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
docs(feedback): report batch-search candidate smoke (#2334)
Jarvisddb1c75a75dc
diff --git a/user-feedback/16-agent-field-reports/2026-10-24-batch-search-candidate.md b/user-feedback/16-agent-field-reports/2026-10-24-batch-search-candidate.mdnew file mode 100644index 000000000..9dc732b56--- /dev/null+++ b/user-feedback/16-agent-field-reports/2026-10-24-batch-search-candidate.md@@ -0,0 +1,13 @@+# Batch retrieval candidate over synthetic advisories (2026-10-24)++**Agent:** OpenAI Codex · **XERJ:** v1.0.0-rc.98 plus the #1801 batch-search candidate · **Platform:** macOS arm64++**Pointed at:** A fresh, isolated node with three synthetic advisory documents, loaded by REST with an explicit keyword ID field. No real advisory pack or autoindex corpus was evaluated.++**Used it for:** Compare release CLI and MCP batch results, exact-ID versus near-ID behavior, per-item failure handling, budget refusal, and restart persistence.++**Verdict:** The explicit hit/abstain/error separation was usable from both transports, and their batch output agreed. Exact IDs require knowing a keyword field; compatibility with existing real-world corpora remains unverified. This fixture does not establish answer quality, calibrated confidence or performance.++**Numbers:** The release CLI returned hits/abstain/hits for the three-item fixture. Adding an invalid item returned exit 3 while retaining the successful sibling. A 50-item batch with max_tokens=32 refused with exit 1 and no partial JSON; `_count` stayed 3 after refusal and restart. Latency and token savings were not measured.++**Filed alongside:** https://github.com/xerj-org/xerj/pull/2333 (batch retrieval portion of #1801)