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.

seo: add contextual internal links to 14 unindexed pages (#2804)

xerj-orgd7cd0957e520
diff --git a/content/answers/batch-lookup-cve-ghsa-ids-offline.md b/content/answers/batch-lookup-cve-ghsa-ids-offline.md
index 53a424a06..5ee07dbaa 100644
--- a/content/answers/batch-lookup-cve-ghsa-ids-offline.md
+++ b/content/answers/batch-lookup-cve-ghsa-ids-offline.md
@@ -7,7 +7,7 @@ cluster: "Answers for coding agents"
question: "How do I batch lookup CVE and GHSA IDs offline without hitting GitHub API rate limits?"
intent: "how-to"
published: "2026-10-19"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -43,6 +43,7 @@ links_out:
- "/blog/one-vulnerability-three-names"
- "/blog/the-exploit-hub"
- "/compare/xerj-vs-elasticsearch-local"
+ - "/case-studies/wordpress-security-audit"
faq:
- q: "How do I batch lookup CVE IDs without the GitHub API?"
a: "Download the signed rust-vulns pack once, load it with `xerj corpus add` and `xerj corpus index`, then send all IDs in one `terms` query on `id` and `aliases`. Nothing calls the GitHub API after the download."
diff --git a/content/answers/code-search-mcp-for-claude-code.md b/content/answers/code-search-mcp-for-claude-code.md
index c64720dea..c60de0bfc 100644
--- a/content/answers/code-search-mcp-for-claude-code.md
+++ b/content/answers/code-search-mcp-for-claude-code.md
@@ -7,7 +7,7 @@ cluster: "Agent access and memory"
question: "I need my agent to find a passage and cite the file, not dump the whole tree. How?"
intent: "how-to"
published: "2026-08-21"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -37,6 +37,9 @@ links_out:
- "/compare/xerj-vs-ast-grep"
- "local-code-search-mcp-fewer-tokens-than-grep"
- "ask-questions-about-a-github-repo-without-cloning"
+ - "/answers/claude-code-memory"
+ - "/answers/what-is-an-mcp-memory-server"
+ - "/compare/xerj-vs-claude-mem-and-mempalace"
faq:
- q: "I need my agent to find a passage and cite the file, not dump the whole tree. How?"
a: "Register `xerj mcp` and let the agent call `xerj_search`. Each hit carries a `symbols` array with `name`, `kind` and `line`, so the answer can cite the declaration site."
diff --git a/content/answers/coding-agent-memory-across-sessions.md b/content/answers/coding-agent-memory-across-sessions.md
index 5523c8399..b21404b5e 100644
--- a/content/answers/coding-agent-memory-across-sessions.md
+++ b/content/answers/coding-agent-memory-across-sessions.md
@@ -7,7 +7,7 @@ cluster: "Agent memory: setup"
question: "How do people store user preferences so an agent can recall them next week?"
intent: "how-to"
published: "2026-08-21"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -35,6 +35,9 @@ links_out:
- "/docs/recipes/agentic-memory"
- "what-is-agentic-memory"
- "what-is-an-mcp-memory-server"
+ - "/compare/xerj-vs-agent-memory-tools"
+ - "/compare/xerj-vs-mem0"
+ - "/answers/claude-code-memory"
faq:
- q: "How do people store user preferences so an agent can recall them next week?"
a: "Write each preference with `POST /_memory/{namespace}` and read it back with `POST /_memory/{namespace}/_recall`. The record lives in a reserved index in the data directory and survives a full node restart."
diff --git a/content/answers/do-search-embeddings-help.md b/content/answers/do-search-embeddings-help.md
index 3e8330a4e..8ac73c00d 100644
--- a/content/answers/do-search-embeddings-help.md
+++ b/content/answers/do-search-embeddings-help.md
@@ -7,7 +7,7 @@ cluster: "Embeddings: selection"
question: "Do I even need embeddings for search, or is full-text enough?"
intent: "informational"
published: "2026-08-21"
-updated: "2026-10-19"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -35,6 +35,8 @@ links_out:
- "xerj-vs-vector-database"
- "vector-database-vs-full-text-search"
- "reduce-embedding-api-cost"
+ - "/answers/bm25-vs-vector-search"
+ - "/answers/classify-text-from-labelled-history-without-llm"
faq:
- q: "Do I even need embeddings for search, or is full-text enough?"
a: "Full-text is enough until the queries stop sharing tokens with the files. XERJ's default embedder is lexical feature hashing, and on our judged set it never recovered a synonym BM25 missed."
diff --git a/content/answers/full-text-search-over-jsonl-files-cli.md b/content/answers/full-text-search-over-jsonl-files-cli.md
index c289feee8..2594313cb 100644
--- a/content/answers/full-text-search-over-jsonl-files-cli.md
+++ b/content/answers/full-text-search-over-jsonl-files-cli.md
@@ -7,7 +7,7 @@ cluster: "Answers for coding agents"
question: "How do I run full-text search over JSONL files from the CLI and query a folder with BM25?"
intent: "how-to"
published: "2026-10-19"
-updated: "2026-10-20"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -42,6 +42,7 @@ links_out:
- "batch-lookup-cve-ghsa-ids-offline"
- "/compare/xerj-vs-elasticsearch-local"
- "/compare/xerj-vs-vector-database"
+ - "/compare/best-local-full-text-search-jsonl-csv-2026"
faq:
- q: "How do I run full-text search over JSONL files from the CLI?"
a: "Run `xerj autoindex` on the folder, then query the node with curl or `xerj search`. No mapping is needed: each record becomes a document with its own fields."
diff --git a/content/answers/how-xerj-combines-search.md b/content/answers/how-xerj-combines-search.md
index e02465f40..8375989d7 100644
--- a/content/answers/how-xerj-combines-search.md
+++ b/content/answers/how-xerj-combines-search.md
@@ -7,7 +7,7 @@ cluster: "Hybrid search"
question: "How does XERJ combine BM25 and kNN?"
intent: "informational"
published: "2026-08-21"
-updated: "2026-10-19"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -36,6 +36,8 @@ links_out:
- "what-is-xerj"
- "/docs/recipes/hybrid-search"
- "/docs/playbooks/vector-search"
+ - "/answers/what-is-hybrid-search"
+ - "/answers/bm25-vs-vector-search"
faq:
- q: "How does XERJ combine BM25 and kNN?"
a: "It sends both as sub-queries inside one `hybrid` clause against one index, then merges the two ranked lists with Reciprocal Rank Fusion. One request, one response, one result list."
diff --git a/content/answers/hybrid-search-quality-measured-ndcg.md b/content/answers/hybrid-search-quality-measured-ndcg.md
index b8776126e..deeb3cc38 100644
--- a/content/answers/hybrid-search-quality-measured-ndcg.md
+++ b/content/answers/hybrid-search-quality-measured-ndcg.md
@@ -7,7 +7,7 @@ cluster: "Hybrid retrieval: quality"
question: "How much better is hybrid search than BM25 alone?"
intent: "informational"
published: "2026-09-18"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -35,6 +35,10 @@ links_out:
- "local-embeddings-without-openai-api"
- "retrieval-that-says-not-found-with-file-line-citations"
- "/answers/improve-basic-faiss-rag-pipeline"
+ - "/answers/what-is-hybrid-search"
+ - "/answers/bm25-vs-vector-search"
+ - "/answers/classify-text-from-labelled-history-without-llm"
+ - "/blog/the-rc80-gates"
evidence:
- claim: "On BEIR SciFact (300 test queries, 5,183 documents) nDCG@10 was 0.6572 for BM25, 0.6764 for MiniLM vectors only, 0.6855 for a BM25 top-30 shortlist reordered by MiniLM, and 0.6993 for hybrid RRF."
source: "benchmarks/beir-hybrid/results/scifact.txt"
diff --git a/content/answers/improve-basic-faiss-rag-pipeline.md b/content/answers/improve-basic-faiss-rag-pipeline.md
index be81bc1f6..aee96a143 100644
--- a/content/answers/improve-basic-faiss-rag-pipeline.md
+++ b/content/answers/improve-basic-faiss-rag-pipeline.md
@@ -7,7 +7,7 @@ cluster: "RAG: troubleshooting"
question: "How can I improve a basic FAISS RAG pipeline?"
intent: "troubleshooting"
published: "2026-08-21"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -32,6 +32,9 @@ links_out:
- "rag-without-vector-database"
- "reciprocal-rank-fusion-when-to-use"
- "vector-database-vs-full-text-search"
+ - "/answers/agentic-rag-vs-rag"
+ - "/answers/what-is-hybrid-search"
+ - "/answers/agentic-memory-vs-rag"
faq:
- q: "Why is my FAISS RAG pipeline missing obvious answers?"
a: "A vector-only pipeline has no term signal. On our judged set BM25 found literal tokens that the vector clause missed, and fusion recovered both."
diff --git a/content/answers/rag-without-vector-database.md b/content/answers/rag-without-vector-database.md
index bac881ad3..7a971b8ce 100644
--- a/content/answers/rag-without-vector-database.md
+++ b/content/answers/rag-without-vector-database.md
@@ -7,7 +7,7 @@ cluster: "Hybrid retrieval: architecture"
question: "How do I build RAG without a vector database?"
intent: "how-to"
published: "2026-08-21"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -35,6 +35,8 @@ links_out:
- "/docs/recipes/hybrid-search"
- "retrieval-that-says-not-found-with-file-line-citations"
- "agentic-memory-vs-rag"
+ - "/answers/agentic-rag-vs-rag"
+ - "/answers/what-is-hybrid-search"
faq:
- q: "Does XERJ need a separate vector database?"
a: "No. XERJ stores the dense vector beside the inverted index in one index, and one `_search` request queries both. Our capture ran with no second service."
diff --git a/content/answers/reduce-embedding-api-cost.md b/content/answers/reduce-embedding-api-cost.md
index 135062eff..5f727a790 100644
--- a/content/answers/reduce-embedding-api-cost.md
+++ b/content/answers/reduce-embedding-api-cost.md
@@ -7,7 +7,7 @@ cluster: "Embeddings: cost"
question: "My embedding bill is too high"
intent: "cost"
published: "2026-08-21"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -31,6 +31,7 @@ links_out:
- "do-search-embeddings-help"
- "local-embeddings-without-openai-api"
- "xerj-vs-vector-database"
+ - "/answers/classify-text-from-labelled-history-without-llm"
faq:
- q: "Does XERJ call an external embedding API?"
a: "No, not by default. A network sampler watched the default node for its whole life and observed 0 non-loopback peers while it embedded 100 documents."
diff --git a/content/answers/store-agent-memory-without-llm-call.md b/content/answers/store-agent-memory-without-llm-call.md
index 5e721c9de..509a50685 100644
--- a/content/answers/store-agent-memory-without-llm-call.md
+++ b/content/answers/store-agent-memory-without-llm-call.md
@@ -7,7 +7,7 @@ cluster: "Agent memory: model dependency"
question: "How do I dedup agent memory?"
intent: "how-to"
published: "2026-08-21"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -37,6 +37,8 @@ links_out:
- "do-search-embeddings-help"
- "local-embeddings-without-openai-api"
- "/compare/xerj-vs-agent-memory-tools"
+ - "/compare/xerj-vs-mem0"
+ - "/answers/what-is-an-mcp-memory-server"
faq:
- q: "How do I dedup agent memory?"
a: "Send `dedup: true` on the store call. When the cosine score against an existing memory meets `dedup_threshold`, XERJ returns that id with `\"deduplicated\": true` and writes nothing new."
diff --git a/content/answers/vector-database-vs-full-text-search.md b/content/answers/vector-database-vs-full-text-search.md
index 7d14ca2de..ce7365d61 100644
--- a/content/answers/vector-database-vs-full-text-search.md
+++ b/content/answers/vector-database-vs-full-text-search.md
@@ -7,7 +7,7 @@ cluster: "Retrieval: selection"
question: "What's the actual difference between a vector database and full-text search?"
intent: "informational"
published: "2026-08-21"
-updated: "2026-10-19"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -35,6 +35,8 @@ links_out:
- "do-search-embeddings-help"
- "mcp-api-search-or-vector-search"
- "xerj-vs-vector-database"
+ - "/answers/bm25-vs-vector-search"
+ - "/answers/what-is-hybrid-search"
faq:
- q: "What's the actual difference between a vector database and full-text search?"
a: "Full-text search matches tokens and ranks with BM25. Vector search compares one embedding against stored dense vectors. XERJ serves both from one index."
diff --git a/content/answers/what-is-agentic-memory.md b/content/answers/what-is-agentic-memory.md
index 0c143d2b4..8bfa17609 100644
--- a/content/answers/what-is-agentic-memory.md
+++ b/content/answers/what-is-agentic-memory.md
@@ -7,7 +7,7 @@ cluster: "Agent memory: concepts"
question: "What is agentic memory?"
intent: "definition"
published: "2026-10-28"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -44,6 +44,10 @@ links_out:
- "agent-memory-without-postgres-qdrant"
- "store-agent-memory-without-llm-call"
- "/docs/recipes/agentic-memory"
+ - "/answers/agentic-rag-vs-rag"
+ - "/compare/xerj-vs-mem0"
+ - "/compare/xerj-vs-claude-mem-and-mempalace"
+ - "/answers/claude-code-memory"
faq:
- q: "What is agentic memory in simple words?"
a: "Notes an AI agent keeps for itself. The agent saves a fact with a tool call and looks it up later, so the fact outlives the chat that produced it."
diff --git a/content/compare/best-grep-alternative-for-ai-coding-agents-2026.md b/content/compare/best-grep-alternative-for-ai-coding-agents-2026.md
index c6836db1d..e90799431 100644
--- a/content/compare/best-grep-alternative-for-ai-coding-agents-2026.md
+++ b/content/compare/best-grep-alternative-for-ai-coding-agents-2026.md
@@ -7,7 +7,7 @@ cluster: "Comparison: code retrieval"
question: "What is the best grep or ripgrep alternative for an AI coding agent in 2026?"
intent: "comparison"
published: "2026-10-19"
-updated: "2026-10-28"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -64,6 +64,7 @@ links_out:
- /compare/best-mcp-server-local-code-search-2026
- "/compare/xerj-vs-localsynapse"
- "/answers/search-jupyter-notebook-cells"
+ - "/compare/best-local-full-text-search-jsonl-csv-2026"
evidence:
- claim: "ripgrep installs with brew install ripgrep, apt-get install ripgrep or cargo install ripgrep."
