Bug: Mis-configured RAG reranking model name floods the AI gateway during re-index

[steilerDev]

Summary

An invalid model name (e.g. claude-sonnet-4-7) left in the RAG config can cause tons of requests even on non retry-able errors. Noticed when misconfiguring reranking

[markseibal-png]

This looks like an Anthropic rate-limit problem. npx ai-doctor might catch missing Retry-After handling and unsafe retry logic.

[krissetto]

This looks like an Anthropic rate-limit problem. npx ai-doctor might catch missing Retry-After handling and unsafe retry logic.

This is not an anthropic issue.

We need to detect non retry-able errors and not continue making reranking and embeddings requests for all the chunks