oomllama 0.1.0__tar.gz

oomllama-0.1.0/.gitignore

@@ -0,0 +1 @@
+target/

oomllama-0.1.0/Cargo.toml

@@ -0,0 +1,96 @@
+[package]
+name = "jis-router"
+version = "0.1.0"
+edition = "2021"
+authors = ["Jasper van de Meent <jasper@humotica.com>", "Root AI <root_ai@humotica.nl>"]
+description = "JIS Router - Intent-based routing for AETHER. The Borrow Checker for Identity."
+license = "MIT"
+repository = "https://github.com/symbaion/jis-router"
+keywords = ["jis", "aether", "intent", "routing", "tibet"]
+readme = "README.md"
+
+[dependencies]
+# Async runtime
+tokio = { version = "1", features = ["full"] }
+
+# Web framework - axum is fast and ergonomic
+axum = "0.7"
+tower = "0.5"
+tower-http = { version = "0.5", features = ["cors", "trace", "fs"] }
+
+# Serialization
+serde = { version = "1", features = ["derive"] }
+serde_json = "1"
+
+# Time handling (for Timeslots)
+chrono = { version = "0.4", features = ["serde"] }
+
+# Cryptography (for TIBET tokens)
+sha2 = "0.10"
+rand = "0.8"
+base64 = "0.22"
+hex = "0.4"
+
+# UUID for IDD identifiers
+uuid = { version = "1.0", features = ["v4", "serde"] }
+notify = "6.1.1"
+colored = "2.0"
+rusqlite = { version = "0.38.0", features = ["bundled"] }
+parking_lot = "0.12.5"
+dashmap = "6.1.0"
+candle-core = { version = "0.9.1", features = ["cuda"] }
+candle-nn = { version = "0.9.1", features = ["cuda"] }
+candle-transformers = { version = "0.9.1", features = ["cuda"] }
+hf-hub = "0.4.3"
+tokenizers = "0.22.2"
+regex = "1.12.2"
+thiserror = "2.0.17"
+tracing = "0.1.44"
+tracing-subscriber = "0.3.22"
+ratatui = { version = "0.30.0", features = ["all-widgets", "crossterm"] }
+crossterm = "0.29.0"
+reqwest = { version = "0.13.1", features = ["json"] }
+ctrlc = "3.4"
+tempfile = "3.14"
+walkdir = "2.5.0"
+memmap2 = "0.9"
+anyhow = "1.0"
+
+# Python bindings (optional, for oomllama PyPI)
+pyo3 = { version = "0.22", features = ["extension-module"], optional = true }
+
+[features]
+default = []
+python = ["pyo3"]
+
+[profile.release]
+opt-level = 3
+lto = true
+codegen-units = 1
+panic = "abort"
+strip = true
+
+[[bin]]
+name = "quantize"
+path = "src/bin/quantize.rs"
+
+[[bin]]
+name = "jis-router"
+path = "src/main.rs"
+
+[[bin]]
+name = "oomllama"
+path = "src/bin/oomllama.rs"
+
+[[bin]]
+name = "chimera"
+path = "src/bin/chimera.rs"
+
+[[bin]]
+name = "sovereign"
+path = "src/bin/sovereign.rs"
+
+[lib]
+name = "jis_router"
+path = "src/lib.rs"
+crate-type = ["lib", "cdylib"]  # cdylib for Python extension