source: "https://github.com/BurntSushi/ripgrep"
diff --git a/content/compare/best-lightweight-elasticsearch-compatible-server-2026.md b/content/compare/best-lightweight-elasticsearch-compatible-server-2026.md
index f21d3d133..917ab5012 100644
--- a/content/compare/best-lightweight-elasticsearch-compatible-server-2026.md
+++ b/content/compare/best-lightweight-elasticsearch-compatible-server-2026.md
@@ -7,7 +7,7 @@ cluster: "Comparison: search engines"
question: "What is the best lightweight Elasticsearch-compatible server in 2026?"
intent: "comparison"
published: "2026-10-19"
-updated: "2026-10-23"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -61,6 +61,7 @@ links_out:
- /compare/xerj-vs-meilisearch
- /compare/xerj-vs-typesense
- /compare/best-local-full-text-search-jsonl-csv-2026
+ - "/benchmarks/elasticsearch"
evidence:
- claim: "OpenSearch is Apache-2.0, includes Apache-licensed Elasticsearch code, and runs as a single node with docker run and discovery.type=single-node."
source: "https://github.com/opensearch-project/OpenSearch"
diff --git a/content/compare/xerj-vs-elasticsearch-local.md b/content/compare/xerj-vs-elasticsearch-local.md
index 114c6c77d..0778a5e81 100644
--- a/content/compare/xerj-vs-elasticsearch-local.md
+++ b/content/compare/xerj-vs-elasticsearch-local.md
@@ -7,7 +7,7 @@ cluster: "Comparison: local search engines"
question: "Is there an Elasticsearch-compatible engine that isn't Elasticsearch?"
intent: "comparison"
published: "2026-08-22"
-updated: "2026-10-23"
+updated: "2026-10-31"
author: "XERJ documentation team"
reviewer: "XERJ engineering team"
schema_type: "TechArticle"
@@ -36,6 +36,7 @@ links_out:
- "/docs/recipes/migrate-from-elasticsearch"
- "/answers/full-text-search-over-jsonl-files-cli"
- "/answers/batch-lookup-cve-ghsa-ids-offline"
+ - "/benchmarks/elasticsearch"
evidence:
- claim: "The published board is 55 WIN, 4 LOSE, 26 TIE and 3 N/A over 88 cells at 100,000 documents against a live Elasticsearch 8.13.4 on one host."
source: "demo/playbooks/SCORECARD.md"
diff --git a/landing/answers/batch-lookup-cve-ghsa-ids-offline.html b/landing/answers/batch-lookup-cve-ghsa-ids-offline.html
index 4c84e9c0b..4423f415a 100644
--- a/landing/answers/batch-lookup-cve-ghsa-ids-offline.html
+++ b/landing/answers/batch-lookup-cve-ghsa-ids-offline.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/answers-batch-lookup-cve-ghsa-ids-offline.png",
"inLanguage": "en",
"datePublished": "2026-10-19",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -224,7 +224,7 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>BATCH LOOKUP CVE IDS AND GHSA IDS OFFLINE, WITHOUT GITHUB API RATE LIMITS</div>
<div class="kicker"><span>ANSWERS · ANSWERS FOR CODING AGENTS</span></div>
<h1>Batch lookup CVE IDs and GHSA IDs offline, without GitHub API rate limits</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-19 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-19 · UPDATED · 2026-10-31</div>
<p><strong>TL;DR</strong> — To batch lookup CVE and GHSA IDs offline without GitHub API rate limits, load the signed rust-vulns advisory pack into xerj and resolve every ID in one <code class="inline">terms</code> query. Each answer names the source file; unknown IDs abstain. On 2026-10-19, 6 IDs took 15 ms: 4 resolved, 2 abstained. Coverage is Rust crates only.</p>
<div class="prompt-card">
<span class="prompt-label">RUN THIS XERJ COMMAND</span>
@@ -347,6 +347,7 @@ CVE-2099-00001 NOT FOUND in this pack: abstain
<li><a href="/blog/one-vulnerability-three-names">One vulnerability, three names</a></li>
<li><a href="/blog/the-exploit-hub">The exploit hub</a></li>
<li><a href="/compare/xerj-vs-elasticsearch-local">Is there an Elasticsearch-compatible engine that isn't Elasticsearch?</a></li>
+ <li><a href="/case-studies/wordpress-security-audit">WordPress security audit</a></li>
</ul>
</main>
<nav class="docs-nav" aria-label="Article sections">
diff --git a/landing/answers/batch-lookup-cve-ghsa-ids-offline.md b/landing/answers/batch-lookup-cve-ghsa-ids-offline.md
index 69bfa3dbe..5fdbfd46a 100644
--- a/landing/answers/batch-lookup-cve-ghsa-ids-offline.md
+++ b/landing/answers/batch-lookup-cve-ghsa-ids-offline.md
@@ -2,7 +2,7 @@
---
title: "Batch lookup CVE and GHSA IDs offline in one query"
canonical: "https://xerj.org/answers/batch-lookup-cve-ghsa-ids-offline"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/batch-lookup-cve-ghsa-ids-offline.md"
---
@@ -169,3 +169,4 @@ Not reliably. `xerj code rust-vulns "CVE-2021-25900"` ranked an lz4 advisory fir
- [One vulnerability, three names](/blog/one-vulnerability-three-names)
- [The exploit hub](/blog/the-exploit-hub)
- [Is there an Elasticsearch-compatible engine that isn't Elasticsearch?](/compare/xerj-vs-elasticsearch-local)
+- [WordPress security audit](/case-studies/wordpress-security-audit)
diff --git a/landing/answers/code-search-mcp-for-claude-code.html b/landing/answers/code-search-mcp-for-claude-code.html
index 1b0e41c90..14427fb39 100644
--- a/landing/answers/code-search-mcp-for-claude-code.html
+++ b/landing/answers/code-search-mcp-for-claude-code.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -224,9 +224,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>I NEED MY AGENT TO FIND A PASSAGE AND CITE THE FILE, NOT DUMP THE WHOLE TREE. HOW?</div>
<div class="kicker"><span>ANSWERS · AGENT ACCESS AND MEMORY</span></div>
<h1 class="h1-long">I need my agent to find a passage and cite the file, not dump the whole tree. How?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>How do I make my agent find a passage and cite the file, not dump the whole tree? Index the repository with xerj autoindex, register xerj mcp in the agent host, and query the defs field. One captured xerj_search call returned 2 documents whose symbols arrays gave file and line, including merge_segments at src/lib.rs line 17.</p>
<p class="key-fact"><strong>Key number:</strong> XERJ's tree-sitter registry covers 36 languages, a count pinned by a unit test so the docs cannot drift. <span class="mute">(as of 2026-10-19 · source: <a href="https://github.com/xerj-org/xerj/blob/main/engine/crates/xerj-autoindex/src/extract/code.rs">engine/crates/xerj-autoindex/src/extract/code.rs</a>)</span></p>
</div>
@@ -338,6 +338,9 @@ none of postgres/qdrant/redis/weaviate/milvus/chroma/elasticsearch/opensearch/ol
<li><a href="/compare/xerj-vs-ast-grep">ast-grep vs a code index for an agent?</a></li>
<li><a href="/answers/local-code-search-mcp-fewer-tokens-than-grep">Local code search MCP for Claude Code that uses fewer tokens than grep</a></li>
<li><a href="/answers/ask-questions-about-a-github-repo-without-cloning">Ask questions about a GitHub repo without cloning it</a></li>
+ <li><a href="/answers/claude-code-memory">How does Claude Code memory work, and when do you need more?</a></li>
+ <li><a href="/answers/what-is-an-mcp-memory-server">What is an MCP memory server?</a></li>
+ <li><a href="/compare/xerj-vs-claude-mem-and-mempalace">claude-mem vs MemPalace vs XERJ: which memory should Claude Code use?</a></li>
<li><a href="/answers/audit-php-codebase-with-an-index">How do I audit a PHP codebase with an index instead of grepping 1,400 files?</a> <span class="mute">· same intent: code search for agents</span></li>
<li><a href="/compare/best-mcp-server-local-code-search-2026">The best MCP servers for searching a local codebase in 2026</a> <span class="mute">· same intent: code search for agents</span></li>
<li><a href="/compare/xerj-vs-ripgrep-for-code-agents">Is ripgrep enough for a coding agent, or do I need a code index?</a> <span class="mute">· same intent: code search for agents</span></li>
diff --git a/landing/answers/code-search-mcp-for-claude-code.md b/landing/answers/code-search-mcp-for-claude-code.md
index e089a3226..eb82cc940 100644
--- a/landing/answers/code-search-mcp-for-claude-code.md
+++ b/landing/answers/code-search-mcp-for-claude-code.md
@@ -2,13 +2,13 @@
---
title: "Code search MCP for Claude Code: cite file:line, not grep"
canonical: "https://xerj.org/answers/code-search-mcp-for-claude-code"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/code-search-mcp-for-claude-code.md"
---
# I need my agent to find a passage and cite the file, not dump the whole tree. How?
-**Short answer** (updated 2026-10-28): How do I make my agent find a passage and cite the file, not dump the whole tree? Index the repository with xerj autoindex, register xerj mcp in the agent host, and query the defs field. One captured xerj_search call returned 2 documents whose symbols arrays gave file and line, including merge_segments at src/lib.rs line 17.
+**Short answer** (updated 2026-10-31): How do I make my agent find a passage and cite the file, not dump the whole tree? Index the repository with xerj autoindex, register xerj mcp in the agent host, and query the defs field. One captured xerj_search call returned 2 documents whose symbols arrays gave file and line, including merge_segments at src/lib.rs line 17.
**Key number:** XERJ's tree-sitter registry covers 36 languages, a count pinned by a unit test so the docs cannot drift. (as of 2026-10-19; source: [engine/crates/xerj-autoindex/src/extract/code.rs](https://github.com/xerj-org/xerj/blob/main/engine/crates/xerj-autoindex/src/extract/code.rs))
@@ -184,6 +184,9 @@ No. Both list 6 tools, while on main the binary serves 23 (checked 2026-10-28).
- [ast-grep vs a code index for an agent?](/compare/xerj-vs-ast-grep)
- [Local code search MCP for Claude Code that uses fewer tokens than grep](/answers/local-code-search-mcp-fewer-tokens-than-grep)
- [Ask questions about a GitHub repo without cloning it](/answers/ask-questions-about-a-github-repo-without-cloning)
+- [How does Claude Code memory work, and when do you need more?](/answers/claude-code-memory)
+- [What is an MCP memory server?](/answers/what-is-an-mcp-memory-server)
+- [claude-mem vs MemPalace vs XERJ: which memory should Claude Code use?](/compare/xerj-vs-claude-mem-and-mempalace)
- [How do I audit a PHP codebase with an index instead of grepping 1,400 files?](/answers/audit-php-codebase-with-an-index) (same intent: code search for agents)
- [The best MCP servers for searching a local codebase in 2026](/compare/best-mcp-server-local-code-search-2026) (same intent: code search for agents)
- [Is ripgrep enough for a coding agent, or do I need a code index?](/compare/xerj-vs-ripgrep-for-code-agents) (same intent: code search for agents)
diff --git a/landing/answers/coding-agent-memory-across-sessions.html b/landing/answers/coding-agent-memory-across-sessions.html
index 9ffd55e39..b1a02123f 100644
--- a/landing/answers/coding-agent-memory-across-sessions.html
+++ b/landing/answers/coding-agent-memory-across-sessions.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -224,9 +224,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>HOW DO PEOPLE STORE USER PREFERENCES SO AN AGENT CAN RECALL THEM NEXT WEEK?</div>
<div class="kicker"><span>ANSWERS · AGENT MEMORY: SETUP</span></div>
<h1>How do people store user preferences so an agent can recall them next week?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>How do people store user preferences so an agent can recall them next week? XERJ stores each memory as a document in a reserved index, .xerj-memory-{namespace}, inside the node's data directory. Our capture stored two memories, restarted the node on the same directory, recalled both by id, and one forget call removed exactly one.</p>
<p class="key-fact"><strong>Key number:</strong> A memory store call deduplicates at a default cosine similarity of 0.95 when <code class="inline">dedup</code> is on. <span class="mute">(as of 2026-10-19 · source: <a href="https://github.com/xerj-org/xerj/blob/main/engine/crates/xerj-api/src/memory_api.rs">engine/crates/xerj-api/src/memory_api.rs</a>)</span></p>
</div>
@@ -314,9 +314,12 @@
<li><a href="/docs/recipes/agentic-memory">Agent memory</a></li>
<li><a href="/answers/what-is-agentic-memory">What is agentic memory?</a></li>
<li><a href="/answers/what-is-an-mcp-memory-server">What is an MCP memory server?</a></li>
+ <li><a href="/compare/xerj-vs-agent-memory-tools">XERJ vs agent memory tools: Mem0, Zep, Letta and Cognee</a></li>
+ <li><a href="/compare/xerj-vs-mem0">XERJ vs Mem0: which memory fits your AI agent?</a></li>
+ <li><a href="/answers/claude-code-memory">How does Claude Code memory work, and when do you need more?</a></li>
<li><a href="/compare/xerj-vs-vector-database">Should my agent call a search API or stand up Qdrant for persistent memory?</a> <span class="mute">· same intent: agent memory</span></li>
- <li><a href="/answers/claude-code-memory">How does Claude Code memory work, and when do you need more?</a> <span class="mute">· same intent: agent memory</span></li>
<li><a href="/compare/xerj-vs-claude-mem-and-mempalace">claude-mem vs MemPalace vs XERJ: which memory should Claude Code use?</a> <span class="mute">· same intent: agent memory</span></li>
+ <li><a href="/answers/agentic-memory-vs-rag">Agentic memory vs RAG: what is the difference?</a> <span class="mute">· same intent: agent memory</span></li>
</ul>
</main>
<nav class="docs-nav" aria-label="Article sections">
diff --git a/landing/answers/coding-agent-memory-across-sessions.md b/landing/answers/coding-agent-memory-across-sessions.md
index 8cf94ab38..7d350f537 100644
--- a/landing/answers/coding-agent-memory-across-sessions.md
+++ b/landing/answers/coding-agent-memory-across-sessions.md
@@ -2,13 +2,13 @@
---
title: "Store user preferences an agent can recall later"
canonical: "https://xerj.org/answers/coding-agent-memory-across-sessions"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/coding-agent-memory-across-sessions.md"
---
# How do people store user preferences so an agent can recall them next week?
-**Short answer** (updated 2026-10-28): How do people store user preferences so an agent can recall them next week? XERJ stores each memory as a document in a reserved index, .xerj-memory-{namespace}, inside the node's data directory. Our capture stored two memories, restarted the node on the same directory, recalled both by id, and one forget call removed exactly one.
