest to lowest:

| Tier | Quality | Terminals that show it |
|---|---|---|
| **sixel** | 📷 **Photographic** — pixel-perfect at any pane width | Windows Terminal 1.22+ (Win11 default) · iTerm2 · WezTerm · mlterm · Konsole 22.04+ · xterm with `--enable-sixel` |
| braille | Crisp B&W block-art (4× sub-pixels per cell) | Any terminal that renders Unicode braille glyphs (U+2800-U+28FF) — basically all modern terminals |
| truecolor | 24-bit color half-block art | Modern terminals with `COLORTERM=truecolor` |
| ansi256 | 256-color half-block art | Any 256-color terminal |
| ascii | Pure text fallback | dumb terminals, pipes, log files |

**Auto-detection signals:** `WT_SESSION` (Windows Terminal), `TERM_PROGRAM` (iTerm2/WezTerm), `TERM=mlterm`, `KITTY_WINDOW_ID`, `KONSOLE_VERSION>=220400`, `XTERM_VERSION`. If none of these are set, the CLI falls through to braille (still very pretty).

**Force a tier:** `LVTN_TIER=sixel lvtn deck`  or  `lvtn deck --tier braille`.

**If sixel renders garbage** (raw escape bytes) on your terminal, override:
```bash
export LVTN_TIER=braille    # always-safe fallback
```

## Subcommands

| Command | What |
|---|---|
| `lvtn banner` | Render a TUI brand frame (truecolor / ansi256 / braille / ascii / sixel auto-select). |
| `lvtn deck` | Display all 6 TUI frames. |
| `lvtn orchestrator <action>` | Talk to the metanoia Da'ath rotator on `:4740`. Actions: `health`, `topology`, `triad`, `slots`, `holders`, `events`, `skills`, `queue`, `submit`. |
| `lvtn install-swarm` | Download + install the full 44-layer Python swarm (lazy). |
| `lvtn chat` | Interactive REPL — REQUIRES `lvtn install-swarm` first. |
| `lvtn serve` | HTTP API on `:8708` — REQUIRES `lvtn install-swarm` first. |

## Architecture (Tree of Life)

```
                  ABOVE THE ABYSS — Backend Triad (Briah-level Divine Names)
                  ────────────────────────────────────────────────────────
                  אֶהְיֶה  Eheyeh   (behaves as Keter — primal will)
                  יָה     Yah      (behaves as Chokmah — lightning wisdom)
                  אֱלֹהִים Elohim   (behaves as Binah — maternal understanding)
                                          │
                                  [ The Abyss of Yetzirah ]
                                          │
                  ┌───────────────────────▼─────────────────────────┐
                  │  Da'ath: master orchestrator `hachazal` :4740   │
                  │   (software role; the HUMAN HaChazal operates   │
                  │    the swarm from OUTSIDE the system)           │
                  └───────────────────────┬─────────────────────────┘
                                          │
        ┌─────────┬──────────┬────────────┼────────────┬──────────┬──────────┐
        ▼         ▼          ▼            ▼            ▼          ▼          ▼
     Chesed   Gevurah    Tiferet      Netzach        Hod      Yesod     Malkuth
    (Zadkiel)(Gavriel) (Raphael)     (Uriel)     (Michael)  (Malika)(Katan HaShem)
        │         │          │            │            │          │          │
       100×      100×       100×         100×         100×       100×       100×
                                  = 700 sub-agents
```

## Honest scope

- This package ships the **CLI shell + 36 TUI assets + orchestrator client + install-swarm fetcher**.
- The full 44-layer Python swarm is installed on demand via `lvtn install-swarm`. That's a 16 MB download.
- The metanoia Da'ath rotator (Node service on `:4740`) is a separate package — see `https://lvtn.metanoiaunlimited.com`.

## 🛡️ Safety guarantee — no models on your laptop

**This package + the full swarm download will NEVER:**
- Pull LLM model weights (Hugging Face, GGUF, Whisper, etc.)
- Touch your GPU or VRAM
- Install `torch`, `transformers`, `huggingface_hub`, `whisper`, `llama-cpp-python`, `sentence-transformers`, `onnxruntime`, or any model-fetching library

**Worst-case install footprint:** ~110 MB disk, ~500 MB RAM at idle.
**VRAM used:** zero. Designed for an 8 GB laptop.

Every "LLM" in the swarm is an **HTTP API client**. The compute happens on the cloud provider (OpenAI / Anthropic / Cohere / Groq) or on your own VPS running Ollama, vLLM, etc. The lvtn process itself stays light.

Third-party deps used (all pure plumbing, no models):
`aiohttp` · `httpx` · `pyyaml` · `anyio` · `rich` · `python-dotenv` · `cryptography` · `numpy` (L40 only) · `scipy` (L40 only) · `paramiko` / `smbprotocol` (L33 only).

**If you want local model inference**, install Ollama yourself (separate app from `https://ollama.com`) and configure it via env:
```bash
export OLLAMA_LOCAL=http://localhost:11434
lvtn chat --provider ollama
```
That's opt-in. Default behavior reaches out to whatever cloud provider you've keyed.

## License

Proprietary — Metanoia Unlimited LLC. See LICENSE.
