Metadata-Version: 2.5
Name: sfhybrid
Version: 0.0.1
Summary: Spectral fractal hybrid: agent + modelers + FX layer for multiscale, multi-series, multimodal probabilistic forecasting (placeholder; under development)
Author: Junyan Cheng
License-Expression: Apache-2.0
License-File: LICENSE
Keywords: forecasting,llm-agents,neurosymbolic,probabilistic,spatiotemporal
Classifier: Development Status :: 1 - Planning
Classifier: Intended Audience :: Science/Research
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Description-Content-Type: text/markdown

# sfhybrid

**Placeholder release. The library is under development.**

sfhybrid will package Chorochronos, a spatiotemporal neurosymbolic forecaster with three layers:

- **Slow layer:** global signals learned statistically by default. Optionally, an LLM agent reads text sources and
  adds its own signals, which are kept only if they improve held-out likelihood.
- **Modelers:** configurable statistical, ML and neural models, one per frequency band.
- **FX layer (fractal hybrid):** learns how to mix the components into a joint predictive distribution over many
  series, time scales and hierarchy levels.

The goal is an estimator as easy to use as XGBoost or LightGBM, where the output is a full predictive distribution.