+**Short answer** (updated 2026-10-31): How do people store user preferences so an agent can recall them next week? XERJ stores each memory as a document in a reserved index, .xerj-memory-{namespace}, inside the node's data directory. Our capture stored two memories, restarted the node on the same directory, recalled both by id, and one forget call removed exactly one.
**Key number:** A memory store call deduplicates at a default cosine similarity of 0.95 when `dedup` is on. (as of 2026-10-19; source: [engine/crates/xerj-api/src/memory_api.rs](https://github.com/xerj-org/xerj/blob/main/engine/crates/xerj-api/src/memory_api.rs))
@@ -147,6 +147,9 @@ Only if you ask. Plain `query` runs BM25, `semantic: true` embeds server-side, a
- [Agent memory](/docs/recipes/agentic-memory)
- [What is agentic memory?](/answers/what-is-agentic-memory)
- [What is an MCP memory server?](/answers/what-is-an-mcp-memory-server)
+- [XERJ vs agent memory tools: Mem0, Zep, Letta and Cognee](/compare/xerj-vs-agent-memory-tools)
+- [XERJ vs Mem0: which memory fits your AI agent?](/compare/xerj-vs-mem0)
+- [How does Claude Code memory work, and when do you need more?](/answers/claude-code-memory)
- [Should my agent call a search API or stand up Qdrant for persistent memory?](/compare/xerj-vs-vector-database) (same intent: agent memory)
-- [How does Claude Code memory work, and when do you need more?](/answers/claude-code-memory) (same intent: agent memory)
- [claude-mem vs MemPalace vs XERJ: which memory should Claude Code use?](/compare/xerj-vs-claude-mem-and-mempalace) (same intent: agent memory)
+- [Agentic memory vs RAG: what is the difference?](/answers/agentic-memory-vs-rag) (same intent: agent memory)
diff --git a/landing/answers/do-search-embeddings-help.html b/landing/answers/do-search-embeddings-help.html
index d82a840de..c26d5afbc 100644
--- a/landing/answers/do-search-embeddings-help.html
+++ b/landing/answers/do-search-embeddings-help.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-19",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -224,9 +224,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>WHEN SHOULD I TURN ON NEURAL EMBEDDINGS, AND WHEN IS THE LEXICAL DEFAULT ENOUGH?</div>
<div class="kicker"><span>ANSWERS · EMBEDDINGS: SELECTION</span></div>
<h1>When should I turn on neural embeddings, and when is the lexical default enough?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-19</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-19</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>Do I even need embeddings for search, or is full-text enough? On XERJ's default lexical embedder, full-text BM25 is close: 0.4167 precision at 3 against BM25's 0.375 on 8 judged queries. Embeddings clearly helped only after opting into --embed-mode neural, which moved precision at 3 to 0.6667 and recall to 1.0.</p>
<p class="key-fact"><strong>Key number:</strong> On BEIR SciFact, hybrid RRF with the neural embedder scored nDCG@10 0.6993 against 0.6572 for BM25 alone. <span class="mute">(as of 2026-09-18 · source: <a href="https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/scifact.txt">benchmarks/beir-hybrid/results/scifact.txt</a>)</span></p>
</div>
@@ -332,6 +332,8 @@
<li><a href="/compare/xerj-vs-vector-database">Should my agent call a search API or stand up Qdrant for persistent memory?</a></li>
<li><a href="/answers/vector-database-vs-full-text-search">What's the actual difference between a vector database and full-text search when I have a folder?</a></li>
<li><a href="/answers/reduce-embedding-api-cost">How do I reduce embedding API cost?</a></li>
+ <li><a href="/answers/bm25-vs-vector-search">BM25 vs vector search: which one should you use?</a></li>
+ <li><a href="/answers/classify-text-from-labelled-history-without-llm">Can I route tickets or filter spam from labelled history, without an LLM call?</a></li>
<li><a href="/answers/mcp-api-search-or-vector-search">Should an agent use API search or vector search?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/how-xerj-combines-search">How does XERJ combine BM25 and kNN?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/hybrid-search-quality-measured-ndcg">How much better is hybrid search than BM25? Measured nDCG numbers</a> <span class="mute">· same intent: retrieval research and eval</span></li>
diff --git a/landing/answers/do-search-embeddings-help.md b/landing/answers/do-search-embeddings-help.md
index c0f038d97..7ecd7037f 100644
--- a/landing/answers/do-search-embeddings-help.md
+++ b/landing/answers/do-search-embeddings-help.md
@@ -2,13 +2,13 @@
---
title: "When are neural embeddings worth turning on?"
canonical: "https://xerj.org/answers/do-search-embeddings-help"
-updated: "2026-10-19"
+updated: "2026-10-31"
source: "content/answers/do-search-embeddings-help.md"
---
# When should I turn on neural embeddings, and when is the lexical default enough?
-**Short answer** (updated 2026-10-19): Do I even need embeddings for search, or is full-text enough? On XERJ's default lexical embedder, full-text BM25 is close: 0.4167 precision at 3 against BM25's 0.375 on 8 judged queries. Embeddings clearly helped only after opting into --embed-mode neural, which moved precision at 3 to 0.6667 and recall to 1.0.
+**Short answer** (updated 2026-10-31): Do I even need embeddings for search, or is full-text enough? On XERJ's default lexical embedder, full-text BM25 is close: 0.4167 precision at 3 against BM25's 0.375 on 8 judged queries. Embeddings clearly helped only after opting into --embed-mode neural, which moved precision at 3 to 0.6667 and recall to 1.0.
**Key number:** On BEIR SciFact, hybrid RRF with the neural embedder scored nDCG@10 0.6993 against 0.6572 for BM25 alone. (as of 2026-09-18; source: [benchmarks/beir-hybrid/results/scifact.txt](https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/scifact.txt))
@@ -158,6 +158,8 @@ BM25 alone scored 0.375 precision at 3 and 0.6875 recall on our set. It found ev
- [Should my agent call a search API or stand up Qdrant for persistent memory?](/compare/xerj-vs-vector-database)
- [What's the actual difference between a vector database and full-text search when I have a folder?](/answers/vector-database-vs-full-text-search)
- [How do I reduce embedding API cost?](/answers/reduce-embedding-api-cost)
+- [BM25 vs vector search: which one should you use?](/answers/bm25-vs-vector-search)
+- [Can I route tickets or filter spam from labelled history, without an LLM call?](/answers/classify-text-from-labelled-history-without-llm)
- [Should an agent use API search or vector search?](/answers/mcp-api-search-or-vector-search) (same intent: retrieval research and eval)
- [How does XERJ combine BM25 and kNN?](/answers/how-xerj-combines-search) (same intent: retrieval research and eval)
- [How much better is hybrid search than BM25? Measured nDCG numbers](/answers/hybrid-search-quality-measured-ndcg) (same intent: retrieval research and eval)
diff --git a/landing/answers/full-text-search-over-jsonl-files-cli.html b/landing/answers/full-text-search-over-jsonl-files-cli.html
index 07537815e..f0970545d 100644
--- a/landing/answers/full-text-search-over-jsonl-files-cli.html
+++ b/landing/answers/full-text-search-over-jsonl-files-cli.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/answers-full-text-search-over-jsonl-files-cli.png",
"inLanguage": "en",
"datePublished": "2026-10-19",
- "dateModified": "2026-10-20",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -216,7 +216,7 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>FULL-TEXT SEARCH OVER JSONL FILES FROM THE CLI: INDEX A FOLDER AND QUERY IT WITH BM25</div>
<div class="kicker"><span>ANSWERS · ANSWERS FOR CODING AGENTS</span></div>
<h1 class="h1-long">Full-text search over JSONL files from the CLI: index a folder and query it with BM25</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-19 · UPDATED · 2026-10-20</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-19 · UPDATED · 2026-10-31</div>
<p><strong>TL;DR</strong> — For full-text search over JSONL files from the CLI, run <code class="inline">xerj autoindex</code> on the folder and query it with BM25. No schema or mapping is needed; CSV works the same way. On 2026-10-19 a folder of 2,902 JSONL and CSV rows indexed in 12.4 s, and the field queries returned ranked hits in 1 ms.</p>
<div class="prompt-card">
<span class="prompt-label">RUN THIS XERJ COMMAND</span>
@@ -325,6 +325,7 @@ indexed: 2902 documents live; this run submitted 2902 source records from 3 file
<li><a href="/answers/batch-lookup-cve-ghsa-ids-offline">Batch lookup CVE IDs and GHSA IDs offline, without GitHub API rate limits</a></li>
<li><a href="/compare/xerj-vs-elasticsearch-local">Is there an Elasticsearch-compatible engine that isn't Elasticsearch?</a></li>
<li><a href="/compare/xerj-vs-vector-database">Should my agent call a search API or stand up Qdrant for persistent memory?</a></li>
+ <li><a href="/compare/best-local-full-text-search-jsonl-csv-2026">The best local full-text search for JSONL and CSV files in 2026, with a published benchmark</a></li>
</ul>
</main>
<nav class="docs-nav" aria-label="Article sections">
diff --git a/landing/answers/full-text-search-over-jsonl-files-cli.md b/landing/answers/full-text-search-over-jsonl-files-cli.md
index 96efb3b66..781d59272 100644
--- a/landing/answers/full-text-search-over-jsonl-files-cli.md
+++ b/landing/answers/full-text-search-over-jsonl-files-cli.md
@@ -2,7 +2,7 @@
---
title: "Full-text search over JSONL and CSV files from the CLI"
canonical: "https://xerj.org/answers/full-text-search-over-jsonl-files-cli"
-updated: "2026-10-20"
+updated: "2026-10-31"
source: "content/answers/full-text-search-over-jsonl-files-cli.md"
---
@@ -153,3 +153,4 @@ Yes, but check the label. Our 1,385-record file was detected as format `json` ra
- [Batch lookup CVE IDs and GHSA IDs offline, without GitHub API rate limits](/answers/batch-lookup-cve-ghsa-ids-offline)
- [Is there an Elasticsearch-compatible engine that isn't Elasticsearch?](/compare/xerj-vs-elasticsearch-local)
- [Should my agent call a search API or stand up Qdrant for persistent memory?](/compare/xerj-vs-vector-database)
+- [The best local full-text search for JSONL and CSV files in 2026, with a published benchmark](/compare/best-local-full-text-search-jsonl-csv-2026)
diff --git a/landing/answers/how-xerj-combines-search.html b/landing/answers/how-xerj-combines-search.html
index 94f288f0f..ce235eb36 100644
--- a/landing/answers/how-xerj-combines-search.html
+++ b/landing/answers/how-xerj-combines-search.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-19",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -224,9 +224,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>HOW DOES XERJ COMBINE BM25 AND KNN?</div>
<div class="kicker"><span>ANSWERS · HYBRID SEARCH</span></div>
<h1>How does XERJ combine BM25 and kNN?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-19</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-19</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>How does XERJ combine BM25 and kNN? One hybrid request carries a BM25 sub-query and a kNN sub-query against a single index, and Reciprocal Rank Fusion merges the two ranked lists. Fusion reads rank positions, not scores, so the BM25 and cosine scales never need to be normalized against each other.</p>
<p class="key-fact"><strong>Key number:</strong> XERJ's RRF uses a default k of 60, so a document ranked first in both lists scores 1/61 + 1/61 = 0.032787. <span class="mute">(as of 2026-08-23 · source: <a href="/answers/reciprocal-rank-fusion-when-to-use">/answers/reciprocal-rank-fusion-when-to-use</a>)</span></p>
</div>
@@ -306,6 +306,8 @@
<li><a href="/answers/what-is-xerj">What is XERJ?</a></li>
<li><a href="/docs/recipes/hybrid-search">Hybrid search</a></li>
<li><a href="/docs/playbooks/vector-search">Playbook · Vector &amp; RAG</a></li>
+ <li><a href="/answers/what-is-hybrid-search">What is hybrid search?</a></li>
+ <li><a href="/answers/bm25-vs-vector-search">BM25 vs vector search: which one should you use?</a></li>
<li><a href="/answers/mcp-api-search-or-vector-search">Should an agent use API search or vector search?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/do-search-embeddings-help">When should I turn on neural embeddings, and when is the lexical default enough?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/hybrid-search-quality-measured-ndcg">How much better is hybrid search than BM25? Measured nDCG numbers</a> <span class="mute">· same intent: retrieval research and eval</span></li>
diff --git a/landing/answers/how-xerj-combines-search.md b/landing/answers/how-xerj-combines-search.md
index 2cbd56594..b2dca63e8 100644
--- a/landing/answers/how-xerj-combines-search.md
+++ b/landing/answers/how-xerj-combines-search.md
@@ -2,13 +2,13 @@
---
title: "How does XERJ combine BM25 and kNN?"
canonical: "https://xerj.org/answers/how-xerj-combines-search"
-updated: "2026-10-19"
+updated: "2026-10-31"
source: "content/answers/how-xerj-combines-search.md"
---
# How does XERJ combine BM25 and kNN?
-**Short answer** (updated 2026-10-19): How does XERJ combine BM25 and kNN? One hybrid request carries a BM25 sub-query and a kNN sub-query against a single index, and Reciprocal Rank Fusion merges the two ranked lists. Fusion reads rank positions, not scores, so the BM25 and cosine scales never need to be normalized against each other.
+**Short answer** (updated 2026-10-31): How does XERJ combine BM25 and kNN? One hybrid request carries a BM25 sub-query and a kNN sub-query against a single index, and Reciprocal Rank Fusion merges the two ranked lists. Fusion reads rank positions, not scores, so the BM25 and cosine scales never need to be normalized against each other.
