explainTalentMatch's /explain endpoints 503'd because defaultRagAsk had no
real LLM client behind it. talentGraphLlmClient.ts tries Railyard Relay
(OpenAI-compatible, same tensorzero::model_name::anthropic:: pattern
bidstage-tenders already uses) first, falling back to a direct OpenRouter
call on any Relay failure or when Relay isn't configured.
Verified against real endpoints, not mocks: a real call through Relay
(claude-haiku-4-5-20251001) returned a grounded answer citing all 3 real
subgraph edges with zero hallucinated citations; a deliberately broken
Relay key correctly triggered the OpenRouter fallback path (confirmed via
a real 401 -> real fallback attempt -> real OpenRouter 402, no credits).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>