termux-diffusion 1.6.4__tar.gz → 1.6.6__tar.gz

termux_diffusion-1.6.6/PKG-INFO

@@ -0,0 +1,110 @@
+Metadata-Version: 2.4
+Name: termux-diffusion
+Version: 1.6.6
+Summary: Pure CPU On-Device Stable Diffusion Runtime for Android Termux (ARM64 Bionic NEON SIMD)
+Home-page: https://github.com/uno-km/termux-diffusion
+Author: uno-km (AMEVA Foundation)
+Author-email: "uno-km (AMEVA Foundation)" <hosequelbo@gmail.com>
+License: Apache-2.0
+Project-URL: Homepage, https://github.com/uno-km/termux-diffusion
+Project-URL: Documentation, https://uno-km.github.io/termux-diffusion/
+Project-URL: npm Package, https://www.npmjs.com/package/termux-diffusion
+Project-URL: Bug Tracker, https://github.com/uno-km/termux-diffusion/issues
+Project-URL: Source, https://github.com/uno-km/termux-diffusion
+Keywords: stable-diffusion,diffusion,termux,android,samsung-galaxy,edge-ai,on-device-ai,image-generation,text-to-image,gguf,arm64,aarch64,neon,simd,bionic-libc,vae-tiling,galaxy-s25,galaxy-s21
+Classifier: Development Status :: 5 - Production/Stable
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: POSIX :: Linux
+Classifier: Operating System :: Android
+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: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Multimedia :: Graphics
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Provides-Extra: gpu
+Requires-Dist: ameva-runtime>=2.0.0; extra == "gpu"
+Dynamic: author
+Dynamic: classifier
+Dynamic: home-page
+Dynamic: license-file
+Dynamic: requires-python
+
+# Termux-Diffusion (Python)
+
+[![PyPI](https://img.shields.io/pypi/v/termux-diffusion.svg?style=flat-square&color=0369a1)](https://pypi.org/project/termux-diffusion/)
+[![Python](https://img.shields.io/pypi/pyversions/termux-diffusion.svg?style=flat-square)](https://pypi.org/project/termux-diffusion/)
+[![License](https://img.shields.io/badge/License-Apache_2.0-004499.svg?style=flat-square)](https://github.com/uno-km/termux-diffusion)
+
+> **Pure CPU Native On-Device Stable Diffusion Runtime for Android Termux & Samsung Galaxy.**  
+> *Zero PRoot. Zero Virtualization. 100% Native ARM64 Bionic libc & NEON SIMD Vector Acceleration.*
+
+---
+
+## Installation & Native Engine Provisioning
+
+```bash
+# 1. Install Python package from PyPI
+pip install termux-diffusion
+
+# 2. Provision prebuilt Pure CPU native engine & OpenMP runtime
+termux-diffusion install
+```
+
+---
+
+## Quickstart (Python SDK)
+
+```python
+import termux_diffusion as td
+
+# Generate image using pure CPU with LCM anime preset
+image_path = td.generate(
+    prompt="cute polar bear in snow, high quality illustration",
+    model="anime",
+    steps=6,
+    device="cpu",
+    output_path="polar_bear.png"
+)
+print(f"Generated and saved to Android Gallery: {image_path}")
+```
+
+## CLI Usage
+
+```bash
+termux-diffusion generate "cute dolphin, vibrant ocean" -m anime --cpu -o dolphin.png
+```
+
+---
+
+## Key Capabilities
+
+* **ARMv8.2-A DotProd & FP16 NEON SIMD**: Native matrix acceleration compiled directly against Android Bionic libc.
+* **Built-in VAE Tiling**: Reduces peak RAM consumption by ~70% during latent-to-pixel decoding to prevent Out-Of-Memory (OOM) aborts.
+* **Android MediaStore Auto-Indexing**: Automatically synchronizes generated images into `Pictures/TermuxDiffusion` and the native Samsung Gallery.
+* **Fail-Fast Protection**: Blocks unintended on-device compilation cycles to protect mobile devices from severe thermal throttling.
+
+---
+
+## Hardware Benchmarks (Physical Android Devices)
+
+| Device | AP / Architecture | Wall Time | Sampling Time | Peak RAM |
+| :--- | :--- | :--- | :--- | :--- |
+| **Galaxy S25** | Snapdragon 8 Elite | **15m 52s (951s)** | 11m 37s (697s) | ~2.5 GB |
+| **Galaxy S21** | Exynos 2100 | **53m 45s (3221s)** | 7m 47s (467s) | ~2.4 GB |
+
+---
+
+## Documentation & Repository
+- [Official Architecture & API Reference](https://uno-km.vercel.app/lib/diffusion/)
+- [GitHub Repository](https://github.com/uno-km/termux-diffusion)
+
+## License
+Apache-2.0 License. Copyright (c) 2026 Eunho Kim (@uno-km).
+