**Key number:** XERJ's RRF uses a default k of 60, so a document ranked first in both lists scores 1/61 + 1/61 = 0.032787. (as of 2026-08-23; source: [/answers/reciprocal-rank-fusion-when-to-use](/answers/reciprocal-rank-fusion-when-to-use))
@@ -136,6 +136,8 @@ Not on a default node. The default embedder is lexical feature hashing, so fusin
- [What is XERJ?](/answers/what-is-xerj)
- [Hybrid search](/docs/recipes/hybrid-search)
- [Playbook · Vector & RAG](/docs/playbooks/vector-search)
+- [What is hybrid search?](/answers/what-is-hybrid-search)
+- [BM25 vs vector search: which one should you use?](/answers/bm25-vs-vector-search)
- [Should an agent use API search or vector search?](/answers/mcp-api-search-or-vector-search) (same intent: retrieval research and eval)
- [When should I turn on neural embeddings, and when is the lexical default enough?](/answers/do-search-embeddings-help) (same intent: retrieval research and eval)
- [How much better is hybrid search than BM25? Measured nDCG numbers](/answers/hybrid-search-quality-measured-ndcg) (same intent: retrieval research and eval)
diff --git a/landing/answers/hybrid-search-quality-measured-ndcg.html b/landing/answers/hybrid-search-quality-measured-ndcg.html
index 80e510965..a90017efc 100644
--- a/landing/answers/hybrid-search-quality-measured-ndcg.html
+++ b/landing/answers/hybrid-search-quality-measured-ndcg.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-09-18",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -216,9 +216,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>HOW MUCH BETTER IS HYBRID SEARCH THAN BM25? MEASURED NDCG NUMBERS</div>
<div class="kicker"><span>ANSWERS · HYBRID RETRIEVAL: QUALITY</span></div>
<h1>How much better is hybrid search than BM25? Measured nDCG numbers</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-09-18 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-09-18 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>How much better is hybrid search than BM25 alone? On two public BEIR datasets XERJ's hybrid RRF beat both BM25 and vectors: nDCG@10 0.6993 against 0.6572 on SciFact and 0.3448 against 0.3016 on NFCorpus. Those runs used --embed-mode neural with all-MiniLM-L6-v2; the default lexical embedder is not covered.</p>
<p class="key-fact"><strong>Key number:</strong> On BEIR NFCorpus (323 test queries, 3,633 documents), hybrid RRF scored nDCG@10 0.3448 against 0.3016 for BM25. <span class="mute">(as of 2026-09-18 · source: <a href="https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/nfcorpus.txt">benchmarks/beir-hybrid/results/nfcorpus.txt</a>)</span></p>
</div>
@@ -326,6 +326,10 @@
<li><a href="/answers/local-embeddings-without-openai-api">I want local search that works offline, no cloud embeddings. What are my options?</a></li>
<li><a href="/answers/retrieval-that-says-not-found-with-file-line-citations">Retrieval that says not found, and cites file and line when it does answer</a></li>
<li><a href="/answers/improve-basic-faiss-rag-pipeline">How do I improve a basic FAISS RAG pipeline?</a></li>
+ <li><a href="/answers/what-is-hybrid-search">What is hybrid search?</a></li>
+ <li><a href="/answers/bm25-vs-vector-search">BM25 vs vector search: which one should you use?</a></li>
+ <li><a href="/answers/classify-text-from-labelled-history-without-llm">Can I route tickets or filter spam from labelled history, without an LLM call?</a></li>
+ <li><a href="/blog/the-rc80-gates">The rc.80 gates</a></li>
<li><a href="/answers/rag-without-vector-database">How do I build RAG without a vector database?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/how-xerj-combines-search">How does XERJ combine BM25 and kNN?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/mcp-api-search-or-vector-search">Should an agent use API search or vector search?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
diff --git a/landing/answers/hybrid-search-quality-measured-ndcg.md b/landing/answers/hybrid-search-quality-measured-ndcg.md
index 35f43f537..24f9aa3d3 100644
--- a/landing/answers/hybrid-search-quality-measured-ndcg.md
+++ b/landing/answers/hybrid-search-quality-measured-ndcg.md
@@ -2,13 +2,13 @@
---
title: "How much better is hybrid search than BM25? Measured."
canonical: "https://xerj.org/answers/hybrid-search-quality-measured-ndcg"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/hybrid-search-quality-measured-ndcg.md"
---
# How much better is hybrid search than BM25? Measured nDCG numbers
-**Short answer** (updated 2026-10-28): How much better is hybrid search than BM25 alone? On two public BEIR datasets XERJ's hybrid RRF beat both BM25 and vectors: nDCG@10 0.6993 against 0.6572 on SciFact and 0.3448 against 0.3016 on NFCorpus. Those runs used --embed-mode neural with all-MiniLM-L6-v2; the default lexical embedder is not covered.
+**Short answer** (updated 2026-10-31): How much better is hybrid search than BM25 alone? On two public BEIR datasets XERJ's hybrid RRF beat both BM25 and vectors: nDCG@10 0.6993 against 0.6572 on SciFact and 0.3448 against 0.3016 on NFCorpus. Those runs used --embed-mode neural with all-MiniLM-L6-v2; the default lexical embedder is not covered.
**Key number:** On BEIR NFCorpus (323 test queries, 3,633 documents), hybrid RRF scored nDCG@10 0.3448 against 0.3016 for BM25. (as of 2026-09-18; source: [benchmarks/beir-hybrid/results/nfcorpus.txt](https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/nfcorpus.txt))
@@ -157,6 +157,10 @@ About 3 to 7 documents per second on CPU in this run. SciFact's 5,183 short abst
- [I want local search that works offline, no cloud embeddings. What are my options?](/answers/local-embeddings-without-openai-api)
- [Retrieval that says not found, and cites file and line when it does answer](/answers/retrieval-that-says-not-found-with-file-line-citations)
- [How do I improve a basic FAISS RAG pipeline?](/answers/improve-basic-faiss-rag-pipeline)
+- [What is hybrid search?](/answers/what-is-hybrid-search)
+- [BM25 vs vector search: which one should you use?](/answers/bm25-vs-vector-search)
+- [Can I route tickets or filter spam from labelled history, without an LLM call?](/answers/classify-text-from-labelled-history-without-llm)
+- [The rc.80 gates](/blog/the-rc80-gates)
- [How do I build RAG without a vector database?](/answers/rag-without-vector-database) (same intent: retrieval research and eval)
- [How does XERJ combine BM25 and kNN?](/answers/how-xerj-combines-search) (same intent: retrieval research and eval)
- [Should an agent use API search or vector search?](/answers/mcp-api-search-or-vector-search) (same intent: retrieval research and eval)
diff --git a/landing/answers/improve-basic-faiss-rag-pipeline.html b/landing/answers/improve-basic-faiss-rag-pipeline.html
index cae14b96e..4d97582b9 100644
--- a/landing/answers/improve-basic-faiss-rag-pipeline.html
+++ b/landing/answers/improve-basic-faiss-rag-pipeline.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -224,9 +224,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>HOW DO I IMPROVE A BASIC FAISS RAG PIPELINE?</div>
<div class="kicker"><span>ANSWERS · RAG: TROUBLESHOOTING</span></div>
<h1>How do I improve a basic FAISS RAG pipeline?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>How can I improve a basic FAISS RAG pipeline? Load the chunk store into one XERJ index with a semantic_text field and retrieve with the hybrid clause, which fuses BM25 and kNN, instead of vectors alone. Our 20-chunk migration indexed cleanly, and the two request shapes FAISS users write first both returned HTTP 400.</p>
<p class="key-fact"><strong>Key number:</strong> With the neural embedder, hybrid RRF beat vectors alone on BEIR SciFact: nDCG@10 0.6993 against 0.6764. <span class="mute">(as of 2026-09-18 · source: <a href="https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/scifact.txt">benchmarks/beir-hybrid/results/scifact.txt</a>)</span></p>
</div>
@@ -319,6 +319,9 @@
<li><a href="/answers/rag-without-vector-database">How do I build RAG without a vector database?</a></li>
<li><a href="/answers/reciprocal-rank-fusion-when-to-use">What is Reciprocal Rank Fusion?</a></li>
<li><a href="/answers/vector-database-vs-full-text-search">What's the actual difference between a vector database and full-text search when I have a folder?</a></li>
+ <li><a href="/answers/agentic-rag-vs-rag">Agentic RAG vs RAG: what is the difference?</a></li>
+ <li><a href="/answers/what-is-hybrid-search">What is hybrid search?</a></li>
+ <li><a href="/answers/agentic-memory-vs-rag">Agentic memory vs RAG: what is the difference?</a></li>
<li><a href="/answers/how-xerj-combines-search">How does XERJ combine BM25 and kNN?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/hybrid-search-quality-measured-ndcg">How much better is hybrid search than BM25? Measured nDCG numbers</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/rerank-search-results-calibrated-judge">How do I rerank search results with a relevance judge?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
diff --git a/landing/answers/improve-basic-faiss-rag-pipeline.md b/landing/answers/improve-basic-faiss-rag-pipeline.md
index 6b2d8d2b0..f118d2763 100644
--- a/landing/answers/improve-basic-faiss-rag-pipeline.md
+++ b/landing/answers/improve-basic-faiss-rag-pipeline.md
@@ -2,13 +2,13 @@
---
title: "Improve a basic FAISS RAG pipeline with hybrid search"
canonical: "https://xerj.org/answers/improve-basic-faiss-rag-pipeline"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/improve-basic-faiss-rag-pipeline.md"
---
# How do I improve a basic FAISS RAG pipeline?
-**Short answer** (updated 2026-10-28): How can I improve a basic FAISS RAG pipeline? Load the chunk store into one XERJ index with a semantic_text field and retrieve with the hybrid clause, which fuses BM25 and kNN, instead of vectors alone. Our 20-chunk migration indexed cleanly, and the two request shapes FAISS users write first both returned HTTP 400.
+**Short answer** (updated 2026-10-31): How can I improve a basic FAISS RAG pipeline? Load the chunk store into one XERJ index with a semantic_text field and retrieve with the hybrid clause, which fuses BM25 and kNN, instead of vectors alone. Our 20-chunk migration indexed cleanly, and the two request shapes FAISS users write first both returned HTTP 400.
**Key number:** With the neural embedder, hybrid RRF beat vectors alone on BEIR SciFact: nDCG@10 0.6993 against 0.6764. (as of 2026-09-18; source: [benchmarks/beir-hybrid/results/scifact.txt](https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/scifact.txt))
@@ -153,6 +153,9 @@ Partly. Weighted `linear` fusion reached 0.4583 precision at 3 on the lexical de
- [How do I build RAG without a vector database?](/answers/rag-without-vector-database)
- [What is Reciprocal Rank Fusion?](/answers/reciprocal-rank-fusion-when-to-use)
- [What's the actual difference between a vector database and full-text search when I have a folder?](/answers/vector-database-vs-full-text-search)
+- [Agentic RAG vs RAG: what is the difference?](/answers/agentic-rag-vs-rag)
+- [What is hybrid search?](/answers/what-is-hybrid-search)
+- [Agentic memory vs RAG: what is the difference?](/answers/agentic-memory-vs-rag)
- [How does XERJ combine BM25 and kNN?](/answers/how-xerj-combines-search) (same intent: retrieval research and eval)
- [How much better is hybrid search than BM25? Measured nDCG numbers](/answers/hybrid-search-quality-measured-ndcg) (same intent: retrieval research and eval)
- [How do I rerank search results with a relevance judge?](/answers/rerank-search-results-calibrated-judge) (same intent: retrieval research and eval)
diff --git a/landing/answers/index.html b/landing/answers/index.html
index c8b0f93f5..7b1e1af39 100644
--- a/landing/answers/index.html
+++ b/landing/answers/index.html
@@ -77,7 +77,7 @@
"url": "https://xerj.org/answers/",
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
- "dateModified": "2026-10-30",
+ "dateModified": "2026-10-31",
"isPartOf": {
"@id": "https://xerj.org/#website"
},
diff --git a/landing/answers/index.json b/landing/answers/index.json
index 7e2cf2076..fa4715a50 100644
--- a/landing/answers/index.json
+++ b/landing/answers/index.json
@@ -41,7 +41,7 @@
"canonical_url": "https://xerj.org/answers/code-search-mcp-for-claude-code",
"markdown_url": "https://xerj.org/answers/code-search-mcp-for-claude-code.md",
"cluster": "Agent access and memory",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "Local code search MCP for Claude Code with file:line citations: index with xerj autoindex, call xerj_search. One captured call cited 2 symbols without opening files.",
"intent_group": "code search for agents",
"answer": "How do I make my agent find a passage and cite the file, not dump the whole tree? Index the repository with xerj autoindex, register xerj mcp in the agent host, and query the defs field. One captured xerj_search call returned 2 documents whose symbols arrays gave file and line, including merge_segments at src/lib.rs line 17.",
@@ -107,7 +107,7 @@
"canonical_url": "https://xerj.org/answers/what-is-agentic-memory",
"markdown_url": "https://xerj.org/answers/what-is-agentic-memory.md",
"cluster": "Agent memory: concepts",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "Agentic memory is memory an AI agent writes and reads itself through tools, so facts survive past one chat. What it is, the common types, and how to run it locally.",
"intent_group": "agent memory",
"answer": "What is agentic memory? It is memory that an AI agent manages itself: the agent decides what to save, writes it through a tool call, and searches it again in a later task or session. It differs from a chat history because it is stored outside the model and survives restarts. XERJ serves it as a search endpoint.",
@@ -157,7 +157,7 @@
"canonical_url": "https://xerj.org/answers/store-agent-memory-without-llm-call",
"markdown_url": "https://xerj.org/answers/store-agent-memory-without-llm-call.md",
"cluster": "Agent memory: model dependency",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "Send dedup true on the store call and XERJ returns the existing id instead of a second document. The default lexical embedder needs no model; neural or proxy embedding is opt-in.",
"intent_group": "agent memory",
"answer": "How do I dedup agent memory so it stops filling with duplicate notes? Send dedup: true on the store call. XERJ compares the new memory against what the namespace already holds and returns the existing id with deduplicated: true instead of writing a second document. There is no generative summarization step; the configured embedder still runs.",
@@ -235,7 +235,7 @@
"canonical_url": "https://xerj.org/answers/coding-agent-memory-across-sessions",
"markdown_url": "https://xerj.org/answers/coding-agent-memory-across-sessions.md",
"cluster": "Agent memory: setup",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "XERJ keeps agent memory in a reserved index inside the data directory. We stored two memories, restarted the node, recalled both, then forgot one by id.",
"intent_group": "agent memory",
