v2.1: The Agentic Architecture

We are thrilled to announce the release of ProMoAI v2.1. This update marks a fundamental shift in the project's architecture, evolving into a fully Agentic Framework for Process Mining.

With the introduction of the PMAx engine, ProMoAI no longer just "chats" about processes—it reasons, plans, and executes complex data science tasks autonomously.

🌟 What’s New?

1. PMAx: The Agentic Analytics Engine

   The core of v2.1 is the PMAx framework. It utilizes a "Divide-and-Conquer" multi-agent architecture to handle complex business questions:

   • 🧠 Engineer Agent: Generates and executes Python code (powered by PM4Py) locally to compute exact metrics.

   • 📊 Analyst Agent: Interprets the results and generates a narrative report with visualizations.

2. Privacy-First Architecture

   Data never leaves your environment. ProMoAI v2.1 only sends metadata (column names, data types, high-level summaries and samples) to the LLM. Your actual event log stays within your local execution environment.

3. Redesigned User Interface:

   We’ve completely refreshed the UI to support the new agentic workflow:

   • Cleaner layout with clearer separation between analysis, actions, and outputs
   • More intuitive interaction with inputs, exports, and refinement tools

🔥 What This Means

ProMoAI is no longer just an assistant, it’s an autonomous process mining analyst.

✨ What’s Next

We’re continuing to expand agent collaboration to enable deeper process insights. v2.1 is the foundation for a much larger shift.

More to come 🔜

v1.3

What's Changed

• Integrate POWL 2.0
• Added support for log and model as input
• Added BPMNAnalyzer
• Improved exception handling
• Updated models and AI providers
• Bugfixes

Full Changelog: https://github.com/humam-kourani/ProMoAI/commits/v1.3