AI search engine and MCP server for bots and agents. One Rust binary indexes code, docs, logs, JSONL/CSV and PDFs for BM25 full-text, vector and hybrid search, code search without grep, agent memory and RAG. Elasticsearch-compatible API; Corpus Hub serves hundreds of reference memory corpora.
Rust agent-memoryai-agentsbm25claude-codeclicode-searchcursordatabaseelasticsearchfull-text-searchhybrid-searchjsonllocal-firstmcpmcp-servermodel-context-protocolragrustsearch-enginevector-search
perf(query): admit highlight into the scored-columnar plan (highlight cell LOSS→WIN, bool/multi_match+highlight scores → ES-exact)
xerj-orgff855f9f3a14
diff --git a/engine/crates/xerj-engine/src/index.rs b/engine/crates/xerj-engine/src/index.rsindex 24c0917d2..8f68ecd24 100644--- a/engine/crates/xerj-engine/src/index.rs+++ b/engine/crates/xerj-engine/src/index.rs@@ -21109,7 +21109,14 @@ fn scored_fast_plan(numeric_fields: &HashSet<String>,boolean_fields: &HashSet<String>,) -> Option<ScoredPlan> {- // Request-level plainness — identical to `function_score_fast_fvf`.+ // Request-level plainness — like `function_score_fast_fvf`, except+ // `highlight` is admitted: highlighting is pure page post-processing+ // (`apply_highlight(page, hl_req, query)` runs on the final rendered+ // page regardless of which executor produced it, and reads only the+ // hit `_source` + the query AST — no scan-time state). Bailing on it+ // dropped e.g. `match(status)+highlight` to the brute FTS walk over+ // every match (~2.5× ES); the columnar page is byte-identical, so the+ // shared highlighter emits byte-identical fragments.if size == 0|| request.aggs.is_some()|| request.min_score.is_some()@@ -21117,7 +21124,6 @@ fn scored_fast_plan(|| !request.rescore.is_empty()|| request.collapse.is_some()|| request.search_after.is_some()- || request.highlight.is_some()|| request.explain|| request.script_fields.is_some()|| !request.fields.is_empty()