"answer": "How do people store user preferences so an agent can recall them next week? XERJ stores each memory as a document in a reserved index, .xerj-memory-{namespace}, inside the node's data directory. Our capture stored two memories, restarted the node on the same directory, recalled both by id, and one forget call removed exactly one.",
@@ -292,7 +292,7 @@
"canonical_url": "https://xerj.org/answers/batch-lookup-cve-ghsa-ids-offline",
"markdown_url": "https://xerj.org/answers/batch-lookup-cve-ghsa-ids-offline.md",
"cluster": "Answers for coding agents",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "Load a signed advisory pack and look up CVE, GHSA and RUSTSEC IDs offline with source citations. The recorded six-ID run answered four and abstained on two.",
"queries": [
"batch lookup CVE and GHSA IDs offline",
@@ -305,7 +305,7 @@
"canonical_url": "https://xerj.org/answers/full-text-search-over-jsonl-files-cli",
"markdown_url": "https://xerj.org/answers/full-text-search-over-jsonl-files-cli.md",
"cluster": "Answers for coding agents",
- "updated": "2026-10-20",
+ "updated": "2026-10-31",
"summary": "xerj autoindex reads JSONL and CSV with no schema, then answers BM25 per field from the CLI or MCP. 2,902 rows indexed in 12.4 s; queries took 1 ms.",
"queries": [
"full-text search over JSONL files from the CLI",
@@ -453,7 +453,7 @@
"canonical_url": "https://xerj.org/answers/reduce-embedding-api-cost",
"markdown_url": "https://xerj.org/answers/reduce-embedding-api-cost.md",
"cluster": "Embeddings: cost",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "XERJ embeds documents inside the node by default, so a captured run saw 0 non-loopback peers. Proxy mode sent exactly 2 requests for 100 documents.",
"intent_group": "retrieval research and eval",
"answer": "My embedding bill is too high; how do I reduce embedding API cost? Embed inside the search node. XERJ embeds semantic_text fields in-process by default, so the default configuration bills nothing to an embedding API: a capture of a default node saw 0 non-loopback peers in 3.51 s. Proxy mode sent exactly 2 requests for 100 documents.",
@@ -482,7 +482,7 @@
"canonical_url": "https://xerj.org/answers/do-search-embeddings-help",
"markdown_url": "https://xerj.org/answers/do-search-embeddings-help.md",
"cluster": "Embeddings: selection",
- "updated": "2026-10-19",
+ "updated": "2026-10-31",
"summary": "Embeddings only helped when we opted into the neural embedder. On XERJ's lexical default, kNN promoted coffee storage above the synonyms for the query car.",
"intent_group": "retrieval research and eval",
"answer": "Do I even need embeddings for search, or is full-text enough? On XERJ's default lexical embedder, full-text BM25 is close: 0.4167 precision at 3 against BM25's 0.375 on 8 judged queries. Embeddings clearly helped only after opting into --embed-mode neural, which moved precision at 3 to 0.6667 and recall to 1.0.",
@@ -1005,7 +1005,7 @@
"canonical_url": "https://xerj.org/answers/rag-without-vector-database",
"markdown_url": "https://xerj.org/answers/rag-without-vector-database.md",
"cluster": "Hybrid retrieval: architecture",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "Build a RAG retrieval layer on one XERJ index. One hybrid request fuses BM25 and kNN with Reciprocal Rank Fusion, and no second service runs.",
"intent_group": "retrieval research and eval",
"answer": "How do I build RAG without a vector database? Keep the inverted index and the dense vectors in one XERJ index and retrieve with one hybrid request, which fuses a BM25 sub-query and a kNN sub-query with Reciprocal Rank Fusion. No second service runs. Our 20-document capture returned d08, d04 and d10.",
@@ -1020,7 +1020,7 @@
"canonical_url": "https://xerj.org/answers/hybrid-search-quality-measured-ndcg",
"markdown_url": "https://xerj.org/answers/hybrid-search-quality-measured-ndcg.md",
"cluster": "Hybrid retrieval: quality",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "XERJ measured BM25, vectors and hybrid RRF on two public BEIR datasets. Hybrid won both: nDCG@10 0.6993 against 0.6572 on SciFact, with a neural embedder.",
"intent_group": "retrieval research and eval",
"answer": "How much better is hybrid search than BM25 alone? On two public BEIR datasets XERJ's hybrid RRF beat both BM25 and vectors: nDCG@10 0.6993 against 0.6572 on SciFact and 0.3448 against 0.3016 on NFCorpus. Those runs used --embed-mode neural with all-MiniLM-L6-v2; the default lexical embedder is not covered.",
@@ -1064,7 +1064,7 @@
"canonical_url": "https://xerj.org/answers/how-xerj-combines-search",
"markdown_url": "https://xerj.org/answers/how-xerj-combines-search.md",
"cluster": "Hybrid search",
- "updated": "2026-10-19",
+ "updated": "2026-10-31",
"summary": "One hybrid request carries a BM25 sub-query and a kNN sub-query against one index, and Reciprocal Rank Fusion merges their ranked lists into a single result list.",
"intent_group": "retrieval research and eval",
"answer": "How does XERJ combine BM25 and kNN? One hybrid request carries a BM25 sub-query and a kNN sub-query against a single index, and Reciprocal Rank Fusion merges the two ranked lists. Fusion reads rank positions, not scores, so the BM25 and cosine scales never need to be normalized against each other.",
@@ -1277,7 +1277,7 @@
"canonical_url": "https://xerj.org/answers/improve-basic-faiss-rag-pipeline",
"markdown_url": "https://xerj.org/answers/improve-basic-faiss-rag-pipeline.md",
"cluster": "RAG: troubleshooting",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "Move the FAISS chunk store into one XERJ index and fuse BM25 with vectors in a hybrid query. Two request shapes FAISS users try first return HTTP 400.",
"intent_group": "retrieval research and eval",
"answer": "How can I improve a basic FAISS RAG pipeline? Load the chunk store into one XERJ index with a semantic_text field and retrieve with the hybrid clause, which fuses BM25 and kNN, instead of vectors alone. Our 20-chunk migration indexed cleanly, and the two request shapes FAISS users write first both returned HTTP 400.",
@@ -1303,7 +1303,7 @@
"canonical_url": "https://xerj.org/answers/vector-database-vs-full-text-search",
"markdown_url": "https://xerj.org/answers/vector-database-vs-full-text-search.md",
"cluster": "Retrieval: selection",
- "updated": "2026-10-19",
+ "updated": "2026-10-31",
"summary": "Full-text search matches tokens, kNN matches vectors, hybrid fuses both. XERJ serves all three from one index over an indexed folder, so the choice is a clause.",
"intent_group": "retrieval research and eval",
"answer": "What's the actual difference between a vector database and full-text search? Full-text search matches tokens and ranks with BM25; a kNN query compares one embedding against stored dense vectors. XERJ holds both in one index, so the choice is which clause to send. On our 8 judged queries the three modes disagreed on the top hit once.",
diff --git a/landing/answers/rag-without-vector-database.html b/landing/answers/rag-without-vector-database.html
index 02be46a9d..a31568453 100644
--- a/landing/answers/rag-without-vector-database.html
+++ b/landing/answers/rag-without-vector-database.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -216,9 +216,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>HOW DO I BUILD RAG WITHOUT A VECTOR DATABASE?</div>
<div class="kicker"><span>ANSWERS · HYBRID RETRIEVAL: ARCHITECTURE</span></div>
<h1>How do I build RAG without a vector database?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>How do I build RAG without a vector database? Keep the inverted index and the dense vectors in one XERJ index and retrieve with one hybrid request, which fuses a BM25 sub-query and a kNN sub-query with Reciprocal Rank Fusion. No second service runs. Our 20-document capture returned d08, d04 and d10.</p>
<p class="key-fact"><strong>Key number:</strong> On BEIR SciFact, hybrid RRF with the opt-in neural embedder scored nDCG@10 0.6993, above both BM25 (0.6572) and vectors alone (0.6764). <span class="mute">(as of 2026-09-18 · source: <a href="https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/scifact.txt">benchmarks/beir-hybrid/results/scifact.txt</a>)</span></p>
</div>
@@ -322,6 +322,8 @@
<li><a href="/docs/recipes/hybrid-search">Hybrid search</a></li>
<li><a href="/answers/retrieval-that-says-not-found-with-file-line-citations">Retrieval that says not found, and cites file and line when it does answer</a></li>
<li><a href="/answers/agentic-memory-vs-rag">Agentic memory vs RAG: what is the difference?</a></li>
+ <li><a href="/answers/agentic-rag-vs-rag">Agentic RAG vs RAG: what is the difference?</a></li>
+ <li><a href="/answers/what-is-hybrid-search">What is hybrid search?</a></li>
<li><a href="/answers/mcp-api-search-or-vector-search">Should an agent use API search or vector search?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/how-xerj-combines-search">How does XERJ combine BM25 and kNN?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/hybrid-search-quality-measured-ndcg">How much better is hybrid search than BM25? Measured nDCG numbers</a> <span class="mute">· same intent: retrieval research and eval</span></li>
diff --git a/landing/answers/rag-without-vector-database.md b/landing/answers/rag-without-vector-database.md
index eafd39241..29b631521 100644
--- a/landing/answers/rag-without-vector-database.md
+++ b/landing/answers/rag-without-vector-database.md
@@ -2,13 +2,13 @@
---
title: "RAG without a vector database: one hybrid index"
canonical: "https://xerj.org/answers/rag-without-vector-database"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/rag-without-vector-database.md"
---
# How do I build RAG without a vector database?
-**Short answer** (updated 2026-10-28): How do I build RAG without a vector database? Keep the inverted index and the dense vectors in one XERJ index and retrieve with one hybrid request, which fuses a BM25 sub-query and a kNN sub-query with Reciprocal Rank Fusion. No second service runs. Our 20-document capture returned d08, d04 and d10.
+**Short answer** (updated 2026-10-31): How do I build RAG without a vector database? Keep the inverted index and the dense vectors in one XERJ index and retrieve with one hybrid request, which fuses a BM25 sub-query and a kNN sub-query with Reciprocal Rank Fusion. No second service runs. Our 20-document capture returned d08, d04 and d10.
**Key number:** On BEIR SciFact, hybrid RRF with the opt-in neural embedder scored nDCG@10 0.6993, above both BM25 (0.6572) and vectors alone (0.6764). (as of 2026-09-18; source: [benchmarks/beir-hybrid/results/scifact.txt](https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/scifact.txt))
@@ -162,6 +162,8 @@ One. XERJ is single-node, with no replication, no sharding and no failover, so t
- [Hybrid search](/docs/recipes/hybrid-search)
- [Retrieval that says not found, and cites file and line when it does answer](/answers/retrieval-that-says-not-found-with-file-line-citations)
- [Agentic memory vs RAG: what is the difference?](/answers/agentic-memory-vs-rag)
+- [Agentic RAG vs RAG: what is the difference?](/answers/agentic-rag-vs-rag)
+- [What is hybrid search?](/answers/what-is-hybrid-search)
- [Should an agent use API search or vector search?](/answers/mcp-api-search-or-vector-search) (same intent: retrieval research and eval)
- [How does XERJ combine BM25 and kNN?](/answers/how-xerj-combines-search) (same intent: retrieval research and eval)
- [How much better is hybrid search than BM25? Measured nDCG numbers](/answers/hybrid-search-quality-measured-ndcg) (same intent: retrieval research and eval)
diff --git a/landing/answers/reduce-embedding-api-cost.html b/landing/answers/reduce-embedding-api-cost.html
index 485343b74..95aa955e9 100644
--- a/landing/answers/reduce-embedding-api-cost.html
+++ b/landing/answers/reduce-embedding-api-cost.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -216,9 +216,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>HOW DO I REDUCE EMBEDDING API COST?</div>
<div class="kicker"><span>ANSWERS · EMBEDDINGS: COST</span></div>
<h1>How do I reduce embedding API cost?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>My embedding bill is too high; how do I reduce embedding API cost? Embed inside the search node. XERJ embeds semantic_text fields in-process by default, so the default configuration bills nothing to an embedding API: a capture of a default node saw 0 non-loopback peers in 3.51 s. Proxy mode sent exactly 2 requests for 100 documents.</p>
<p class="key-fact"><strong>Key number:</strong> Proxy embedding mode batched 100 documents into exactly 2 outbound requests in the captured run. <span class="mute">(as of 2026-08-23 · source: the captured run documented on this page)</span></p>
</div>
@@ -306,6 +306,7 @@
<li><a href="/answers/do-search-embeddings-help">When should I turn on neural embeddings, and when is the lexical default enough?</a></li>
<li><a href="/answers/local-embeddings-without-openai-api">I want local search that works offline, no cloud embeddings. What are my options?</a></li>
<li><a href="/compare/xerj-vs-vector-database">Should my agent call a search API or stand up Qdrant for persistent memory?</a></li>
+ <li><a href="/answers/classify-text-from-labelled-history-without-llm">Can I route tickets or filter spam from labelled history, without an LLM call?</a></li>
<li><a href="/answers/mcp-api-search-or-vector-search">Should an agent use API search or vector search?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/how-xerj-combines-search">How does XERJ combine BM25 and kNN?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/hybrid-search-quality-measured-ndcg">How much better is hybrid search than BM25? Measured nDCG numbers</a> <span class="mute">· same intent: retrieval research and eval</span></li>
diff --git a/landing/answers/reduce-embedding-api-cost.md b/landing/answers/reduce-embedding-api-cost.md
index ba903bb0a..b84b56724 100644
--- a/landing/answers/reduce-embedding-api-cost.md
+++ b/landing/answers/reduce-embedding-api-cost.md
@@ -2,13 +2,13 @@
---
title: "Reduce embedding API cost: embed inside the search engine"
canonical: "https://xerj.org/answers/reduce-embedding-api-cost"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/reduce-embedding-api-cost.md"
---
# How do I reduce embedding API cost?
-**Short answer** (updated 2026-10-28): My embedding bill is too high; how do I reduce embedding API cost? Embed inside the search node. XERJ embeds semantic_text fields in-process by default, so the default configuration bills nothing to an embedding API: a capture of a default node saw 0 non-loopback peers in 3.51 s. Proxy mode sent exactly 2 requests for 100 documents.
+**Short answer** (updated 2026-10-31): My embedding bill is too high; how do I reduce embedding API cost? Embed inside the search node. XERJ embeds semantic_text fields in-process by default, so the default configuration bills nothing to an embedding API: a capture of a default node saw 0 non-loopback peers in 3.51 s. Proxy mode sent exactly 2 requests for 100 documents.
