Skip to content

armgit / xerj

Public Imported from GitHub
0

AI search engine and MCP server for bots and agents. One Rust binary indexes code, docs, logs, JSONL/CSV and PDFs for BM25 full-text, vector and hybrid search, code search without grep, agent memory and RAG. Elasticsearch-compatible API; Corpus Hub serves hundreds of reference memory corpora.

docs(website): publish the Rust self-audit case study + measured token economics

xerj-orgdc2f13ed441e
diff --git a/docs/case-studies/xerj-self-audit/FINDINGS.md b/docs/case-studies/xerj-self-audit/FINDINGS.md
index f8e56a020..b8667bf3d 100644
--- a/docs/case-studies/xerj-self-audit/FINDINGS.md
+++ b/docs/case-studies/xerj-self-audit/FINDINGS.md
@@ -5,7 +5,9 @@
### `POST /_sql` unauthenticated stack-overflow → process abort
**Severity:** Critical (unauthenticated, single request, whole-process abort).
-**Status:** confirmed by crashing a live server. **Not fixed** on current `main`.
+**Status:** confirmed by crashing a live server; **fixed** on branch
+`feat/rust-self-audit` (commit `657db52`) and the fix proven against the same
+payload — see "The fix" below.
**The code.** `engine/crates/xerj-engine/src/sql.rs`, the WHERE-clause
recursive-descent parser:
@@ -59,10 +61,23 @@ fatal runtime error: stack overflow, aborting
A stack-overflow abort is not a catchable panic — `catch_unwind` cannot save it,
and the whole server (every tenant, every index) goes down.
-**Suggested fix.** Enter a `DepthGuard` at the top of `parse_condition` (and/or
-`parse_or_expr`), returning a parse error past `MAX_QUERY_DEPTH`, exactly as PR
-#69 did for `parse_qs_unary`. A parse-time token-count/paren-count cap on the raw
-SQL string is a cheap defense-in-depth second layer.
+**The fix (shipped).** An explicit `depth` parameter is threaded through
+`parse_or_expr` / `parse_and_expr` / `parse_condition`, rejected past
+`MAX_SQL_DEPTH = 64` (mirroring `xerj-query`'s `MAX_QUERY_DEPTH`). Proof on the
+rebuilt binary:
+
+```
+# baseline still works
+POST /_sql {"query":"SELECT * FROM t WHERE (a = 1)"} → HTTP 200
+# the exact payload that aborted the unpatched server
+POST /_sql {"query":"SELECT * FROM t WHERE ((…50k…))"} → HTTP 400 in ~2ms
+ {"reason":"SQL parse error: WHERE clause nesting exceeds max depth of 64"}
+# server still answering afterwards
+GET / → HTTP 200
+```
+
+Three regression tests (deep parens, deep `NOT`, moderate nesting still parses)
+plus the full `xerj-engine` lib suite pass; `fmt` + `clippy -D warnings` clean.
---
diff --git a/docs/case-studies/xerj-self-audit/README.md b/docs/case-studies/xerj-self-audit/README.md
index 9cbcd7e89..6f9a738a1 100644
--- a/docs/case-studies/xerj-self-audit/README.md
+++ b/docs/case-studies/xerj-self-audit/README.md
@@ -42,16 +42,21 @@ the code is fixed, the query goes quiet.
| | Agent reads all files | XERJ-assisted audit |
|---|---|---|
-| **Tokens to load the codebase once** | ~2,000,000 (approx: 8.0 MB source ÷ 4) | index once, then read only flagged functions |
-| **Fits one context?** | No — ~10× a 200k window; must chunk | Yes — targeted queries return tens of candidates |
+| **Tokens to reach the `/_sql` finding** | **1,805,277** (every `.rs` under `engine/`) | **10,533** (cycle-finder output + 4 parser fns read + route lookup) |
+| **Ratio** | — | **171× fewer tokens**, measured, to the same bug |
+| **Fits one context?** | No — ~9× a 200k window; must chunk | Yes — targeted queries return tens of candidates |
| **Interprocedural reach** | Lost at chunk boundaries | 87,137 call edges preserved across all files |
