Automated Pipelines & Reports¶

Dyno includes an automated execution and reporting engine that compiles model verification, perturbation solving, moment evaluation, and simulation into reproducible output reports.

The `@run:` Directive¶

Declare automated pipeline steps directly within your model file:

# model.dyno
@name: RBC_Report_Example
@run: steady
@run: check
@run: solve
@run: simul: {T: 40}

# Parameters & equations follow...
alpha <- 0.36
...

The @run: command format is stable: models written with it will keep working in future releases.

Commands are executed in the order they appear. Each @run: line holds one command, written as:

Form	Example
Bare command name	`@run: steady`
Command with options (a YAML mapping)	`@run: simul: {T: 40}`
Command with null options	`@run: steady: null`

Appending ; to the end of a command line (e.g. @run: solve; or @run: simul: {T: 40};) mutes that command's output in reports. Semicolons are only permitted at the end of the line (for instance, @run: simul:; {T: 40} is invalid and raises an error). In YAML models, the same commands are listed under a top-level run: key, and the explicit form {command: simul, options: {T: 40}} is also accepted.

Supported @run: commands:

Command	Action
`steady`	Solves for steady state numerically if needed (`model.steady(**options)`)
`resid`	Computes steady-state residuals
`check`	Verifies residuals and checks Blanchard-Kahn eigenvalues
`solve` / `perturb`	Computes first-order perturbation policy function
`simul` / `simulate`	Generates stochastic simulations, or a perfect-foresight path for deterministic models (options such as `T`, `N`, `mode`, `units`)
`analyze` / `stoch_simul`	Solves the model, computes moments and impulse responses (options `T`/`irf`, `type`/`units`, `variables`, `nograph`)
`plot`	Plots the current simulation (options such as `variables`, `engine`)
`variants`	Runs the remaining commands on each calibration variant (e.g. `@run: variants: {beta: [0.95, 0.99]}`)

Executing Pipelines with `model.run()`¶

In Python, execute the pipeline with one call:

from dyno import DynoModel

model = DynoModel("examples/RBC.dyno")
results = model.run()

Inspecting `RunResults`¶

The returned RunResults object encapsulates the complete execution state:

# Access components:
results.model           # Model after steady-state evaluation
results.residuals       # Equation residual vector
results.solution        # PerturbationSolution decision rule
results.eigenvalues     # Generalized eigenvalues array
results.bk_check        # True if Blanchard-Kahn conditions hold
results.moments         # Asymptotic covariance matrices
results.simulation      # Dictionary of IRFs or simulation DataFrame

Rich Display in Notebooks & Terminal¶

RunResults automatically renders clean, formatted summaries across all environments:

Terminal (Plain Text)¶

Printing the result outputs a formatted diagnostic summary:

print(results)

RunResults
==========
Model
-----
name: RBC_Report_Example
filename: model.dyno
symbols: variables=8, endogenous=8, exogenous=1, parameters=7

Steady State
------------
residuals: max|res|=0.000e+00, mean|res|=0.000e+00

Solution
--------
method: qz
Eigenvalues: computed (n=8, |lambda|>1: 4, |lambda|≈1: 0)
Blanchard-Kahn conditions: met

Simulation
----------
Simulation: computed (IRFs, shocks=1, horizon=40)
  shocks: e_a

Simulation Plots
----------------
[ASCII / terminal graphs rendered via plotext]

You can also customize or extract terminal graphs directly:

# Print or retrieve text plots with plotext:
print(results.plot_text(cols=2, width=80))

# Control graph inclusion in text reports:
print(results.to_text(graphs=False))

Jupyter Notebooks (HTML & Markdown)¶

In JupyterLab or VSCode Notebooks, simply returning results renders interactive HTML tables, equation renderings, and embedded charts:

# In a Jupyter cell:
results

Programmatic Report Generation & Export¶

RunResults can be directly exported to HTML or Markdown strings:

html_content = results.to_html()
markdown_content = results.to_markdown()

# Save report to disk
with open("model_report.html", "w") as f:
    f.write(html_content)