**Key number:** Proxy embedding mode batched 100 documents into exactly 2 outbound requests in the captured run. (as of 2026-08-23; source: the captured run documented on this page)
@@ -138,6 +138,7 @@ Call GET /v1/embedding/identity on the native REST port. The backend field reads
- [When should I turn on neural embeddings, and when is the lexical default enough?](/answers/do-search-embeddings-help)
- [I want local search that works offline, no cloud embeddings. What are my options?](/answers/local-embeddings-without-openai-api)
- [Should my agent call a search API or stand up Qdrant for persistent memory?](/compare/xerj-vs-vector-database)
+- [Can I route tickets or filter spam from labelled history, without an LLM call?](/answers/classify-text-from-labelled-history-without-llm)
- [Should an agent use API search or vector search?](/answers/mcp-api-search-or-vector-search) (same intent: retrieval research and eval)
- [How does XERJ combine BM25 and kNN?](/answers/how-xerj-combines-search) (same intent: retrieval research and eval)
- [How much better is hybrid search than BM25? Measured nDCG numbers](/answers/hybrid-search-quality-measured-ndcg) (same intent: retrieval research and eval)
diff --git a/landing/answers/store-agent-memory-without-llm-call.html b/landing/answers/store-agent-memory-without-llm-call.html
index fd0cd6b42..6b1bea6ab 100644
--- a/landing/answers/store-agent-memory-without-llm-call.html
+++ b/landing/answers/store-agent-memory-without-llm-call.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -224,9 +224,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>HOW DO I STOP AN AGENT MEMORY STORE FROM FILLING WITH NEAR-DUPLICATE NOTES?</div>
<div class="kicker"><span>ANSWERS · AGENT MEMORY: MODEL DEPENDENCY</span></div>
<h1>How do I stop an agent memory store from filling with near-duplicate notes?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>How do I dedup agent memory so it stops filling with duplicate notes? Send dedup: true on the store call. XERJ compares the new memory against what the namespace already holds and returns the existing id with deduplicated: true instead of writing a second document. There is no generative summarization step; the configured embedder still runs.</p>
<p class="key-fact"><strong>Key number:</strong> The default dedup threshold is a cosine similarity of 0.95. <span class="mute">(as of 2026-10-19 · source: <a href="https://github.com/xerj-org/xerj/blob/main/engine/crates/xerj-api/src/memory_api.rs">engine/crates/xerj-api/src/memory_api.rs</a>)</span></p>
</div>
@@ -330,6 +330,8 @@
<li><a href="/answers/do-search-embeddings-help">When should I turn on neural embeddings, and when is the lexical default enough?</a></li>
<li><a href="/answers/local-embeddings-without-openai-api">I want local search that works offline, no cloud embeddings. What are my options?</a></li>
<li><a href="/compare/xerj-vs-agent-memory-tools">XERJ vs agent memory tools: Mem0, Zep, Letta and Cognee</a></li>
+ <li><a href="/compare/xerj-vs-mem0">XERJ vs Mem0: which memory fits your AI agent?</a></li>
+ <li><a href="/answers/what-is-an-mcp-memory-server">What is an MCP memory server?</a></li>
<li><a href="/answers/agentic-memory-vs-rag">Agentic memory vs RAG: what is the difference?</a> <span class="mute">· same intent: agent memory</span></li>
<li><a href="/answers/claude-code-memory">How does Claude Code memory work, and when do you need more?</a> <span class="mute">· same intent: agent memory</span></li>
<li><a href="/answers/coding-agent-memory-across-sessions">How do people store user preferences so an agent can recall them next week?</a> <span class="mute">· same intent: agent memory</span></li>
diff --git a/landing/answers/store-agent-memory-without-llm-call.md b/landing/answers/store-agent-memory-without-llm-call.md
index 0842bcc32..386e26a13 100644
--- a/landing/answers/store-agent-memory-without-llm-call.md
+++ b/landing/answers/store-agent-memory-without-llm-call.md
@@ -2,13 +2,13 @@
---
title: "Stop agent memory filling with duplicate notes"
canonical: "https://xerj.org/answers/store-agent-memory-without-llm-call"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/store-agent-memory-without-llm-call.md"
---
# How do I stop an agent memory store from filling with near-duplicate notes?
-**Short answer** (updated 2026-10-28): How do I dedup agent memory so it stops filling with duplicate notes? Send dedup: true on the store call. XERJ compares the new memory against what the namespace already holds and returns the existing id with deduplicated: true instead of writing a second document. There is no generative summarization step; the configured embedder still runs.
+**Short answer** (updated 2026-10-31): How do I dedup agent memory so it stops filling with duplicate notes? Send dedup: true on the store call. XERJ compares the new memory against what the namespace already holds and returns the existing id with deduplicated: true instead of writing a second document. There is no generative summarization step; the configured embedder still runs.
**Key number:** The default dedup threshold is a cosine similarity of 0.95. (as of 2026-10-19; source: [engine/crates/xerj-api/src/memory_api.rs](https://github.com/xerj-org/xerj/blob/main/engine/crates/xerj-api/src/memory_api.rs))
@@ -171,6 +171,8 @@ No. The stored `text` is the text you sent. Embedding creates a companion vector
- [When should I turn on neural embeddings, and when is the lexical default enough?](/answers/do-search-embeddings-help)
- [I want local search that works offline, no cloud embeddings. What are my options?](/answers/local-embeddings-without-openai-api)
- [XERJ vs agent memory tools: Mem0, Zep, Letta and Cognee](/compare/xerj-vs-agent-memory-tools)
+- [XERJ vs Mem0: which memory fits your AI agent?](/compare/xerj-vs-mem0)
+- [What is an MCP memory server?](/answers/what-is-an-mcp-memory-server)
- [Agentic memory vs RAG: what is the difference?](/answers/agentic-memory-vs-rag) (same intent: agent memory)
- [How does Claude Code memory work, and when do you need more?](/answers/claude-code-memory) (same intent: agent memory)
- [How do people store user preferences so an agent can recall them next week?](/answers/coding-agent-memory-across-sessions) (same intent: agent memory)
diff --git a/landing/answers/vector-database-vs-full-text-search.html b/landing/answers/vector-database-vs-full-text-search.html
index 3aff44449..4f552a3a3 100644
--- a/landing/answers/vector-database-vs-full-text-search.html
+++ b/landing/answers/vector-database-vs-full-text-search.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-21",
- "dateModified": "2026-10-19",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -224,9 +224,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>WHAT'S THE ACTUAL DIFFERENCE BETWEEN A VECTOR DATABASE AND FULL-TEXT SEARCH WHEN I HAVE A FOLDER?</div>
<div class="kicker"><span>ANSWERS · RETRIEVAL: SELECTION</span></div>
<h1 class="h1-long">What's the actual difference between a vector database and full-text search when I have a folder?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-19</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-21 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-19</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>What's the actual difference between a vector database and full-text search? Full-text search matches tokens and ranks with BM25; a kNN query compares one embedding against stored dense vectors. XERJ holds both in one index, so the choice is which clause to send. On our 8 judged queries the three modes disagreed on the top hit once.</p>
<p class="key-fact"><strong>Key number:</strong> On BEIR SciFact, BM25 scored nDCG@10 0.6572, vectors alone 0.6764, and hybrid RRF 0.6993. <span class="mute">(as of 2026-09-18 · source: <a href="https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/scifact.txt">benchmarks/beir-hybrid/results/scifact.txt</a>)</span></p>
</div>
@@ -335,6 +335,8 @@
<li><a href="/answers/do-search-embeddings-help">When should I turn on neural embeddings, and when is the lexical default enough?</a></li>
<li><a href="/answers/mcp-api-search-or-vector-search">Should an agent use API search or vector search?</a></li>
<li><a href="/compare/xerj-vs-vector-database">Should my agent call a search API or stand up Qdrant for persistent memory?</a></li>
+ <li><a href="/answers/bm25-vs-vector-search">BM25 vs vector search: which one should you use?</a></li>
+ <li><a href="/answers/what-is-hybrid-search">What is hybrid search?</a></li>
<li><a href="/answers/how-xerj-combines-search">How does XERJ combine BM25 and kNN?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/classify-text-from-labelled-history-without-llm">Can I route tickets or filter spam from labelled history, without an LLM call?</a> <span class="mute">· same intent: retrieval research and eval</span></li>
<li><a href="/answers/hybrid-search-quality-measured-ndcg">How much better is hybrid search than BM25? Measured nDCG numbers</a> <span class="mute">· same intent: retrieval research and eval</span></li>
diff --git a/landing/answers/vector-database-vs-full-text-search.md b/landing/answers/vector-database-vs-full-text-search.md
index 64cac2596..ee7661359 100644
--- a/landing/answers/vector-database-vs-full-text-search.md
+++ b/landing/answers/vector-database-vs-full-text-search.md
@@ -2,13 +2,13 @@
---
title: "Vector database or full-text search for a folder?"
canonical: "https://xerj.org/answers/vector-database-vs-full-text-search"
-updated: "2026-10-19"
+updated: "2026-10-31"
source: "content/answers/vector-database-vs-full-text-search.md"
---
# What's the actual difference between a vector database and full-text search when I have a folder?
-**Short answer** (updated 2026-10-19): What's the actual difference between a vector database and full-text search? Full-text search matches tokens and ranks with BM25; a kNN query compares one embedding against stored dense vectors. XERJ holds both in one index, so the choice is which clause to send. On our 8 judged queries the three modes disagreed on the top hit once.
+**Short answer** (updated 2026-10-31): What's the actual difference between a vector database and full-text search? Full-text search matches tokens and ranks with BM25; a kNN query compares one embedding against stored dense vectors. XERJ holds both in one index, so the choice is which clause to send. On our 8 judged queries the three modes disagreed on the top hit once.
**Key number:** On BEIR SciFact, BM25 scored nDCG@10 0.6572, vectors alone 0.6764, and hybrid RRF 0.6993. (as of 2026-09-18; source: [benchmarks/beir-hybrid/results/scifact.txt](https://github.com/xerj-org/xerj/blob/main/benchmarks/beir-hybrid/results/scifact.txt))
@@ -167,6 +167,8 @@ Only when seven conditions hold, including cosine similarity, no filter and at l
- [When should I turn on neural embeddings, and when is the lexical default enough?](/answers/do-search-embeddings-help)
- [Should an agent use API search or vector search?](/answers/mcp-api-search-or-vector-search)
- [Should my agent call a search API or stand up Qdrant for persistent memory?](/compare/xerj-vs-vector-database)
+- [BM25 vs vector search: which one should you use?](/answers/bm25-vs-vector-search)
+- [What is hybrid search?](/answers/what-is-hybrid-search)
- [How does XERJ combine BM25 and kNN?](/answers/how-xerj-combines-search) (same intent: retrieval research and eval)
- [Can I route tickets or filter spam from labelled history, without an LLM call?](/answers/classify-text-from-labelled-history-without-llm) (same intent: retrieval research and eval)
- [How much better is hybrid search than BM25? Measured nDCG numbers](/answers/hybrid-search-quality-measured-ndcg) (same intent: retrieval research and eval)
diff --git a/landing/answers/what-is-agentic-memory.html b/landing/answers/what-is-agentic-memory.html
index d03c1a6cf..9022fd757 100644
--- a/landing/answers/what-is-agentic-memory.html
+++ b/landing/answers/what-is-agentic-memory.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-10-28",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -216,9 +216,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/answers/">ANSWERS</a><span class="sep">·</span>WHAT IS AGENTIC MEMORY?</div>
<div class="kicker"><span>ANSWERS · AGENT MEMORY: CONCEPTS</span></div>
<h1>What is agentic memory?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-28 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-28 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>What is agentic memory? It is memory that an AI agent manages itself: the agent decides what to save, writes it through a tool call, and searches it again in a later task or session. It differs from a chat history because it is stored outside the model and survives restarts. XERJ serves it as a search endpoint.</p>
</div>
<p><strong>TL;DR</strong>: Agentic memory is memory an AI agent manages for itself. The agent decides what to keep, saves it with a tool call, and searches it again in a later task. Recall is a search problem, so XERJ serves agent memory as two endpoints on one local search engine.</p>
@@ -302,9 +302,13 @@
<li><a href="/answers/agent-memory-without-postgres-qdrant">What's a simple way to give an agent long-term memory on my laptop without Qdrant?</a></li>
<li><a href="/answers/store-agent-memory-without-llm-call">How do I stop an agent memory store from filling with near-duplicate notes?</a></li>
<li><a href="/docs/recipes/agentic-memory">Agent memory</a></li>
+ <li><a href="/answers/agentic-rag-vs-rag">Agentic RAG vs RAG: what is the difference?</a></li>
+ <li><a href="/compare/xerj-vs-mem0">XERJ vs Mem0: which memory fits your AI agent?</a></li>
+ <li><a href="/compare/xerj-vs-claude-mem-and-mempalace">claude-mem vs MemPalace vs XERJ: which memory should Claude Code use?</a></li>
+ <li><a href="/answers/claude-code-memory">How does Claude Code memory work, and when do you need more?</a></li>
<li><a href="/answers/coding-agent-memory-across-sessions">How do people store user preferences so an agent can recall them next week?</a> <span class="mute">· same intent: agent memory</span></li>
- <li><a href="/answers/claude-code-memory">How does Claude Code memory work, and when do you need more?</a> <span class="mute">· same intent: agent memory</span></li>
<li><a href="/answers/private-agent-memory-namespaces">How do I give two agents private memory on one laptop?</a> <span class="mute">· same intent: agent memory</span></li>
+ <li><a href="/answers/set-mcp-memory-storage-path">How do I set the MCP memory storage path?</a> <span class="mute">· same intent: agent memory</span></li>
</ul>
</main>
<nav class="docs-nav" aria-label="Article sections">
diff --git a/landing/answers/what-is-agentic-memory.md b/landing/answers/what-is-agentic-memory.md
index aefcc7bbc..080a37f22 100644
--- a/landing/answers/what-is-agentic-memory.md
+++ b/landing/answers/what-is-agentic-memory.md
@@ -2,13 +2,13 @@
---
title: "What is agentic memory? A plain answer"
canonical: "https://xerj.org/answers/what-is-agentic-memory"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/answers/what-is-agentic-memory.md"
---
# What is agentic memory?
-**Short answer** (updated 2026-10-28): What is agentic memory? It is memory that an AI agent manages itself: the agent decides what to save, writes it through a tool call, and searches it again in a later task or session. It differs from a chat history because it is stored outside the model and survives restarts. XERJ serves it as a search endpoint.
