OMNeT++ Preview Versions

This technology preview introduces Large Language Model (LLM) integration to OMNeT++.

Key features include an integrated chat interface within Qtenv for direct interaction with LLM services, and a dedicated Model Context Protocol (MCP) server. The MCP server allows an LLM to query and manipulate active simulations in real time. For configuration details, please refer to the User Guide.

The integration includes an optional JIT C++ interpreter, allowing the LLM to generate and execute code on the fly to programmatically inspect the simulation and answer complex queries. To ensure security while dynamic code execution is enabled, the simulation must be run within a sandboxed environment to prevent arbitrary access to the host system.

Explore the Showcases

Discover the capabilities of this build through these practical examples:

• Finding a 3-Way Collision in Aloha – Using AI to identify specific network events.
• Diagnosing Clock Sync Anomalies in TSN – Root-cause analysis in Time-Sensitive Networking.

Important Notes

• Platform Support: This snapshot is limited to Linux (x86_64) and macOS (aarch64).
• Dependencies: This version introduces new requirements. Run the install.sh script in the root directory to update your environment.
• Local Inference: If you prefer local models over cloud APIs, ensure ollama is installed and running.
• C++ Interpreter: Interactive C++ functionality is Linux-only and requires launching the simulation via the opp_sandbox script.

See the GitHub release page and the discussion on GitHub Discussions for more details.