You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Breaking — Indexer.IndexBatchAsync no longer drops a document it could not index without telling the caller.
When a document's preparation or embedding failed (an embedding server refusing a long passage, say), the batch logged
it and returned the ids that were written — a caller learned of the gap only by comparing that list with its input. The
batch still writes every other document, then throws BatchIndexingException: IndexedIds (what was written, in input
order) and Failures (each document left out, with its exception). Migration: a caller that tolerates partial batches
catches BatchIndexingException and reads IndexedIds; progress reports and BatchCompleted are unchanged.
Added
BatchIndexingException (see above).
Documentation
The guide's «Basic Search» example says what it is: Retriever.SearchAsync(query, maxResults, minScore) is vector-only
(it was labelled adaptive); hybrid and strategy choice are named next to it. The README says Fusion is on IHybridSearchService results, not on the SDK's SearchAsync results.
docs/QUALITY-BENCHMARKS.md names the public benchmarks for retrieval quality: MTEB(kor, v2) led by MIRACL-ko
(nDCG@10, Recall@100 against the BM25 baseline), MIRACLReranking for the reranker, BEIR for English.
Every C# example in docs/ compiles as its guide assembles it — the design document and the documentation index
are now compiled too. The «Storage Modes» example in docs/README.md declared var ctx twice in one block, so a reader
who copied it got a compile error; the two contexts are now local and production.