termux_diffusion-1.6.6/README.md

@@ -0,0 +1,121 @@
+# Termux-Diffusion
+
+[![PyPI](https://img.shields.io/pypi/v/termux-diffusion.svg?style=flat-square&color=0369a1)](https://pypi.org/project/termux-diffusion/)
+[![Python](https://img.shields.io/pypi/pyversions/termux-diffusion.svg?style=flat-square)](https://pypi.org/project/termux-diffusion/)
+[![npm](https://img.shields.io/npm/v/termux-diffusion.svg?style=flat-square&color=b91c1c)](https://www.npmjs.com/package/termux-diffusion)
+[![npm downloads](https://img.shields.io/npm/dm/termux-diffusion.svg?style=flat-square&color=b91c1c)](https://www.npmjs.com/package/termux-diffusion)
+[![License](https://img.shields.io/badge/License-Apache_2.0-004499.svg?style=flat-square)](https://github.com/uno-km/termux-diffusion)
+
+> **Pure CPU Native On-Device Stable Diffusion Runtime for Android Termux & Samsung Galaxy.**  
+> *Zero PRoot. Zero Virtualization. 100% Native ARM64 Bionic libc & NEON SIMD Vector Acceleration.*
+
+---
+
+## 1. Architecture & Key Engineering Features
+
+`termux-diffusion` is a standalone on-device image synthesis engine engineered specifically for Android smartphones running Termux. It operates directly against Android's native Bionic libc ABI with zero virtualization layers.
+
+* **Pure CPU Native Bionic Execution**: Bypasses heavy PRoot Linux containers and virtualization overhead, binding directly to native ARM64 CPU instructions.
+* **One-Touch Prebuilt Provisioning**: Provisions precompiled native binaries (`sd-cli-cpu`) and the OpenMP runtime (`libomp.so`) in seconds via a single command, completely eliminating 20-minute on-device C++ compilation cycles.
+* **ARMv8.2-A DotProd & FP16 NEON SIMD**: Employs optimized vector matrix multiplication kernels tailored for modern mobile microarchitectures (Cortex-X, Cortex-A78, Oryon).
+* **Built-in VAE Tiling**: Eliminates the ~1.2 GB memory spike during latent-to-pixel decoding, reducing peak RAM consumption by ~70% and preventing Android low-memory killer (LMK) aborts.
+* **Automated Android MediaStore Indexing**: Synchronizes synthesized images directly to `Pictures/TermuxDiffusion` and broadcasts media scan intents for real-time visibility in the Samsung Gallery app.
+* **Fail-Fast Thermal Protection**: On-device compilation on Termux is blocked by default without explicit opt-in (`TERMUX_DIFFUSION_ALLOW_SOURCE_BUILD=1`), guarding the device against excessive thermal throttling.
+
+---
+
+## 2. Installation & One-Touch Native Provisioning
+
+### 2.1 System Prerequisites (Termux)
+In the Termux terminal, install the core runtime dependencies:
+```bash
+pkg update && pkg install -y python nodejs clang termux-api
+```
+
+### 2.2 Package Installation
+
+* **Python SDK (PyPI)**:
+  ```bash
+  pip install termux-diffusion
