Metadata-Version: 2.4
Name: maestro-economics
Version: 0.8.3
Summary: CLI and SDK for RA Data platform — datasets, compute, and experiments.
Author-email: Maestro AI <hello@maestro.onl>
License: MIT
Keywords: cli,economics,experiments,panel-data,research-data
Classifier: Development Status :: 4 - Beta
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Information Analysis
Requires-Python: >=3.11
Requires-Dist: click>=8.1
Requires-Dist: httpx>=0.27
Requires-Dist: tomli>=2.0; python_version < '3.11'
Provides-Extra: data
Requires-Dist: duckdb>=1.0; extra == 'data'
Requires-Dist: pandas>=2.2; extra == 'data'
Requires-Dist: pyarrow>=16.0; extra == 'data'
Description-Content-Type: text/markdown

# maestro-economics

CLI and SDK for the RA Data platform — datasets, compute, and experiments.

```bash
pip install maestro-economics
mecon --help
mecon resources
mecon submit . --resource cpu-4c-16gb
```

CPU runs support two modes:

```bash
# Lower-cost default; may restart, so long jobs must resume from durable progress.
mecon submit . --resource cpu-16c-32gb --mode standard

# Fewer infrastructure interruptions; CPU only, billed at the Stable rate.
mecon submit . --resource cpu-16c-64gb --mode stable
```

For CPU timeouts above four hours, save progress during the run, load it at
startup, and confirm the contract at submission:

```python
def run(ctx):
    state = ctx.load_checkpoint() or {"step": 0}
    for step in range(int(state["step"]), 100):
        # compute one restart-safe unit
        ctx.save_checkpoint(step=step + 1)
    return {"completed_steps": 100}
```

```bash
mecon submit . --resource cpu-16c-32gb --timeout 21600 --durable-resume
```

`ctx.output_dir` is also durable. Stable CPU reduces interruptions but does not
replace checkpoint/resume logic for long jobs.
