|
|
|
|
Release of Stochos Flow and Stochos 7.2.0
|
STOCHOS Flow is a graphic AI workbench for STOCHOS with agent-based automation and natural language interaction
|
|
|
|
|
Dear Customers and Partners
We are proud to announce that our new product Stochos Flow an agentic workflowGUI and finally a GUI for all Stochos functionalities has been released today.
|
|
Some of the highlights it covers:
- A node based graphical user interface for all Stochos functionalities which allows you to build Model training pipelines, optimization task, data preparation and much more by just adding and connecting nodes. Connection to Ansys Workbench Projects is also included.
- A Python solver node to add new functionality and allows very flexible customization which can be exported to a permanent node
- Import / Export of nodes to share your customization with colleagues or receive them from us or our partners
- Workflows can be exported as Python code to run them without the UI
- Full support of an LLM Agent which can build workflows for you just describe what you want to achieve. It also supports writing python solver nodes to extend missing functionalities, so you don't have to code yourself. We support local llms via Ollama or you can use Open AI / Anthropic API Keys to use models via the cloud
- An integrated RAG system to add documents to enhance the knowledge of the agent and add context if needed, for example about how to use certain packages / APIs. We ship it with full Pyansys documentation and Stochos documentation prebuilt
- It all runs locally if you are using the local Agent via Ollama no internet connection required (highly recommend a 12-16 GB GPU) !
|
|
|
|
|
|
|
|
|
To get an overview of the functionalities we have prepared a Video for you:
|
|
This platform will mark a new beginning of how you can work with data and building workflows. We have a lot of ideas for future updates like sub agents for specific tasks and want to get your feedback about missing functionalities you would like to see in the future.
|
|
Note that you can use this GUI also without the Agent to use all Stochos capabilities without needing to program.
|
|
|
|
|
How to Start
|
|
|
You will find Stochos Flow in our Customer Portal under Releases (like the StochosPython module). We highly recommend having a look at the supplementary material provided in our customer portal for Stochos Flow the Stochos_Flow_examples.zipwhich includes all examples / templates you already know from the Stochos Python package as already built project (.spfj files) and a prompt example how to build this project with the help of the agent (.prompt.txt file). Additionally, in the help menu of Stochos Flow you find a lot of helpful tips and there is a dedicated prompting guide on how to work with the agent. In the future we will release also additional training materials.
|
|
|
|
|
Important Issue in Stochos Python Pacakge releases from June 2025 till now
|
Besides that, we want also to inform you that we have released a new version of Stochos Python package in Version 7.2.0 which has a lot of new stuff and improvements. It also fixes one critical bug that has been there since June 2025 in the Bayesian_opt class:
|
- Critical fix for the simple Bayesian optimization API (config based). Since the release of this new API, trained models were not updated during adaptations, causing the optimization to not work as intended and leading to poor results. This occurred only when the simple interface was used together with bayesian_optimization.optimize(), or when next_sample() was used in a for-loop without updating the output_config and re-initializing the bayesian_opt object with the retrained model inside the loop.
|
So, all of you which have been using the new config-based optimization setup for the Bayesian_opt module have very likely got very poor results and should rerun them with 7.2.0.
|
|
|
|
|
We hope you enjoy working with our new product STOCHOS Flow, and we warmly invite you to share your feedback, ideas, and impressions with us to help shape its future development.
|
|
Best regards and thank you,
|
Your PI Probaligence Team
|
|
|
|
|
|
|
|
Am Schammacher Feld 37, 85567 Grafing b. München, Germany
|
|
mail: info@probaligence.de
|
|
|
|
|
|