+  termux-diffusion install
+  ```
+
+* **Node.js SDK & CLI (NPM)**:
+  ```bash
+  npm install -g termux-diffusion
+  npx termux-diffusion install
+  ```
+
+> The `install` command provisions the standalone CPU native engine (`sd-cli-cpu`) and companion OpenMP runtime (`libomp.so`) directly into `~/.cache/termux-diffusion/bin`.
+
+---
+
+## 3. Quickstart & Usage
+
+### 3.1 CLI Interface
+```bash
+# Generate image using pure CPU with default LCM anime preset (6 steps)
+termux-diffusion generate "cute polar bear, high quality illustration" -m anime --cpu -o polar_bear.png
+
+# Custom dimensions, steps, and threads
+termux-diffusion generate "cute dolphin, vibrant ocean, 8k resolution" -W 512 -H 512 --steps 6 -t 4 -o dolphin.png
+```
+
+### 3.2 Python SDK
+```python
+import termux_diffusion as td
+
+# Synthesize image via Pure CPU Native Engine
+image_path = td.generate(
+    prompt="cute polar bear in snow, high quality illustration",
+    model="anime",
+    steps=6,
+    device="cpu",
+    output_path="polar_bear.png"
+)
+print(f"Generated artifact: {image_path}")
+```
+
+### 3.3 Node.js / TypeScript SDK
+```typescript
+import { generate } from "termux-diffusion";
+
+async function main() {
+  const result = await generate({
+    prompt: "cute dolphin leaping through waves, high quality illustration",
+    model: "anime",
+    steps: 6,
+    device: "cpu",
+    outputPath: "dolphin.png"
+  });
+  console.log("Image synthesized:", result.outputPath);
+}
+
+main();
+```
+
+---
+
+## 4. Verified Hardware Benchmarks (Ground Truth)
+
+All benchmarks measured on physical Android hardware using pure CPU execution (`sd-cli-cpu`, 4 threads, NEON SIMD) with the `anime` preset (DreamShaper 8 LCM Q4_0, 512x512, 6 steps):
+
+| Device | SoC & CPU Architecture | Total Time (Wall Time) | Sampling Time | Peak RAM (VmRSS) | Output Resolution |
+| :--- | :--- | :--- | :--- | :--- | :--- |
+| **Samsung Galaxy S25** | Snapdragon 8 Elite (Oryon CPU) | **15m 52s (951s)** | **11m 37s (697s)** | **~2.5 GB** | 512 × 512 PNG |
+| **Samsung Galaxy S21** | Exynos 2100 (Cortex-X1 + A78) | **53m 45s (3221s)** | **7m 47s (467s)** | **~2.4 GB** | 512 × 512 PNG |
+
+*Both devices completed end-to-end inference and synchronized outputs to the Android MediaStore without memory leaks or crash events.*
+
+---
+
+## 5. Official Ecosystem Documentation
+- [Official Architecture & API Reference](https://uno-km.vercel.app/lib/diffusion/)
+- [Ecosystem Metrics & Registry Stats](https://uno-km.vercel.app/foundation/metrics)
+- [AMEVA Open-Source Foundation Portal](https://uno-km.vercel.app/foundation/index.html)
+
+---
+
+## 6. License
+Licensed under the Apache-2.0 License. Copyright (c) 2026 Eunho Kim ([@uno-km](https://github.com/uno-km)).
+

