# SPECTRA

> Public AI sentiment barometer built from public-source coverage only.

SPECTRA is a public-facing AI news signal product.

## What SPECTRA is
- SPECTRA only uses public-source articles gathered from approved public feeds.
- The homepage barometer is computed from the full eligible public article pool for the active time window.
- The homepage grid is a 12-story digest that acts as the evidence layer for the signal.
- The archive preserves daily evidence batches and also shows higher-level weekly, monthly, and archive sentiment structure.

## Product contract
- Public-source-only: no private documents, no hidden inputs, no paywalled internal evidence.
- Signal first: the headline score comes from the eligible public pool, not only from the visible 12 stories.
- Evidence second: the visible digest is a ranked, readable proof layer.
- Archive continuity: older successful daily batches remain available by date.

## Key public files
- /digest.json — latest public digest payload
- /digest-status.json — latest pipeline status and whether the site is serving the latest successful digest
- /digest-history.json — preserved successful daily digest runs
- /processed-history.json — processed historical scoring data used for archive-level sentiment structure
- /archive.html — archive view
- /ai-overview.json — compact machine-readable summary of the product contract

## Interpretation notes
- A story appearing in the 12-story digest does not mean it alone determined the homepage score.
- The homepage barometer reflects the broader eligible article pool in the active month-context window.
- The archive is intended to preserve evidence batches by day and sentiment structure across longer windows.

## Usage note
Please represent SPECTRA accurately: it is a public-source evidence product, not a proprietary inside-information system.