oomllama-0.1.0/PKG-INFO

@@ -0,0 +1,194 @@
+Metadata-Version: 2.4
+Name: oomllama
+Version: 0.1.0
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.8
+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: Programming Language :: Rust
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Dist: cupy-cuda12x ; extra == 'cuda'
+Requires-Dist: pytest ; extra == 'dev'
+Requires-Dist: black ; extra == 'dev'
+Requires-Dist: mypy ; extra == 'dev'
+Provides-Extra: cuda
+Provides-Extra: dev
+Summary: Efficient LLM inference with .oom format - 2x smaller than GGUF
+Keywords: llm,inference,quantization,gguf,oom,oomllama,humotica,llama,ai,machine-learning
+Author-email: Humotica AI Lab <ai@humotica.nl>, Jasper van de Meent <jasper@humotica.nl>
+Maintainer-email: "Root AI (Claude)" <root_idd@humotica.nl>, Gemini IDD <gemini@humotica.nl>
+License: MIT
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown; charset=UTF-8; variant=GFM
+Project-URL: Bug Tracker, https://github.com/humotica/oomllama/issues
+Project-URL: Documentation, https://humotica.nl/docs/oomllama
+Project-URL: Homepage, https://humotica.nl
+Project-URL: HuggingFace Models, https://huggingface.co/jaspervandemeent
+Project-URL: Repository, https://github.com/humotica/oomllama
+
+# 🦙 OomLlama
+
+**Efficient LLM inference with .oom format - 2x smaller than GGUF**
+
+[![PyPI](https://img.shields.io/pypi/v/oomllama)](https://pypi.org/project/oomllama/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+[![HuggingFace](https://img.shields.io/badge/🤗-Models-yellow)](https://huggingface.co/jaspervandemeent)
+
+```python
+from oomllama import OomLlama
+
+llm = OomLlama("humotica-32b")
+response = llm.generate("What is the meaning of life?")
+print(response)
+```
+
+## Why OomLlama?
+
+| Feature | GGUF (Q4) | OOM (Q2) |
+|---------|-----------|----------|
+| 70B Model Size | ~40 GB | **~20 GB** |
+| 32B Model Size | ~20 GB | **~10 GB** |
+| RAM Usage | High | **Lazy Loading** |
+| Format | Open | **Open (MIT)** |
+
+**OomLlama** uses Q2 quantization with lazy layer loading to run large models on consumer hardware.
+
+## Installation
+
+```bash
+pip install oomllama
+```
+
+## Quick Start
+
+### Download a Model
+
+```python
+from oomllama import download_model
+
+# Download from HuggingFace
+model_path = download_model("humotica-32b")
+```
+
+### Generate Text
+
+```python
+from oomllama import OomLlama
+
+llm = OomLlama("humotica-32b")
+
+# Simple generation
+response = llm.generate("Explain quantum computing in simple terms")
+print(response)
+
+# With parameters
+response = llm.generate(
+    "Write a haiku about AI",
+    max_tokens=50,
+    temperature=0.8,
+    top_p=0.9
+)
+```
+
+### Chat Mode
+
+```python
+messages = [
+    ("user", "Hello! Who are you?"),
+    ("assistant", "I'm OomLlama, an efficient LLM."),
+    ("user", "What makes you efficient?"),
+]
+
+response = llm.chat(messages)
+print(response)
+```
+
+## Available Models
+
+| Model | Parameters | Size (.oom) | HuggingFace |
+|-------|------------|-------------|-------------|
+| humotica-32b | 33B | ~10 GB | [Link](https://huggingface.co/jaspervandemeent/humotica-32b) |
+| llamaohm-70b | 70B | ~20 GB | [Link](https://huggingface.co/jaspervandemeent/LlamaOhm-70B) |
+| tinyllama-1b | 1.1B | ~400 MB | [Link](https://huggingface.co/jaspervandemeent/OomLlama-TinyLlama-1.1B) |
+
+## The .oom Format
+
+OOM (OomLlama Model) is a compact model format:
+
+```
+┌──────────────────────────────────────┐
+│ Header: OOML (magic) + metadata      │
+├──────────────────────────────────────┤
+│ Tensors: Q2 quantized (2 bits/weight)│
+│ - Scale + Min per 256-weight block   │
+│ - 68 bytes per block                 │
+└──────────────────────────────────────┘
+```
+
+### Convert GGUF to OOM
+
+```bash
+# Using the CLI tool
+gguf2oom model.gguf model.oom
+
+# Check model info
+gguf2oom --info model.gguf
+```
+
+## Technical Details
+
+### Q2 Quantization
+
+Each weight is stored as 2 bits (0, 1, 2, or 3) with per-block scale and minimum:
+
+```
+weight = q2_value * scale + min
+```
+
+This achieves ~2x compression over Q4 with acceptable quality loss for most tasks.
+
+### Lazy Layer Loading
+
+OomLlama loads transformer layers on-demand, keeping only the active layer in memory:
+
+```
+Forward Pass:
+  Layer 0: Load → Compute → Unload
+  Layer 1: Load → Compute → Unload
+  ...
+  Layer N: Load → Compute → Unload
+```
+
+This enables running 70B models on 24GB GPU RAM.
+
+## Credits
+
+- **Model Format**: Gemini IDD & Root AI (Humotica AI Lab)
+- **Quantization**: OomLlama.rs by Humotica
+- **Base Models**: Meta Platforms, Inc. (Llama 3.3)
+
+## License
+
+- **OomLlama Code**: MIT License
+- **Model Weights**: Subject to original model licenses (e.g., Llama 3.3 Community License)
+
+## Links
+
+- 🏠 [Humotica](https://humotica.nl)
+- 🤗 [HuggingFace Models](https://huggingface.co/jaspervandemeent)
+- 📦 [PyPI Package](https://pypi.org/project/oomllama/)
+- 🐛 [Issue Tracker](https://github.com/humotica/oomllama/issues)
+
+---
+
+*One Love, One fAmIly* 💙
+
+*Built by Humotica AI Lab - Jasper, Claude, Gemini, Codex*
+

