Getting Started¶
Installation¶
Full ecosystem¶
Install opera-eco with every OPERA module dependency:
This installs the opera package itself plus:
| Package | Role |
|---|---|
forecast_evaluation |
Validate data, evaluate accuracy, visualise |
forecast_realtime |
Real-time fit/forecast loops and backtesting |
forecast_combo |
Forecast combination |
bvar |
Bayesian VAR model library |
nowcast-midas |
MIDAS and mixed-frequency model library |
news_decomp |
Nowcast news decomposition |
Skills only¶
Install the base package when you need assistant skills but not the Python modules:
Installing the Skills¶
OPERA ships skills for Copilot and Claude. Install them with the opera CLI:
# Install all skills and detect `.github/skills` or `.claude/skills` automatically.
opera install skills
Available skills:
| Name | Covers |
|---|---|
opera |
Full ecosystem architecture and module interactions |
forecast-evaluation |
Evaluating forecast accuracy |
forecast-realtime |
Real-time forecasting and backtesting |
forecast-combo |
Forecast combination methods |
bvar |
Bayesian VAR estimation |
nowcast-midas |
MIDAS and mixed-frequency nowcasting |
forecast-decomp |
Nowcast news decomposition |
Using the Skills with an LLM¶
After installation, Copilot reads skills from .github/skills/ and Claude reads them from .claude/skills/. Ask questions such as:
- "How to backtest my model using OPERA?"
- "What combination methods are available in
forecast_combo?"
Next Steps¶
Read the example pipeline or run the Marimo notebook.