termux_diffusion-1.6.6/README.pypi.md

@@ -0,0 +1,72 @@
+# Termux-Diffusion (Python)
+
+[![PyPI](https://img.shields.io/pypi/v/termux-diffusion.svg?style=flat-square&color=0369a1)](https://pypi.org/project/termux-diffusion/)
+[![Python](https://img.shields.io/pypi/pyversions/termux-diffusion.svg?style=flat-square)](https://pypi.org/project/termux-diffusion/)
+[![License](https://img.shields.io/badge/License-Apache_2.0-004499.svg?style=flat-square)](https://github.com/uno-km/termux-diffusion)
+
+> **Pure CPU Native On-Device Stable Diffusion Runtime for Android Termux & Samsung Galaxy.**  
+> *Zero PRoot. Zero Virtualization. 100% Native ARM64 Bionic libc & NEON SIMD Vector Acceleration.*
+
+---
+
+## Installation & Native Engine Provisioning
+
+```bash
+# 1. Install Python package from PyPI
+pip install termux-diffusion
+
+# 2. Provision prebuilt Pure CPU native engine & OpenMP runtime
+termux-diffusion install
+```
+
+---
+
+## Quickstart (Python SDK)
+
+```python
+import termux_diffusion as td
+
+# Generate image using pure CPU with LCM anime preset
+image_path = td.generate(
+    prompt="cute polar bear in snow, high quality illustration",
+    model="anime",
+    steps=6,
+    device="cpu",
+    output_path="polar_bear.png"
+)
+print(f"Generated and saved to Android Gallery: {image_path}")
+```
+
+## CLI Usage
+
+```bash
+termux-diffusion generate "cute dolphin, vibrant ocean" -m anime --cpu -o dolphin.png
+```
+
+---
+
+## Key Capabilities
+
+* **ARMv8.2-A DotProd & FP16 NEON SIMD**: Native matrix acceleration compiled directly against Android Bionic libc.
+* **Built-in VAE Tiling**: Reduces peak RAM consumption by ~70% during latent-to-pixel decoding to prevent Out-Of-Memory (OOM) aborts.
+* **Android MediaStore Auto-Indexing**: Automatically synchronizes generated images into `Pictures/TermuxDiffusion` and the native Samsung Gallery.
+* **Fail-Fast Protection**: Blocks unintended on-device compilation cycles to protect mobile devices from severe thermal throttling.
+
+---
+
+## Hardware Benchmarks (Physical Android Devices)
+
+| Device | AP / Architecture | Wall Time | Sampling Time | Peak RAM |
+| :--- | :--- | :--- | :--- | :--- |
+| **Galaxy S25** | Snapdragon 8 Elite | **15m 52s (951s)** | 11m 37s (697s) | ~2.5 GB |
+| **Galaxy S21** | Exynos 2100 | **53m 45s (3221s)** | 7m 47s (467s) | ~2.4 GB |
+
+---
+
+## Documentation & Repository
+- [Official Architecture & API Reference](https://uno-km.vercel.app/lib/diffusion/)
+- [GitHub Repository](https://github.com/uno-km/termux-diffusion)
+
+## License
+Apache-2.0 License. Copyright (c) 2026 Eunho Kim (@uno-km).
+

{termux_diffusion-1.6.4 → termux_diffusion-1.6.6}/pyproject.toml

@@ -4,11 +4,11 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "termux-diffusion"
-version = "1.6.4"
-description = "On-device Stable Diffusion runtime utilizing device resources for Android Termux & Samsung Galaxy (Dual-Engine Python & Node.js)"
-readme = "README.md"
+version = "1.6.6"
+description = "Pure CPU On-Device Stable Diffusion Runtime for Android Termux (ARM64 Bionic NEON SIMD)"
+readme = "README.pypi.md"
 requires-python = ">=3.8"
-license = { text = "MIT" }
+license = { text = "Apache-2.0" }
 authors = [
     { name = "uno-km (AMEVA Foundation)", email = "hosequelbo@gmail.com" }
 ]
@@ -22,37 +22,21 @@ keywords = [
     "on-device-ai",
     "image-generation",
     "text-to-image",
-    "txt2img",
-    "img2img",
     "gguf",
     "arm64",
     "aarch64",
-    "vulkan",
-    "vulkan-compute",
-    "spirv",
-    "adreno",
-    "mali",
-    "snapdragon",
-    "exynos",
+    "neon",
+    "simd",
     "bionic-libc",
-    "taesd",
     "vae-tiling",
-    "lora",
-    "controlnet",
-    "sdxs",
-    "sd-turbo",
-    "dreamshaper",
-    "mobile-inference",
-    "camera-roll",
     "galaxy-s25",
-    "galaxy-s20",
-    "private-ai"
+    "galaxy-s21"
 ]
 dynamic = ["classifiers"]
-dependencies = [
-    "ameva-runtime>=2.0.0",
-    "ameva-component-sdk>=0.1.0,<2.0",
-]
+dependencies = []
+
+[project.optional-dependencies]
+gpu = ["ameva-runtime>=2.0.0"]
 
