Fixed
- provider-aware LLM client, cost attribution per provider, multi-provider pricing
Changed
- docs: update eval results to 12/19 (63%), 26% hallucination — post all fixes
- chore: update hello_world_source eval and cost report post-rebase
- chore: sync hello_world_output with post-10B regeneration — 12/19 (63%), 26% hallucination