Unverified50% confidenceFactExact time
When the provider layer throws an exception, schemagate's reranker catches the exception by design and returns the mathematical ranking result (fallback), causing the output the user sees to be exactly the same as when --rerank is not enabled.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/16/2026
First Seen
Valid until: 1/14/2027
Sources
Related Entities
Related Claims
UnverifiedThe core logic of a multi-provider failover mechanism is: when the primary provider returns 429 or other retryable errors, automatically switch to a backup provider instead of letting the request fail outright.66% similarUnverifiedAfter the fix, an events_dropped counter was added to the read-back response, making event eviction explicit and visible data.65% similarUnverified当底层 provider 协议调整（如消息 role 字段取值、tool call args 是否流式分块返回），上层 schema 若不同步更新会导致类型与真实数据错位，这类错误不会在编译期报出而是运行时触发校验失败64% similarVerifiedReranking technology first uses coarse retrieval to recall a large number of candidate segments, then uses a fine-ranking model to re-score them, keeping only the most relevant content to feed into the model.64% similarUnverifiedWhen re-delivering after manual correction on the error branch, if the failure point occurred after the CRM had already written, directly re-running the entire lead may create duplicate records, so idempotency design is needed (such as a unique identifier plus upsert or deduplication).63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1049513API
curl https://kongchang.com/api/v1/knowledge/claims/1049513MCP
get_claim(id=1049513)