+**Short answer** (updated 2026-10-31): What is agentic memory? It is memory that an AI agent manages itself: the agent decides what to save, writes it through a tool call, and searches it again in a later task or session. It differs from a chat history because it is stored outside the model and survives restarts. XERJ serves it as a search endpoint.
**TL;DR**: Agentic memory is memory an AI agent manages for itself. The agent decides what to keep, saves it with a tool call, and searches it again in a later task. Recall is a search problem, so XERJ serves agent memory as two endpoints on one local search engine.
@@ -125,6 +125,10 @@ No. There is no decay and no TTL. A memory is permanent until explicitly forgott
- [What's a simple way to give an agent long-term memory on my laptop without Qdrant?](/answers/agent-memory-without-postgres-qdrant)
- [How do I stop an agent memory store from filling with near-duplicate notes?](/answers/store-agent-memory-without-llm-call)
- [Agent memory](/docs/recipes/agentic-memory)
+- [Agentic RAG vs RAG: what is the difference?](/answers/agentic-rag-vs-rag)
+- [XERJ vs Mem0: which memory fits your AI agent?](/compare/xerj-vs-mem0)
+- [claude-mem vs MemPalace vs XERJ: which memory should Claude Code use?](/compare/xerj-vs-claude-mem-and-mempalace)
+- [How does Claude Code memory work, and when do you need more?](/answers/claude-code-memory)
- [How do people store user preferences so an agent can recall them next week?](/answers/coding-agent-memory-across-sessions) (same intent: agent memory)
-- [How does Claude Code memory work, and when do you need more?](/answers/claude-code-memory) (same intent: agent memory)
- [How do I give two agents private memory on one laptop?](/answers/private-agent-memory-namespaces) (same intent: agent memory)
+- [How do I set the MCP memory storage path?](/answers/set-mcp-memory-storage-path) (same intent: agent memory)
diff --git a/landing/compare/best-grep-alternative-for-ai-coding-agents-2026.html b/landing/compare/best-grep-alternative-for-ai-coding-agents-2026.html
index 46d54aa78..8f053cc68 100644
--- a/landing/compare/best-grep-alternative-for-ai-coding-agents-2026.html
+++ b/landing/compare/best-grep-alternative-for-ai-coding-agents-2026.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/compare-best-grep-alternative-for-ai-coding-agents-2026.png",
"inLanguage": "en",
"datePublished": "2026-10-19",
- "dateModified": "2026-10-28",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -312,9 +312,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/compare/">COMPARISONS</a><span class="sep">·</span>THE BEST GREP AND RIPGREP ALTERNATIVES FOR AI CODING AGENTS IN 2026</div>
<div class="kicker"><span>COMPARE · COMPARISON: CODE RETRIEVAL</span></div>
<h1>The best grep and ripgrep alternatives for AI coding agents in 2026</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-19 · UPDATED · 2026-10-28</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-19 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-28</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>What is the best grep or ripgrep alternative for an AI coding agent in 2026? For one-off literal searches, none: ripgrep is still the default. ugrep adds Boolean and fuzzy queries, ast-grep and Semgrep match syntax, zoekt ranks over a trigram index, and XERJ returns ranked passages through MCP.</p>
<p class="key-fact"><strong>Key number:</strong> ripgrep has 69,257 GitHub stars and its latest release is 15.2.0 (2026-07-15), the most-starred tool on this list. <span class="mute">(as of 2026-10-19 · source: <a href="https://github.com/BurntSushi/ripgrep">github.com/BurntSushi/ripgrep</a>)</span></p>
<p class="mute">Last checked: 2026-10-19. Every third-party fact on this page was re-read from the tool's own docs or repository on that date.</p>
@@ -399,6 +399,7 @@
<li><a href="/compare/best-mcp-server-local-code-search-2026">The best MCP servers for searching a local codebase in 2026</a></li>
<li><a href="/compare/xerj-vs-localsynapse">What's a local MCP search engine for Claude?</a></li>
<li><a href="/answers/search-jupyter-notebook-cells">I need to find a function I wrote in a notebook months ago. How do I search all .ipynb files and the .py next to them?</a></li>
+ <li><a href="/compare/best-local-full-text-search-jsonl-csv-2026">The best local full-text search for JSONL and CSV files in 2026, with a published benchmark</a></li>
<li><a href="/answers/code-search-mcp-for-claude-code">I need my agent to find a passage and cite the file, not dump the whole tree. How?</a> <span class="mute">· same intent: code search for agents</span></li>
<li><a href="/answers/audit-php-codebase-with-an-index">How do I audit a PHP codebase with an index instead of grepping 1,400 files?</a> <span class="mute">· same intent: code search for agents</span></li>
<li><a href="/answers/index-reference-library-for-coding-agent">How do I set up reference coding so the agent retrieves a mechanism before writing code?</a> <span class="mute">· same intent: code search for agents</span></li>
diff --git a/landing/compare/best-grep-alternative-for-ai-coding-agents-2026.md b/landing/compare/best-grep-alternative-for-ai-coding-agents-2026.md
index 4b75f7b08..afa583de4 100644
--- a/landing/compare/best-grep-alternative-for-ai-coding-agents-2026.md
+++ b/landing/compare/best-grep-alternative-for-ai-coding-agents-2026.md
@@ -2,13 +2,13 @@
---
title: "Best grep alternative for AI coding agents (2026)"
canonical: "https://xerj.org/compare/best-grep-alternative-for-ai-coding-agents-2026"
-updated: "2026-10-28"
+updated: "2026-10-31"
source: "content/compare/best-grep-alternative-for-ai-coding-agents-2026.md"
---
# The best grep and ripgrep alternatives for AI coding agents in 2026
-**Short answer** (updated 2026-10-28): What is the best grep or ripgrep alternative for an AI coding agent in 2026? For one-off literal searches, none: ripgrep is still the default. ugrep adds Boolean and fuzzy queries, ast-grep and Semgrep match syntax, zoekt ranks over a trigram index, and XERJ returns ranked passages through MCP.
+**Short answer** (updated 2026-10-31): What is the best grep or ripgrep alternative for an AI coding agent in 2026? For one-off literal searches, none: ripgrep is still the default. ugrep adds Boolean and fuzzy queries, ast-grep and Semgrep match syntax, zoekt ranks over a trigram index, and XERJ returns ranked passages through MCP.
**Key number:** ripgrep has 69,257 GitHub stars and its latest release is 15.2.0 (2026-07-15), the most-starred tool on this list. (as of 2026-10-19; source: [github.com/BurntSushi/ripgrep](https://github.com/BurntSushi/ripgrep))
@@ -128,6 +128,7 @@ No head-to-head benchmark was run for this page. The ripgrep numbers come from t
- [The best MCP servers for searching a local codebase in 2026](/compare/best-mcp-server-local-code-search-2026)
- [What's a local MCP search engine for Claude?](/compare/xerj-vs-localsynapse)
- [I need to find a function I wrote in a notebook months ago. How do I search all .ipynb files and the .py next to them?](/answers/search-jupyter-notebook-cells)
+- [The best local full-text search for JSONL and CSV files in 2026, with a published benchmark](/compare/best-local-full-text-search-jsonl-csv-2026)
- [I need my agent to find a passage and cite the file, not dump the whole tree. How?](/answers/code-search-mcp-for-claude-code) (same intent: code search for agents)
- [How do I audit a PHP codebase with an index instead of grepping 1,400 files?](/answers/audit-php-codebase-with-an-index) (same intent: code search for agents)
- [How do I set up reference coding so the agent retrieves a mechanism before writing code?](/answers/index-reference-library-for-coding-agent) (same intent: code search for agents)
diff --git a/landing/compare/best-lightweight-elasticsearch-compatible-server-2026.html b/landing/compare/best-lightweight-elasticsearch-compatible-server-2026.html
index b0a2a8a71..f89e1cdc7 100644
--- a/landing/compare/best-lightweight-elasticsearch-compatible-server-2026.html
+++ b/landing/compare/best-lightweight-elasticsearch-compatible-server-2026.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/compare-best-lightweight-elasticsearch-compatible-server-2026.png",
"inLanguage": "en",
"datePublished": "2026-10-19",
- "dateModified": "2026-10-23",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -312,9 +312,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/compare/">COMPARISONS</a><span class="sep">·</span>THE BEST LIGHTWEIGHT ELASTICSEARCH-COMPATIBLE SEARCH SERVERS IN 2026</div>
<div class="kicker"><span>COMPARE · COMPARISON: SEARCH ENGINES</span></div>
<h1>The best lightweight Elasticsearch-compatible search servers in 2026</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-19 · UPDATED · 2026-10-23</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-10-19 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-23</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>What is the best lightweight Elasticsearch-compatible server in 2026? If you need the whole API and a cluster, OpenSearch. For one small node, ZincSearch or XERJ. For logs on object storage, Quickwit or OpenObserve. Manticore suits teams that prefer SQL. Each covers a different slice of the Elasticsearch API.</p>
<p class="key-fact"><strong>Key number:</strong> XERJ passes 1444 of 1454 Elasticsearch YAML conformance cases on every commit; OpenSearch, the full-API option, has 13,892 GitHub stars and released 3.9.0 on 2026-09-29. <span class="mute">(as of 2026-10-23 · source: <a href="https://github.com/xerj-org/xerj/blob/main/README.md">README.md</a>)</span></p>
<p class="mute">Last checked: 2026-10-19. Every third-party fact on this page was re-read from the tool's own docs or repository on that date.</p>
@@ -397,6 +397,7 @@
<li><a href="/compare/xerj-vs-meilisearch">What's a local alternative to Meilisearch that indexes a folder by itself?</a></li>
<li><a href="/compare/xerj-vs-typesense">Do I need Typesense for an agent to search my docs?</a></li>
<li><a href="/compare/best-local-full-text-search-jsonl-csv-2026">The best local full-text search for JSONL and CSV files in 2026, with a published benchmark</a></li>
+ <li><a href="/benchmarks/elasticsearch">ES scorecard</a></li>
<li><a href="/answers/does-xerj-support-s3-alerting-plugins">Does XERJ support S3 storage, alerting or custom plugins yet?</a> <span class="mute">· same intent: search engine alternatives</span></li>
<li><a href="/answers/search-engine-without-docker">Is there a single-binary local search engine I can just run?</a> <span class="mute">· same intent: search engine alternatives</span></li>
</ul>
diff --git a/landing/compare/best-lightweight-elasticsearch-compatible-server-2026.md b/landing/compare/best-lightweight-elasticsearch-compatible-server-2026.md
index 4afa219e4..63695f176 100644
--- a/landing/compare/best-lightweight-elasticsearch-compatible-server-2026.md
+++ b/landing/compare/best-lightweight-elasticsearch-compatible-server-2026.md
@@ -2,13 +2,13 @@
---
title: "Best lightweight Elasticsearch-compatible server (2026)"
canonical: "https://xerj.org/compare/best-lightweight-elasticsearch-compatible-server-2026"
-updated: "2026-10-23"
+updated: "2026-10-31"
source: "content/compare/best-lightweight-elasticsearch-compatible-server-2026.md"
---
# The best lightweight Elasticsearch-compatible search servers in 2026
-**Short answer** (updated 2026-10-23): What is the best lightweight Elasticsearch-compatible server in 2026? If you need the whole API and a cluster, OpenSearch. For one small node, ZincSearch or XERJ. For logs on object storage, Quickwit or OpenObserve. Manticore suits teams that prefer SQL. Each covers a different slice of the Elasticsearch API.
+**Short answer** (updated 2026-10-31): What is the best lightweight Elasticsearch-compatible server in 2026? If you need the whole API and a cluster, OpenSearch. For one small node, ZincSearch or XERJ. For logs on object storage, Quickwit or OpenObserve. Manticore suits teams that prefer SQL. Each covers a different slice of the Elasticsearch API.
