ForeGlass makes experimental forecasts and evidence available for inspection. The foreglass Python library helps developers read supported public artifacts, preserve their original meaning and retain the bytes used by an application.
Our mission is to help people and organizations examine a forecast before relying on it, and revisit it when outcomes become known. This reader is one small, inspectable part of that work.
Read and retain
Replay a synthetic fixture offline, or explicitly fetch supported public artifacts.
PYTHON REFERENCEIntegrate the client
Bounded reads, snapshots, catalogs and normalization.
Two libraries with different jobs
| Library | Role | Documentation |
|---|---|---|
foreglass |
Optional client for supported ForeGlass public artifacts | Reader guide |
comuvia |
Provider-neutral records, integrity checks, corrections and binary evaluation | Core documentation |
The client depends on comuvia. The core works offline and does not depend on ForeGlass or upload your records.
Start with the evidence, not a score
A file build time is not necessarily a forecast's information cutoff. A market fragility index is not a recession probability. A conditional model path is not an unconditional forecast. The client preserves these distinctions rather than inventing missing metadata.
Read the public ledger and Verification Explorer. A hash commitment can help detect changes against a retained expectation; it does not prove a prediction correct or establish a complete performance record. See record integrity and timing limits.
Availability
foreglass 0.1.0 and its dependency comuvia 0.1.0 are published on PyPI (released 2026-10-01) from the public repository github.com/comuvia/comuvia-sdk, under the Apache License 2.0. Install with pip install "foreglass==0.1.0". The release status page lists the approved file digests and how to verify a download before installing; the core quickstart explains evaluation. 0.1.x is an alpha: pin the version.
These packages do not provide a hosted MCP server, custom forecast generation, a submission service or access to proprietary models. Ask about integration or report a minimal reproducible issue on GitHub issues; report security problems privately as the repository's SECURITY.md describes. Do not include private positions, credentials or restricted source material.