Module feature_encoder 

Encode extraction results into 128-float feature vectors.

Missing data strategy

• 0.0 means “absent/unknown” for all optional features.
• NodeFlags bits distinguish “value is genuinely zero” from “unknown”:
  • features[48] = 0.0 + HAS_PRICE → price is genuinely $0 (free)
  • features[48] = 0.0 + no HAS_PRICE → price not found on page
• Non-USD currency: stored raw in original currency. No conversion.
• Price ranges (“$200–$350”): low end stored in features[48].
• Text ratings (“Excellent”): mapped to numeric 0.0–1.0.

Structs

FeatureEncodeResult

Result of feature encoding, including computed flags.

Functions

encode_features

Encode extraction results + navigation info into a 128-float feature vector.

encode_features_from_patterns

Encode features from pattern-based extraction (Layer 1.5) without browser rendering.

encode_features_from_structured_data

Encode features from structured data (Layer 1) without browser rendering.

encode_features_with_flags

Encode extraction results and also return computed NodeFlags.

merge_features

Merge features from multiple sources, picking the highest-confidence value per dimension.