Skip to content

armgit / xerj

Public Imported from GitHub
0

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.

docs: field report for Windows autoindex and ES-compat search (2026-10-18) (#1763)

Roedy Rustame4ee65a8ecd2
diff --git a/user-feedback/16-agent-field-reports/2026-10-18-windows-autoindex-search.md b/user-feedback/16-agent-field-reports/2026-10-18-windows-autoindex-search.md
new file mode 100644
index 000000000..3bfd2b4b5
--- /dev/null
+++ b/user-feedback/16-agent-field-reports/2026-10-18-windows-autoindex-search.md
@@ -0,0 +1,13 @@
+# Windows autoindex + ES-compat query evaluation (2026-10-18)
+
+**Agent:** Antigravity (Gemini 3.8 Flash, Google DeepMind) · **XERJ:** xerj v1.0.0-rc.96 · **Platform:** Windows 11 x86_64
+
+**Pointed at:** `content/` Markdown corpus under XERJ repo (92 files, 895 KB of prose and comparison docs).
+
+**Used it for:** Verified Windows native install via `get.ps1`, executed `xerj autoindex` with `--prefix test-content`, generated `xerj autoindex map`, and queried HTTP ES REST `_search`.
+
+**Verdict:** The single static binary experience on Windows is excellent: zero dependencies, fast startup, and `autoindex map` generated syntactically accurate ES queries out of the box. BM25 hit retrieval was instant and returned properly scoped passages without manual schema definitions. Minor friction: `get.ps1` failed when `api.github.com` DNS timed out, requiring explicit `$env:XERJ_VERSION='v1.0.0-rc.96'` to bypass the API lookup, and Windows logs warn that total RAM cannot be read so worker counts fall back to core-derived defaults.
+
+**Numbers:** `xerj autoindex`: 91 files / 390 source records -> 481 live docs in 4.5s (107.2 docs/sec, wall 9.9s); BM25 match query returned 20 hits in `test-content-docs`.
+
+**Filed alongside:** nothing broke