Features
Log multimodal, multi-rate data with Python, C++, or Rust SDK
Interactive viewer with 2D, 3D, map, graph, and tensor views
SQL and dataframe queries over recordings and catalog
Transform data with derived columns and schema evolution
Train directly via PyTorch dataloader from .rrd files or Hub streams
Store data as column-chunks in .rrd files for efficiency
Declarative visualization framework with blueprint defaults and overrides
Byte-range indexing and retrieval from object storage (Hub)
Team collaboration with shared recordings and link sharing (Hub)
Open source SDK under Apache-2.0/MIT license
Web viewer for browser-based visualization
Layers and local catalog for organizing recordings
Support for time-series, 3D, and tensor data modalities
Trace visibility for agent steps (prompts, retrievals, tool calls, outputs)
LLM-as-judge evaluation for relevance, toxicity, quality scoring
Dataset creation from traces for reproducible testing
Experiment management and regression benchmarking
Built-in Prompt IDE for iterative prompt optimization
Self-hosted deployment on local, Docker, Kubernetes
Phoenix Cloud managed hosting option
Vendor-agnostic support for any model/framework
Native OpenTelemetry integration
OpenInference specification for LLM telemetry
Human annotation and automated labeling
Ghost trajectories to simulate alternative agent paths
Eval-as-you-test for early quality feedback
One-click integration with LlamaIndex