 [project.urls]
 Homepage = "https://github.com/uno-km/termux-diffusion"

termux_diffusion-1.6.6/termux_diffusion/_version.py

@@ -0,0 +1 @@
+__version__ = "1.6.6"

{termux_diffusion-1.6.4 → termux_diffusion-1.6.6}/termux_diffusion/cli.py

@@ -69,7 +69,6 @@ def main(argv: Optional[List[str]] = None) -> int:
 
     # install command
     inst_parser = subparsers.add_parser("install", help="Provision native Bionic CPU engine (sd-cli-cpu & libomp.so)")
-    inst_parser.add_argument("-b", "--backend", type=str, default="cpu", choices=["cpu", "vulkan", "opencl", "auto"], help="Target compute backend (default: cpu)")
     inst_parser.add_argument("-f", "--force", action="store_true", default=False, help="Force re-installation")
 
     # doctor command
@@ -176,7 +175,7 @@ def main(argv: Optional[List[str]] = None) -> int:
             return ExitCode.BUILD_ERROR
 
     elif args.command == "install":
-        provision_engine(force=args.force, backend=args.backend)
+        provision_engine(force=args.force)
         return 0
 
     elif args.command == "doctor":

{termux_diffusion-1.6.4 → termux_diffusion-1.6.6}/termux_diffusion/core.py

@@ -485,22 +485,10 @@ def generate(
         cmd.extend(["--taesd", str(effective_taesd_path)])
 
     # 5.1 Configure Environment with companion library search paths (Termux native isolation)
-    env = os.environ.copy()
+    from .platform import get_clean_execution_env
+    env = get_clean_execution_env()
     # Bypass debug CPU shadow checks on mobile Vulkan to unlock 10x throughput
     env.setdefault("GGML_VULKAN_SKIP_CHECKS", "999999999")
-    prefix = os.environ.get("PREFIX", "/data/data/com.termux/files/usr")
-    lib_dirs = [
-        str(Path.home() / ".cache" / "termux-diffusion" / "lib"),
-        str(Path.home() / ".cache" / "termux-diffusion" / "staging" / "lib"),
-    ]
-    cur_ld = env.get("LD_LIBRARY_PATH", "")
-    valid_dirs = [d for d in lib_dirs if Path(d).is_dir() and "/usr/lib" not in d]
-    if valid_dirs:
-        env["LD_LIBRARY_PATH"] = ":".join(valid_dirs + ([cur_ld] if cur_ld else []))
-    elif "LD_LIBRARY_PATH" in env:
-        # Sanitize against $PREFIX/lib to prevent Bionic symbol collision with /system/lib64/libunwindstack.so
-        cleaned_ld = ":".join(p for p in cur_ld.split(":") if p and "/usr/lib" not in p)
-        env["LD_LIBRARY_PATH"] = cleaned_ld
 
     # 5.2 ameva-runtime DiffusionAdapter Binding (Vulkan / GPU mode)
     if effective_device in ("vulkan", "gpu") or device_mode in ("vulkan", "gpu"):

{termux_diffusion-1.6.4 → termux_diffusion-1.6.6}/termux_diffusion/downloader.py

@@ -18,9 +18,9 @@ logger = logging.getLogger("termux_diffusion.downloader")
 def atomic_download_file(
     url: str,
     dest_path: Path,
-    expected_sha256: str,
+    expected_sha256: Optional[str] = None,
     expected_size: Optional[int] = None,
-    timeout_sec: float = 30.0,
+    timeout_sec: float = 120.0,
     user_agent: str = "termux-diffusion-downloader/1.2.0"
 ) -> Path:
     """Download url to dest_path atomically using .part -> SHA256 check -> atomic rename -> rollback."""