**Key number:** XERJ passes 1444 of 1454 Elasticsearch YAML conformance cases on every commit; OpenSearch, the full-API option, has 13,892 GitHub stars and released 3.9.0 on 2026-09-29. (as of 2026-10-23; source: [README.md](https://github.com/xerj-org/xerj/blob/main/README.md))
@@ -124,5 +124,6 @@ No. Manticore, Quickwit and OpenObserve publish their own speed and cost compari
- [What's a local alternative to Meilisearch that indexes a folder by itself?](/compare/xerj-vs-meilisearch)
- [Do I need Typesense for an agent to search my docs?](/compare/xerj-vs-typesense)
- [The best local full-text search for JSONL and CSV files in 2026, with a published benchmark](/compare/best-local-full-text-search-jsonl-csv-2026)
+- [ES scorecard](/benchmarks/elasticsearch)
- [Does XERJ support S3 storage, alerting or custom plugins yet?](/answers/does-xerj-support-s3-alerting-plugins) (same intent: search engine alternatives)
- [Is there a single-binary local search engine I can just run?](/answers/search-engine-without-docker) (same intent: search engine alternatives)
diff --git a/landing/compare/index.html b/landing/compare/index.html
index d64f4280b..d45d7b39a 100644
--- a/landing/compare/index.html
+++ b/landing/compare/index.html
@@ -77,7 +77,7 @@
"url": "https://xerj.org/compare/",
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
- "dateModified": "2026-10-30",
+ "dateModified": "2026-10-31",
"isPartOf": {
"@id": "https://xerj.org/#website"
},
diff --git a/landing/compare/index.json b/landing/compare/index.json
index 685287463..ae2a4101c 100644
--- a/landing/compare/index.json
+++ b/landing/compare/index.json
@@ -74,7 +74,7 @@
"canonical_url": "https://xerj.org/compare/best-grep-alternative-for-ai-coding-agents-2026",
"markdown_url": "https://xerj.org/compare/best-grep-alternative-for-ai-coding-agents-2026.md",
"cluster": "Comparison: code retrieval",
- "updated": "2026-10-28",
+ "updated": "2026-10-31",
"summary": "ripgrep, ugrep, ast-grep, Semgrep, zoekt, ag and XERJ compared for coding agents: licence, install, ranking, index and MCP. Stars and releases checked 2026-10-19.",
"intent_group": "code search for agents",
"answer": "What is the best grep or ripgrep alternative for an AI coding agent in 2026? For one-off literal searches, none: ripgrep is still the default. ugrep adds Boolean and fuzzy queries, ast-grep and Semgrep match syntax, zoekt ranks over a trigram index, and XERJ returns ranked passages through MCP.",
@@ -246,7 +246,7 @@
"canonical_url": "https://xerj.org/compare/xerj-vs-elasticsearch-local",
"markdown_url": "https://xerj.org/compare/xerj-vs-elasticsearch-local.md",
"cluster": "Comparison: local search engines",
- "updated": "2026-10-23",
+ "updated": "2026-10-31",
"summary": "The published board is 55 wins, 26 ties, 4 losses and 3 not-applicable across 88 cells against Elasticsearch 8.13.4. All 4 losses are read p99 under a write flood.",
"intent_group": "search engine alternatives",
"answer": "Is there an Elasticsearch-compatible engine that isn't Elasticsearch? XERJ speaks the Elasticsearch REST API from one binary. One published board measured both on one host: 55 wins, 26 ties, 4 losses and 3 not-applicable across 88 cells against Elasticsearch 8.13.4 at 100,000 documents. XERJ is single-node only.",
@@ -290,7 +290,7 @@
"canonical_url": "https://xerj.org/compare/best-lightweight-elasticsearch-compatible-server-2026",
"markdown_url": "https://xerj.org/compare/best-lightweight-elasticsearch-compatible-server-2026.md",
"cluster": "Comparison: search engines",
- "updated": "2026-10-23",
+ "updated": "2026-10-31",
"summary": "OpenSearch, ZincSearch, Quickwit, Manticore, OpenObserve, Elasticsearch and XERJ compared on licence, install, ES API coverage and scaling. Checked 2026-10-19.",
"intent_group": "search engine alternatives",
"answer": "What is the best lightweight Elasticsearch-compatible server in 2026? If you need the whole API and a cluster, OpenSearch. For one small node, ZincSearch or XERJ. For logs on object storage, Quickwit or OpenObserve. Manticore suits teams that prefer SQL. Each covers a different slice of the Elasticsearch API.",
diff --git a/landing/compare/xerj-vs-elasticsearch-local.html b/landing/compare/xerj-vs-elasticsearch-local.html
index b4df1ca5a..be7d992bc 100644
--- a/landing/compare/xerj-vs-elasticsearch-local.html
+++ b/landing/compare/xerj-vs-elasticsearch-local.html
@@ -79,7 +79,7 @@
"image": "https://xerj.org/og/xerj-card.png",
"inLanguage": "en",
"datePublished": "2026-08-22",
- "dateModified": "2026-10-23",
+ "dateModified": "2026-10-31",
"author": {
"@type": "Organization",
"name": "XERJ documentation team",
@@ -232,9 +232,9 @@
<div class="breadcrumb"><a href="/">HOME</a><span class="sep">·</span><a href="/compare/">COMPARISONS</a><span class="sep">·</span>IS THERE AN ELASTICSEARCH-COMPATIBLE ENGINE THAT ISN'T ELASTICSEARCH?</div>
<div class="kicker"><span>COMPARE · COMPARISON: LOCAL SEARCH ENGINES</span></div>
<h1>Is there an Elasticsearch-compatible engine that isn't Elasticsearch?</h1>
- <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-22 · UPDATED · 2026-10-23</div>
+ <div class="source-note">AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-08-22 · UPDATED · 2026-10-31</div>
<div class="prompt-card answer-first" id="short-answer">
- <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-23</span>
+ <span class="prompt-label">SHORT ANSWER · UPDATED 2026-10-31</span>
<p>Is there an Elasticsearch-compatible engine that isn't Elasticsearch? XERJ speaks the Elasticsearch REST API from one binary. One published board measured both on one host: 55 wins, 26 ties, 4 losses and 3 not-applicable across 88 cells against Elasticsearch 8.13.4 at 100,000 documents. XERJ is single-node only.</p>
<p class="key-fact"><strong>Key number:</strong> Elasticsearch has 78,319 GitHub stars and its latest release is v9.5.5 (2026-10-06); the board was measured against 8.13.4. <span class="mute">(as of 2026-10-19 · source: <a href="https://github.com/elastic/elasticsearch">github.com/elastic/elasticsearch</a>)</span></p>
</div>
@@ -353,6 +353,7 @@
<li><a href="/docs/recipes/migrate-from-elasticsearch">Migrate from Elasticsearch</a></li>
<li><a href="/answers/full-text-search-over-jsonl-files-cli">Full-text search over JSONL files from the CLI: index a folder and query it with BM25</a></li>
<li><a href="/answers/batch-lookup-cve-ghsa-ids-offline">Batch lookup CVE IDs and GHSA IDs offline, without GitHub API rate limits</a></li>
+ <li><a href="/benchmarks/elasticsearch">ES scorecard</a></li>
<li><a href="/answers/does-xerj-support-s3-alerting-plugins">Does XERJ support S3 storage, alerting or custom plugins yet?</a> <span class="mute">· same intent: search engine alternatives</span></li>
<li><a href="/compare/best-lightweight-elasticsearch-compatible-server-2026">The best lightweight Elasticsearch-compatible search servers in 2026</a> <span class="mute">· same intent: search engine alternatives</span></li>
<li><a href="/compare/xerj-vs-meilisearch">What's a local alternative to Meilisearch that indexes a folder by itself?</a> <span class="mute">· same intent: search engine alternatives</span></li>
diff --git a/landing/compare/xerj-vs-elasticsearch-local.md b/landing/compare/xerj-vs-elasticsearch-local.md
index 523d27935..afe043c97 100644
--- a/landing/compare/xerj-vs-elasticsearch-local.md
+++ b/landing/compare/xerj-vs-elasticsearch-local.md
@@ -2,13 +2,13 @@
---
title: "Elasticsearch-compatible engine that isn't Elasticsearch"
canonical: "https://xerj.org/compare/xerj-vs-elasticsearch-local"
-updated: "2026-10-23"
+updated: "2026-10-31"
source: "content/compare/xerj-vs-elasticsearch-local.md"
---
# Is there an Elasticsearch-compatible engine that isn't Elasticsearch?
-**Short answer** (updated 2026-10-23): Is there an Elasticsearch-compatible engine that isn't Elasticsearch? XERJ speaks the Elasticsearch REST API from one binary. One published board measured both on one host: 55 wins, 26 ties, 4 losses and 3 not-applicable across 88 cells against Elasticsearch 8.13.4 at 100,000 documents. XERJ is single-node only.
+**Short answer** (updated 2026-10-31): Is there an Elasticsearch-compatible engine that isn't Elasticsearch? XERJ speaks the Elasticsearch REST API from one binary. One published board measured both on one host: 55 wins, 26 ties, 4 losses and 3 not-applicable across 88 cells against Elasticsearch 8.13.4 at 100,000 documents. XERJ is single-node only.
**Key number:** Elasticsearch has 78,319 GitHub stars and its latest release is v9.5.5 (2026-10-06); the board was measured against 8.13.4. (as of 2026-10-19; source: [github.com/elastic/elasticsearch](https://github.com/elastic/elasticsearch))
@@ -198,6 +198,7 @@ When you read while a heavy writer runs, when you need more than one host, or wh
- [Migrate from Elasticsearch](/docs/recipes/migrate-from-elasticsearch)
- [Full-text search over JSONL files from the CLI: index a folder and query it with BM25](/answers/full-text-search-over-jsonl-files-cli)
- [Batch lookup CVE IDs and GHSA IDs offline, without GitHub API rate limits](/answers/batch-lookup-cve-ghsa-ids-offline)
+- [ES scorecard](/benchmarks/elasticsearch)
- [Does XERJ support S3 storage, alerting or custom plugins yet?](/answers/does-xerj-support-s3-alerting-plugins) (same intent: search engine alternatives)
- [The best lightweight Elasticsearch-compatible search servers in 2026](/compare/best-lightweight-elasticsearch-compatible-server-2026) (same intent: search engine alternatives)
- [What's a local alternative to Meilisearch that indexes a folder by itself?](/compare/xerj-vs-meilisearch) (same intent: search engine alternatives)
diff --git a/landing/docs/recipes/agentic-memory.html b/landing/docs/recipes/agentic-memory.html
index c3f7e3e5d..8a7aed01a 100644
--- a/landing/docs/recipes/agentic-memory.html
+++ b/landing/docs/recipes/agentic-memory.html
@@ -384,6 +384,15 @@ All assertions passed.</pre>
<li><strong>Wire-compatible.</strong> Because a namespace is a real index, you can inspect or operate on it with the standard ES-compatible surface if you ever need to — the memory API is a thin, honest adapter over the same machinery.</li>
</ul>
+ <h3>Related reading</h3>
+ <ul>
+ <li><a href="/answers/what-is-an-mcp-memory-server">What is an MCP memory server?</a></li>
+ <li><a href="/answers/agentic-memory-vs-rag">Agentic memory compared with RAG</a></li>
+ <li><a href="/answers/claude-code-memory">How Claude Code memory works</a></li>
+ <li><a href="/compare/xerj-vs-agent-memory-tools">XERJ compared with Mem0, Zep, Letta and Cognee</a></li>
+ <li><a href="/compare/xerj-vs-mem0">XERJ compared with Mem0</a></li>
+ </ul>
+
<h3>Before you trust a recall number</h3>
<p>The recall@1 and recall@3 above come from six hand-labelled probes. That is a smoke test. Before you use a recall number to choose an embedder, a chunking scheme or a recall path for your own agent, apply these field lessons. They were contributed by Chunxiao Wang (Assay / <a href="https://github.com/chunxiaoxx/nautilus-compass">nautilus-compass</a>) in <a href="https://github.com/xerj-org/xerj/issues/1138">#1138</a>, from a production agent-memory system, and are written up with their measurements in <a href="/blog/eval-scars-from-agent-memory">Six eval scars from a production agent-memory system</a>.</p>
<ul>
diff --git a/landing/docs/recipes/hybrid-search.html b/landing/docs/recipes/hybrid-search.html
index 94030fefa..a85670b83 100644
--- a/landing/docs/recipes/hybrid-search.html
+++ b/landing/docs/recipes/hybrid-search.html
@@ -368,6 +368,8 @@ OK</pre>
<li><strong>Vector quality is your embeddings' job.</strong> This recipe supplies vectors directly for a deterministic demo. For real semantic recall, generate <code class="inline">dense_vector</code> values with a production embedding model, or use a <code class="inline">semantic_text</code> field with an external <code class="inline">inference_id</code>.</li>
</ul>
+ <p>For the concepts behind this recipe, read <a href="/answers/what-is-hybrid-search">what hybrid search is</a> and <a href="/answers/bm25-vs-vector-search">BM25 compared with vector search</a>.</p>
+
<!-- Print-only branded footer (hidden on screen, visible in PDF) -->
<div class="print-footer" style="display:none;">
<span>XERJ.AI · V1.0.0-RC.100 · 2026-08-11</span>
diff --git a/landing/index.html b/landing/index.html
index 9367da65a..4bb6fb8a5 100644
--- a/landing/index.html
+++ b/landing/index.html
@@ -423,6 +423,8 @@ $ curl -sXPOST localhost:9200/_memory/agent-demo/_recall \
</a>
</div>
+ <p class="wide">More on memory: <a href="/answers/what-is-an-mcp-memory-server">what an MCP memory server is</a>, <a href="/answers/agentic-memory-vs-rag">agentic memory and RAG</a>, <a href="/answers/claude-code-memory">Claude Code memory</a> and <a href="/compare/xerj-vs-agent-memory-tools">XERJ compared with Mem0, Zep, Letta and Cognee</a>.</p>
+
<p class="viz-caption"><a class="mono" href="/for-agents" style="color:var(--z-mute);">THE FULL AGENT SURFACE — SEVEN OPERATIONS + LLMS.TXT →</a></p>
</section>
@@ -449,7 +451,7 @@ $ curl -sXPOST localhost:9200/_memory/agent-demo/_recall \
</div>
<div class="row">
<div class="k">Hybrid in one pass</div>
- <div class="v">50 query types, machine-checked against the parser dispatch table, including native BM25 + kNN fusion (RRF) in a single query tree — no bolt-on vector database, no fusion glue between two systems</div>
+ <div class="v">50 query types, machine-checked against the parser dispatch table, including native BM25 + kNN fusion (RRF) in a single query tree — no bolt-on vector database, no fusion glue between two systems. Background: <a href="/answers/what-is-hybrid-search">what hybrid search is</a> and <a href="/answers/bm25-vs-vector-search">BM25 compared with vector search</a></div>
</div>
<div class="row">
<div class="k">Semantic without setup</div>
@@ -469,7 +471,7 @@ $ curl -sXPOST localhost:9200/_memory/agent-demo/_recall \
</div>
<div class="row">
<div class="k">One static binary</div>
- <div class="v">Rust, no JVM, sub-second start, Apache-2.0 — laptop to server with the same artifact. Benchmarks with wins <em>and</em> losses published at <a href="/benchmarks/">xerj.org/benchmarks</a></div>
+ <div class="v">Rust, no JVM, sub-second start, Apache-2.0 — laptop to server with the same artifact. Benchmarks with wins <em>and</em> losses published at <a href="/benchmarks/">xerj.org/benchmarks</a>, including the <a href="/benchmarks/elasticsearch">full Elasticsearch scorecard</a></div>
</div>
</section>
diff --git a/landing/use-cases/code-security-audit.html b/landing/use-cases/code-security-audit.html
index 595157794..72b9d0395 100644
--- a/landing/use-cases/code-security-audit.html
+++ b/landing/use-cases/code-security-audit.html
@@ -578,6 +578,7 @@ Host: site.victim.network.example
<section class="scene-section">
<h2 class="scene" style="font-size:var(--fs-56);">READ THE <span class="accent">CASE STUDY.</span></h2>
<p class="wide">Everything above is published: the findings with their honest severities, the reproduction commands, the counter-examples where the index lost, and the verbatim prompts that drove the agent. Setup is roughly two minutes — install the extractor, clone WordPress, run three index scripts — and the method ships as a copyable skill that retargets to another stack by editing the catalog, because the taint model is data, not code.</p>
+ <p class="wide">Read the full write-up of the <a href="/case-studies/wordpress-security-audit">WordPress security audit</a> on this site.</p>
<div style="display:flex; gap:var(--sp-6); margin-top:var(--sp-6); flex-wrap:wrap;">
<a href="https://github.com/xerj-org/xerj/tree/main/docs/case-studies/wordpress-security-audit" rel="external" class="cta" style="font-size:var(--fs-20); padding:var(--sp-4) 0; border-bottom:0; flex:1; min-width:280px;">
<span>OPEN THE CASE STUDY</span>