Inside the Latest Platform and Runtime Updates

Eric, head of product, walks through platform runtime updates in versions 0.9.6 and the preview 0.9.7 (planned for release by week’s end). In 0.9.7, he demos a new runtime dashboard (alongside existing Swagger access) featuring system health, multi-campaign deployment on a single runtime instance for improved infrastructure efficiency and convenience, invocation testing with response registration, and multi-language code snippets. The release also adds a built-in Model Context Protocol (MCP) interface with tools to list campaigns, invoke, update properties, and compile custom Java classes, plus starter cloud/Kubernetes/Docker deployment configs, improved threading performance, and initial automated testing that generates and runs Cucumber tests and environment/config checks. For 0.9.6, he highlights just-in-time Java compilation for custom logic, runtime images with/without deep learning, parallel runtime calls, Mongo client access in pre-scoring logic, combining multiple tables into a customer feature store, custom (and later asymmetric) reward functions for dynamic recommenders, using another runtime as a data source, new behavioral dynamic algorithms, virtual variables, convenience APIs (refresh options, learning trigger, Cassandra config update), and MLflow integration. 00:24 What’s New in 097 01:22 Runtime Dashboard Tour 01:44 Multi Campaign Deployments 04:11 Invocation Testing and Snippets 05:54 Built In MCP Tools 08:20 Automated Testing Dashboard 11:06 097 Summary and Q&A 12:07 096 Just In Time Java 13:22 Parallel Runtime Calls 15:05 Feature Store Enhancements 16:14 Dynamic Rewards and Algorithms 17:06 0961 Data Source Runtimes 19:04 Virtual Variables and APIs 21:14 MLflow and Final Wrap Up