●Change8

v3.5.1

📦 mlflowView on GitHub →
✨ 4 features🐛 16 fixes🔧 17 symbols

Summary

MLflow 3.5.1 is a patch release that adds new CLI and deployment features, introduces a monitoring API, and includes numerous bug fixes across evaluation, tracing, tracking, and UI components.

✨ New Features

  • Add CLI command to list registered scorers by experiment.
  • Add configuration option for long‑running deployments client requests.
  • Create `set_databricks_monitoring_sql_warehouse_id` API.
  • Show instructions for prompt optimization on prompt registry.

🐛 Bug Fixes

  • Validate that trace is not None before accessing its value in `mlflow.genai.evaluate`.
  • Revert the atomicity changes to the `job_start` API.
  • Make `fastmcp` an optional extra under the `mcp` extra.
  • Fix serialization bug in the file store of the Model Registry.
  • Pin `uvloop` to versions <0.22 to maintain compatibility with mlserver.
  • Fix forward‑compatibility issue with `Span.to_dict`.
  • Whitelist notebook trace UI renderer to allow display with default security settings.
  • Fix attribute error in `StrandsAgent` tracing.
  • Adjust truncation logic in trace previews.
  • Revert the response handling fix in `log_spans`.
  • Adjust utility for remote tracking server declaration.
  • Handle Databricks FMAPI style in OpenAI autolog.
  • Fetch configuration after adding the first record.
  • Fix span ID parsing in the UI.
  • Fix Chat message parsing within the trace summary view modal.
  • Fix display issue of the assessments pane in the UI.

Affected Symbols

mlflow.genai.evaluatejob_startfastmcpModelRegistryFileStoreuvloopSpan.to_dictnotebook trace UI rendererStrandsAgenttrace previews truncation logiclog_spansremote tracking server utilopenai autologfetch_configUI span ID parserUI Chat message parserassessments pane UI componentset_databricks_monitoring_sql_warehouse_id