OPERA: Open-Source Prediction Evaluation and Real-Time Analysis

Documentation | Add your own model

OPERA is a modular ecosystem designed to support the use of time-series models in real-time settings and foster open collaboration. It comprises the following modules:

• opera-eco pins compatible releases and supplies documentation, AI skills, and integration tests.
• forecast_evaluation validates vintaged outturns and forecasts and provides evaluation and visualisation capabilities.
• forecast_realtime runs models across data vintages.
• forecast_combo combines forecasts through averaging, regression, error-based weighting, or hierarchies.
• news_decomp attributes nowcast levels and revisions to news, re-estimation, and interaction.

OPERA supports a broad range of models through wrappers for libraries such as scikit-learn and R's fable. You can also add your own model. In addition to these wrappers, OPERA includes models with native support:

• bvar provides tools for working with Bayesian VARs.
• nowcast-midas nowcasts quarterly targets from higher-frequency indicators using MIDAS and combination techniques.

Architecture

Quick Start

pip install opera-eco              # Install the CLI and skills only.
pip install "opera-eco[modules]"  # Install the CLI and all ecosystem packages.
pip install "opera-eco[notebooks]" # Install Marimo notebook tooling.
opera install skills               # Install the bundled AI skills.

Then ask Copilot or Claude to use an installed skill:

Tell me about @opera and how I can use it with my model.

Project Layout

docs/                            # Documentation site.
examples/illustration.py        # Runnable end-to-end example.
examples/illustration_marimo.py # Native Marimo version of the example.
src/opera/                      # Package source, bundled skills, and tests.
  cli.py                        # Command-line interface.
  skills_manager.py             # Skill discovery and installation.
  skills/                       # Bundled Markdown skill files.
pyproject.toml                  # Python package configuration.
zensical.toml                   # Documentation site configuration.

Data Classification

Bank of England Data Classification: OFFICIAL BLUE