Version 2.7.0 , 01/2026

The following libraries have been upgraded to:

• cURL 8.16.0

• Go v1.25.1

• Node.js Package: axios v1.12.2

• OpenSSL 3.6.0

• Pip (pypi) 25.2

• Python 3.10.19

• Python 3.11.14

• Python 3.12.12

• Triton-Inference-Server-Core r25.07

• Not applicable

• AI Inference Server applies mTLS (mutual Transport Layer Security) protocol for the communication between the apps and containers (e.g. AI Inference Server <-> AI Asset Manager)

• AI Inference Server provides pipeline details also for a non-activated pipeline via RestAPI (for AI Asset Manager)

• Type "Secret" is introduced for the pipeline parameters. Its value can be entered on the user interface of AI Inference Server

• Enhanced user interface with updated design elements including icons, labels, and buttons

• Database engine switched from PostgreSQL to Better SQLite

• Issue 1954416 - Connections explicitly mapped by name are no longer automatically remapped.

• Issue 1962551 - Incorrect color conversion when visualizing images

• Issue 1976152 - Prevented input data loss caused by irrelevant inputs during step-to-step communication

• Issue 1836316 - Corrected inaccurate memory usage reporting

• The "Run" function for Python models will be discontinued in January 2027. Use the "process_input" function instead.

• Database engine PostgreSQL is replaced. Upgrade your product to AI Inference Server v2.7 (or newer version) till January 2027, otherwise your data and log messages within AI Inference Server won't be kept.

• Not applicable