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
chore(autoindex): drop the stale #332 KNOWN LIMITATION notes in bvh/unity extractors
xerj-org980340486a2b
diff --git a/engine/crates/xerj-autoindex/src/extract/bvh.rs b/engine/crates/xerj-autoindex/src/extract/bvh.rsindex 86be07a93..fe72f07d1 100644--- a/engine/crates/xerj-autoindex/src/extract/bvh.rs+++ b/engine/crates/xerj-autoindex/src/extract/bvh.rs@@ -12,16 +12,10 @@//! `frames`, `frame_time_s`, `duration_s`. All extractor vocabulary//! (`FieldOrigin::Extractor`), so every BVH file clusters into one dataset.//!-//! KNOWN LIMITATION (engine-side, xerj-org/xerj#332): `joints` is a-//! multi-valued keyword array, and the FTS writer currently joins array-//! elements into ONE space-separated token. So after a segment flush,-//! `{"term": {"joints": "LeftHand"}}` does NOT match — there is no-//! per-element posting. `terms` and other clauses that take the stored-scan-//! path still match, and so does everything before the flush, which is why-//! this is easy to miss. "Which clips animate LeftHand?" is only a reliable-//! one-query answer once #332 lands; it is an engine API change (the FTS-//! writer input is `HashMap<String, String>`), not something this extractor-//! can work around.+//! `joints` is a multi-valued keyword array. Each element is indexed as its+//! own term (xerj-org/xerj#332, fixed in #470), so `{"term": {"joints":+//! "LeftHand"}}` answers "which clips animate LeftHand?" both before and+//! after a segment flush.use super::{ExtractStats, FieldOrigin, RawRecord, Sink};use anyhow::Result;diff --git a/engine/crates/xerj-autoindex/src/extract/unity.rs b/engine/crates/xerj-autoindex/src/extract/unity.rsindex 480d8c00c..aba2e617e 100644--- a/engine/crates/xerj-autoindex/src/extract/unity.rs+++ b/engine/crates/xerj-autoindex/src/extract/unity.rs@@ -27,12 +27,11 @@//! - `script_guid` the `m_Script` reference's guid (MonoBehaviour → .cs)//! - flattened body fields//!-//! KNOWN LIMITATION (engine-side, xerj-org/xerj#332): `ref_guids` is a-//! multi-valued keyword array and the FTS writer joins array elements into-//! ONE space-separated token, so a post-flush `term`/`match` on a single guid-//! does not match. `script_guid`, `script_path` and `script_class` are-//! single-valued and unaffected — the documented "which scenes use this-//! script?" query goes through those, not through `ref_guids`.+//! `ref_guids` is a multi-valued keyword array; each guid is indexed as its+//! own term (xerj-org/xerj#332, fixed in #470), so a `term`/`match` on a+//! single guid matches before and after a flush. The documented "which+//! scenes use this script?" query still goes through the single-valued+//! `script_guid`, `script_path` and `script_class`.//!//! `.meta` sidecars emit ONE record: `guid`, `asset_name`, `importer`, and//! flattened settings. The root-relative `asset_path` (and the