v1.5.0

Signed-off-by: bigcat88 <[email protected]>

Author: bigcat88

CHANGELOG.md

@@ -2,6 +2,22 @@
 
 All notable changes to this project will be documented in this file.
 
+## [1.5.0 - 2025-05-19]
+
+Visionatrix service has been updated from version `2.3.0` to `2.4.1`.
+
+### Added
+
+- New **ACE-Step Audio music flow**.
+- Enabling/disabling ComfyUI `Smart Memory` now should be done from the **Settings -> Workers** UI page. **By default** `Smart memory` from now is enabled.
+- Enabling/disabling ComfyUI `save metadata` option now should be done from the **Settings -> Settings** UI page. **By default** `save metadata` will be disabled.
+- Changing ComfyUI **cache** settings(`lru`, `none`) now should be done from the **Settings -> Workers** UI page.
+- UI option to enable processing `VAE` on the CPU, located in the **Settings -> Workers** page.
+
+### Changed
+
+- CUDA: Updated from `12.6` to `12.8`.
+
 ## [1.4.1 - 2025-05-08]
 
 ### Added

Dockerfile

@@ -59,7 +59,7 @@ RUN --mount=type=cache,target=/root/.cache/pip \
     elif [ "$BUILD_TYPE" = "cpu" ]; then \
         venv/bin/python -m pip install torch==2.7.0 torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu; \
     else \
-        venv/bin/python -m pip install torch==2.7.0 torchvision torchaudio; \
+        venv/bin/python -m pip install torch==2.7.0 torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu128; \
     fi
 
 COPY ex_app/lib/exclude_nodes.py ex_app/lib/exclude_flows.py /ex_app/lib/

README.md

@@ -63,6 +63,22 @@ We offer two types of installation:
 
 For more information, please refer to the original [Visionatrix documentation](https://visionatrix.github.io/VixFlowsDocs/).
 
+## Freeing GPU Memory After Execution
+
+By default, models remain resident in GPU memory after each task. To control this behavior, we’ve added two options on the **Settings → Workers** page:
+
+1. **Smart Memory** (enable/disable)
+2. **Cache Type** (select caching strategy)
+
+If you need GPU memory to be released after task execution, we recommend:
+
+1. **For systems with > 64 GB RAM**
+   Disable **Smart Memory**. This will offload GPU memory to your system RAM, freeing up VRAM.
+
+2. **For systems with ≤ 64 GB RAM**
+   a. Disable **Smart Memory**.
+   b. Set **Cache Type** to **None**.
+
 ## Questions
 
 Feel free to ask questions or report issues.

appinfo/info.xml

@@ -34,7 +34,7 @@ Share your custom workflows, big or small. For guidance on creating AI workflows
 
 📚 For more details, visit the [repository](https://github.com/cloud-py-api/visionatrix) readme.
 	]]></description>
-	<version>1.4.1</version>
+	<version>1.5.0</version>
 	<licence>MIT</licence>
 	<author mail="[email protected]" homepage="https://github.com/andrey18106">Andrey Borysenko</author>
 	<author mail="[email protected]" homepage="https://github.com/bigcat88">Alexander Piskun</author>
@@ -56,7 +56,7 @@ Share your custom workflows, big or small. For guidance on creating AI workflows
 		<docker-install>
 			<registry>ghcr.io</registry>
 			<image>cloud-py-api/visionatrix</image>
-			<image-tag>2.3.0</image-tag>
+			<image-tag>2.4.1</image-tag>
 		</docker-install>
 		<routes>
 			<route>

ex_app/lib/main.py

@@ -393,7 +393,7 @@ def start_visionatrix() -> None:
             # Run worker in background and redirect output to worker.log
             worker_log = open("worker.log", "wb")
             subprocess.Popen(
-                [visionatrix_python, "-m", "visionatrix", "run", "--mode=WORKER", "--disable-smart-memory"],
+                [visionatrix_python, "-m", "visionatrix", "run", "--mode=WORKER"],
                 stdout=worker_log,
                 stderr=subprocess.STDOUT,
             )