[AIGNLAI]

Summary

• expand the bundled catalog from 4 illustrative profiles to 130 current OpenRouter models across 23 providers
• record context length, modalities, tool/structured-output/reasoning support, token-price snapshot, and provenance
• derive conservative strength, cost, and latency tiers while clearly labeling heuristics
• calculate estimated_cost from a documented 2,000-input/1,000-output token reference request
• keep zero-setup routing bounded with a 14-model cross-provider default shortlist
• add a deterministic offline profile generator, methodology documentation, and catalog invariant tests
• update README examples and demo history to current catalog IDs

Data provenance

The catalog was generated from the official OpenRouter Models API snapshot at:

https://openrouter.ai/api/v1/models

Snapshot date: 2026-09-20

Official API fields and ReflexRoute-derived heuristics are separated and documented in docs/MODEL_PROFILES.md. Latency levels are family-name heuristics, not measured telemetry. Prices are a snapshot, not live pricing.

Default behavior

Router() now selects 14 cross-provider defaults instead of putting all 130 profiles into one Jev Choice. Any bundled model remains available through an explicit models=[...] list.

Validation

• 18 passed
• deterministic regeneration check passed for all 130 profiles
• 130 IDs verified against a 446-model official snapshot
• 23 providers represented
• wheel build and isolated installation passed
• installed package successfully loaded all profiles and bundled defaults
• documentation links, credential scan, and Git whitespace checks passed