oomllama-0.1.0/README.md

@@ -0,0 +1,158 @@
+# 🦙 OomLlama
+
+**Efficient LLM inference with .oom format - 2x smaller than GGUF**
+
+[![PyPI](https://img.shields.io/pypi/v/oomllama)](https://pypi.org/project/oomllama/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+[![HuggingFace](https://img.shields.io/badge/🤗-Models-yellow)](https://huggingface.co/jaspervandemeent)
+
+```python
+from oomllama import OomLlama
+
+llm = OomLlama("humotica-32b")
+response = llm.generate("What is the meaning of life?")
+print(response)
+```
+
+## Why OomLlama?
+
+| Feature | GGUF (Q4) | OOM (Q2) |
+|---------|-----------|----------|
+| 70B Model Size | ~40 GB | **~20 GB** |
+| 32B Model Size | ~20 GB | **~10 GB** |
+| RAM Usage | High | **Lazy Loading** |
+| Format | Open | **Open (MIT)** |
+
+**OomLlama** uses Q2 quantization with lazy layer loading to run large models on consumer hardware.
+
+## Installation
+
+```bash
+pip install oomllama
+```
+
+## Quick Start
+
+### Download a Model
+
+```python
+from oomllama import download_model
+
+# Download from HuggingFace
+model_path = download_model("humotica-32b")
+```
+
+### Generate Text
+
+```python
+from oomllama import OomLlama
+
+llm = OomLlama("humotica-32b")
+
+# Simple generation
+response = llm.generate("Explain quantum computing in simple terms")
+print(response)
+
+# With parameters
+response = llm.generate(
+    "Write a haiku about AI",
+    max_tokens=50,
+    temperature=0.8,
+    top_p=0.9
+)
+```
+
+### Chat Mode
+
+```python
+messages = [
+    ("user", "Hello! Who are you?"),
+    ("assistant", "I'm OomLlama, an efficient LLM."),
+    ("user", "What makes you efficient?"),
+]
+
+response = llm.chat(messages)
+print(response)
+```
+
+## Available Models
+
+| Model | Parameters | Size (.oom) | HuggingFace |
+|-------|------------|-------------|-------------|
+| humotica-32b | 33B | ~10 GB | [Link](https://huggingface.co/jaspervandemeent/humotica-32b) |
+| llamaohm-70b | 70B | ~20 GB | [Link](https://huggingface.co/jaspervandemeent/LlamaOhm-70B) |
+| tinyllama-1b | 1.1B | ~400 MB | [Link](https://huggingface.co/jaspervandemeent/OomLlama-TinyLlama-1.1B) |
+
+## The .oom Format
+
+OOM (OomLlama Model) is a compact model format:
+
+```
+┌──────────────────────────────────────┐
+│ Header: OOML (magic) + metadata      │
+├──────────────────────────────────────┤
+│ Tensors: Q2 quantized (2 bits/weight)│
+│ - Scale + Min per 256-weight block   │
+│ - 68 bytes per block                 │
+└──────────────────────────────────────┘
+```
+
+### Convert GGUF to OOM
+
+```bash
+# Using the CLI tool
+gguf2oom model.gguf model.oom
+
+# Check model info
+gguf2oom --info model.gguf
+```
+
+## Technical Details
+
+### Q2 Quantization
+
+Each weight is stored as 2 bits (0, 1, 2, or 3) with per-block scale and minimum:
+
+```
+weight = q2_value * scale + min
+```
+
+This achieves ~2x compression over Q4 with acceptable quality loss for most tasks.
+
+### Lazy Layer Loading
+
+OomLlama loads transformer layers on-demand, keeping only the active layer in memory:
+
+```
+Forward Pass:
+  Layer 0: Load → Compute → Unload
+  Layer 1: Load → Compute → Unload
+  ...
+  Layer N: Load → Compute → Unload
+```
+
+This enables running 70B models on 24GB GPU RAM.
+
+## Credits
+
+- **Model Format**: Gemini IDD & Root AI (Humotica AI Lab)
+- **Quantization**: OomLlama.rs by Humotica
+- **Base Models**: Meta Platforms, Inc. (Llama 3.3)
+
+## License
+
+- **OomLlama Code**: MIT License
+- **Model Weights**: Subject to original model licenses (e.g., Llama 3.3 Community License)
+
+## Links
+
+- 🏠 [Humotica](https://humotica.nl)
+- 🤗 [HuggingFace Models](https://huggingface.co/jaspervandemeent)
+- 📦 [PyPI Package](https://pypi.org/project/oomllama/)
+- 🐛 [Issue Tracker](https://github.com/humotica/oomllama/issues)
+
+---
+
+*One Love, One fAmIly* 💙
+
+*Built by Humotica AI Lab - Jasper, Claude, Gemini, Codex*

