LoginSign Up
← Back to all sparks
M

MLflow

AI-ML-PLATFORMS
Velocity1.6

Open source platform for managing the machine learning lifecycle

mlflow.org ↗

MLflow 3.17.2 records dataset, scorer and agent lineage on every GenAI evaluation run.

llm-observabilityevaluationlineageaccess-controlmlops
◆Current state
The 3.x line centers on LLM tracing and evaluation rather than classic experiment tracking. 3.17 added typed judge feedback, resource-level permissions with explicit DENY and a mandatory schema upgrade; 3.17.1 brought Decision API support in make_judge() and Databricks review-queue management. 3.17.2 records which dataset, scorers and agent each mlflow.genai.evaluate run used, and fixes Optuna integration bugs.
◆Where it's heading
Evaluation is becoming an auditable record rather than a one-off score: typed decisions, review queues and now lineage tie each result back to its inputs. Server hardening for shared deployments continues alongside, and the legacy 2.11 line still gets Databricks-specific patches.
◆Prediction
Expect lineage to surface in the UI for comparing evaluation runs and more human-review tooling in open-source MLflow. The entries don't show a 3.18 date.

◆Recent moves

  1. 8d ago

    3.17.2 records dataset, scorer and agent lineage on evaluate runs

    A patch that adds one meaningful feature: evaluation runs now record the dataset, scorers and agent they used, so results can be traced back to inputs. It continues the 3.17 push toward structured, reviewable evaluation; the Optuna fixes are routine.

    View source ↗
  2. 16d ago

    MLflow 3.17.1

    A patch that extends 3.17's typed-feedback work: make_judge() now supports the Decision API, and review queues on Databricks become manageable from the SDK. The Claude Code OTel rendering fix shows coding-agent traces are now a use case the UI has to handle.

    View source ↗
  3. 16d ago

    Weekly model catalog refresh (CI)

    The CI job that refreshes the per-provider model catalog files ran its weekly update. Useful plumbing for the AI Gateway's model lists, but not a product release.

    View source ↗
  4. 22d ago

    MLflow 3.17.0

    3.17 adds typed Boolean and categorical judge feedback, resource-level permissions with wildcard grants and explicit DENY, and opt-in daily rollups for trace analytics. It continues the shared-server hardening from 3.16, but it brings a mandatory schema upgrade with every writer stopped.

    View source ↗
  5. 1mo ago

    MLflow 2.11.5

    A 2.x maintenance patch adding opt-in Databricks SDK Files API routing for Unity Catalog model artifacts. It shows the legacy line is still serviced for Databricks users while new work lands in 3.x.

    View source ↗
  6. 1mo ago

    MLflow 3.16.1

    Mostly fixes, plus a security-relevant breaking change: the default basic-auth admin password is no longer shipped. Scorer timeouts and Unity Catalog span links extend the 3.16 tracing work.

    View source ↗