Web Search Plus v3.0.2

Web Search Plus v3.0.2 is a focused correctness release for the native v3/Hermes integration.

Fixed

• Restored true multi-provider Research Mode with authoritative per-provider attempts and provenance.
• Preserved the complete Research envelope, truthful timeout/budget degradation, and failed responses when every provider fails.
• Made extraction cache hits lossless for page content, safe raw-content aliases, provider attribution, and routing metadata.
• Keyed extraction cache entries on the full request, attempt budget, effective context limits, and active URL/storage policy.
• Prevented lossy partial-error, image, raw-HTML, and provider-specific payloads from entering the v3 response cache.
• Made retained full-text references immutable and content-versioned, with cache-hit revalidation.
• Applied one deterministic global extraction context budget before cache persistence, operator receipts, and legacy projection.
• Removed stale active Perplexity/Kilo credential and freshness metadata.

Credits

• #104 by @robbyczgw-cla — v3 Research, Extract cache, bounded-context, retained full-text, and operator-receipt repairs.

Verification

• 790 tests passed, plus 6 schema subtests.
• CI passed on Python 3.10, 3.11, and 3.12.
• Provider/routing/contract generators and independent Node schema validation passed.