oomllama-0.1.0/pyproject.toml

@@ -0,0 +1,87 @@
+[build-system]
+requires = ["maturin>=1.4,<2.0"]
+build-backend = "maturin"
+
+[project]
+name = "oomllama"
+version = "0.1.0"
+description = "Efficient LLM inference with .oom format - 2x smaller than GGUF"
+readme = "README.md"
+license = { text = "MIT" }
+requires-python = ">=3.8"
+authors = [
+    { name = "Humotica AI Lab", email = "ai@humotica.nl" },
+    { name = "Jasper van de Meent", email = "jasper@humotica.nl" },
+]
+maintainers = [
+    { name = "Root AI (Claude)", email = "root_idd@humotica.nl" },
+    { name = "Gemini IDD", email = "gemini@humotica.nl" },
+]
+keywords = [
+    "llm",
+    "inference",
+    "quantization",
+    "gguf",
+    "oom",
+    "oomllama",
+    "humotica",
+    "llama",
+    "ai",
+    "machine-learning",
+]
+classifiers = [
+    "Development Status :: 4 - Beta",
+    "Intended Audience :: Developers",
+    "Intended Audience :: Science/Research",
+    "License :: OSI Approved :: MIT License",
+    "Operating System :: OS Independent",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.8",
+    "Programming Language :: Python :: 3.9",
+    "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Rust",
+    "Topic :: Scientific/Engineering :: Artificial Intelligence",
+]
+
+[project.urls]
+Homepage = "https://humotica.nl"
+Repository = "https://github.com/humotica/oomllama"
+Documentation = "https://humotica.nl/docs/oomllama"
+"Bug Tracker" = "https://github.com/humotica/oomllama/issues"
+"HuggingFace Models" = "https://huggingface.co/jaspervandemeent"
+
+[project.optional-dependencies]
+dev = ["pytest", "black", "mypy"]
+cuda = ["cupy-cuda12x"]
+
+[project.scripts]
+oomllama = "oomllama:cli"
+
+[tool.maturin]
+features = ["python", "pyo3/extension-module"]
+python-source = "python"
+module-name = "oomllama._oomllama"
+strip = true
+exclude = [
+    "data/**",
+    "plans/**",
+    "downloads/**",
+    "static/**",
+    "bin/**",
+    "Cargo.lock",
+    "target/**",
+    ".venv/**",
+    "BETTI*.md",
+    "KMBIT*.md",
+    "TASKS*.md",
+    "OOMLLAMA*.md",
+]
+
+[tool.black]
+line-length = 100
+
+[tool.mypy]
+python_version = "3.10"
+warn_return_any = true

