## Purpose

Canonical specification for news-image-relevance-and-fallbacks requirements synced from OpenSpec change deltas.

## Requirements

### Requirement: Image query fallback uses AI-focused defaults
When extracted image keywords are insufficient, the system SHALL use AI-focused default fallback terms.

#### Scenario: Empty keyword extraction
- **WHEN** keyword extraction yields no usable topic keywords
- **THEN** the system uses default fallback terms including AI-domain keywords (for example `ai`, `machine learning`, `deep learning`)

### Requirement: Generic AI image fallback is guaranteed
If provider lookups fail to return a usable summary image, the system SHALL use a generic AI-themed fallback image.

#### Scenario: Provider chain failure
- **WHEN** all configured image providers return no usable image
- **THEN** the system assigns a generic AI fallback image URL/path for summary image

### Requirement: Fallback behavior remains context-aware first
The system SHALL evaluate context-aware candidates before fallback and require refetched summary images to be relevant, non-redundant alternatives.

#### Scenario: Context-aware attempt precedes fallback
- **WHEN** summary image selection runs for a news item
- **THEN** the system first attempts provider queries from extracted context-aware keywords
- **AND** only falls back to generic AI image if candidate evaluation fails

#### Scenario: Finance-story relevance guard
- **WHEN** article topic contains finance/market terms (for example stocks, shares, plunge, earnings)
- **THEN** image selection rejects obviously unrelated animal/portrait outcomes
- **AND** system retries with finance-safe query refinements before final fallback

#### Scenario: Guarded fallback remains deterministic
- **WHEN** provider chain cannot return a relevant finance-safe image
- **THEN** system uses deterministic generic fallback that is topic-safe
- **AND** avoids unrelated imagery classes flagged by guardrails

#### Scenario: Refetch rejects unrelated and duplicate outcomes
- **WHEN** candidate images are evaluated during refetch
- **THEN** the system rejects candidates matching current image identity for the same article
- **AND** rejects clearly unrelated animal/pet imagery classes before final selection