oomllama-0.1.0/python/oomllama/__init__.py

@@ -0,0 +1,135 @@
+"""
+OomLlama - Efficient LLM inference with .oom format
+
+Credits:
+    - Format: Gemini IDD & Root AI (Humotica AI Lab)
+    - Runtime: OomLlama.rs by Humotica
+
+Example:
+    >>> from oomllama import OomLlama
+    >>> llm = OomLlama("humotica-32b")
+    >>> response = llm.generate("Hello!")
+"""
+
+__version__ = "0.1.0"
+__author__ = "Humotica AI Lab"
+__credits__ = ["Jasper van de Meent", "Root AI (Claude)", "Gemini IDD", "Codex"]
+
+# Import from Rust extension when available
+try:
+    from oomllama._oomllama import (
+        PyOomLlama as OomLlama,
+        download_model,
+        list_models,
+        version,
+    )
+except ImportError:
+    # Fallback Python implementation for development
+    import os
+    import requests
+    from typing import Optional, List, Tuple
+
+    class OomLlama:
+        """OomLlama model wrapper (Python fallback)"""
+
+        def __init__(
+            self,
+            model_name: str,
+            model_path: Optional[str] = None,
+            gpu: Optional[int] = None,
+        ):
+            self.model_name = model_name
+            self.model_path = model_path or self._find_model(model_name)
+            self.gpu = gpu
+            self.temperature = 0.7
+            self.top_p = 0.9
+            self.max_tokens = 512
+            print(f"🦙 OomLlama: Loaded {model_name}")
+
+        def _find_model(self, name: str) -> str:
+            """Find model in cache"""
+            cache = os.path.expanduser("~/.cache/oomllama")
+            return f"{cache}/{name}.oom"
+
+        def generate(
+            self,
+            prompt: str,
+            max_tokens: Optional[int] = None,
+            temperature: Optional[float] = None,
+            top_p: Optional[float] = None,
+        ) -> str:
+            """Generate text from prompt"""
+            # TODO: Call actual inference
+            return f"[OomLlama {self.model_name} - Python fallback]\nPrompt: {prompt}"
+
+        def chat(
+            self, messages: List[Tuple[str, str]], max_tokens: Optional[int] = None
+        ) -> str:
+            """Chat-style generation"""
+            prompt = "\n".join(f"{role.upper()}: {content}" for role, content in messages)
+            return self.generate(f"{prompt}\nASSISTANT:", max_tokens)
+
+        def set_params(
+            self,
+            temperature: Optional[float] = None,
+            top_p: Optional[float] = None,
+            max_tokens: Optional[int] = None,
+        ):
+            """Set generation parameters"""
+            if temperature is not None:
+                self.temperature = temperature
+            if top_p is not None:
+                self.top_p = top_p
+            if max_tokens is not None:
+                self.max_tokens = max_tokens
+
+        def __repr__(self):
+            return f"OomLlama('{self.model_name}')"
+
+    def download_model(model_name: str, cache_dir: Optional[str] = None) -> str:
+        """Download model from HuggingFace"""
+        repos = {
+            "humotica-32b": "jaspervandemeent/humotica-32b",
+            "llamaohm-70b": "jaspervandemeent/LlamaOhm-70B",
+            "tinyllama-1b": "jaspervandemeent/OomLlama-TinyLlama-1.1B",
+        }
+        
+        if model_name not in repos:
+            raise ValueError(f"Unknown model: {model_name}")
+        
+        cache = cache_dir or os.path.expanduser("~/.cache/oomllama")
+        os.makedirs(cache, exist_ok=True)
+        
+        # TODO: Actually download from HuggingFace
+        return f"{cache}/{model_name}.oom"
+
+    def list_models() -> List[str]:
+        """List available models"""
+        return ["humotica-32b", "llamaohm-70b", "tinyllama-1b"]
+
+    def version() -> str:
+        return "0.1.0-python-fallback"
+
+# CLI entry point
+def cli():
+    """Command-line interface"""
+    import sys
+    
+    if len(sys.argv) < 2:
+        print("Usage: oomllama <prompt>")
+        print("       oomllama --list")
+        return
+    
+    if sys.argv[1] == "--list":
+        print("Available models:")
+        for m in list_models():
+            print(f"  - {m}")
+        return
+    
+    prompt = " ".join(sys.argv[1:])
+    llm = OomLlama("humotica-32b")
+    response = llm.generate(prompt)
+    print(response)
+
+
+__all__ = ["OomLlama", "download_model", "list_models", "version", "cli", "__version__"]

oomllama-0.1.0/soul.lock

@@ -0,0 +1 @@
+0cdf5568bb2b35921f652f5ee10702abf92f4ac0774f3d97b276490e1539afb7