| **Finds a recursion *cycle*?** | Only if the whole cycle lands in one chunk | One query over the call graph |
-| **Grounded?** | Skimming 2M tokens → guesses | Every finding cites a real `file:line`, verified by reading it |
+| **Grounded?** | Skimming 1.8M tokens → guesses | Every finding cites a real `file:line`, verified by reading it |
| **Reproducible?** | No | Yes — extractor + queries reproduce exactly |
-Substrate build: **~3.2s extract + ~1.9s ingest** for the whole workspace; 31 MB
-on disk. The token figure is a labelled approximation (source bytes ÷ 4), not a
-measured tokenizer count.
+Both token figures are **measured with tiktoken (`cl100k_base`)**, a public
+GPT-4-class BPE used as an LLM-token proxy — not a chars÷4 estimate. Read-all =
+every `.rs` file under `engine/` tokenized (197 files, 8,025,201 bytes). The
+XERJ-assisted figure is the actual path to *this* finding: the cycle-finder's
+stdout (7,511 tok) + the four SQL parser function bodies the auditor read (3,002
+tok) + the route lookup (20 tok). Substrate build: **~3.2s extract + ~1.9s
+ingest**, 31 MB on disk.
## Coverage (the explicit goal was 100%)
diff --git a/docs/examples/rust-ast-audit/sample-token-economics.json b/docs/examples/rust-ast-audit/sample-token-economics.json
new file mode 100644
index 000000000..2eaf355c6
--- /dev/null
+++ b/docs/examples/rust-ast-audit/sample-token-economics.json
@@ -0,0 +1,14 @@
+{
+ "without_xerj_tokens": 1805277,
+ "without_xerj_files": 197,
+ "without_xerj_bytes": 8025201,
+ "with_xerj_tokens": 10533,
+ "with_xerj_breakdown": {
+ "cycle_finder": 7511,
+ "read_functions": 3002,
+ "n_functions_read": 4,
+ "route_lookup": 20
+ },
+ "ratio": 171.4,
+ "tokenizer": "tiktoken cl100k_base (public GPT-4-class BPE, used as an LLM-token proxy)"
+}
\ No newline at end of file
diff --git a/landing/case-studies.html b/landing/case-studies.html
index 162c9b4ff..8d10d2162 100644
--- a/landing/case-studies.html
+++ b/landing/case-studies.html
@@ -56,6 +56,7 @@
<nav class="uc-subnav" aria-label="Case studies">
<a href="/case-studies.html" class="active">ALL CASE STUDIES</a>
+ <a href="/case-studies/xerj-self-audit.html">RUST SELF-AUDIT</a>
<a href="/case-studies/wordpress-security-audit.html">WORDPRESS SECURITY AUDIT</a>
<a href="/case-studies/calltree-analytics.html">SEMANTIC ANALYTICS</a>
<a href="/case-studies/daily-dev-postgres-cdc.html">POSTGRES CDC + HYBRID</a>
@@ -65,20 +66,53 @@
<div class="kicker"><span>CASE STUDIES</span><span class="dash">·</span><span>REPRODUCIBLE WALKTHROUGHS</span></div>
<h1 class="hero">REAL QUESTIONS.<br><span class="accent">RUNNABLE ANSWERS.</span></h1>
<p class="lead" style="max-width:720px;">
- WordPress core — ~619k lines — security-audited by an AI agent for ~26,000
- tokens: that is the one study measured on real code. The other two are
+ XERJ audited its own Rust engine and found a real unauthenticated
+ process-abort DoS — reached for 10,533 tokens instead of the 1.8M it takes to
+ read the source once. WordPress core — ~619k lines — was security-audited for
+ ~26,000 tokens. Those two are measured on real code; the other two are
functional proofs on small synthetic corpora — they show the shape works,
- not how fast.
- Three engineering questions, answered end to end with published code and
- recorded output. The questions came from calltree.ai and daily.dev; neither
- is a customer. Every number traces to a walkthrough in the repo.
+ not how fast. Every number traces to a walkthrough in the repo.
</p>
</section>
-<!-- 01 WORDPRESS SECURITY AUDIT -->
+<!-- 01 RUST SELF-AUDIT -->
<div class="uc-showcase">
<div class="uc-copy">
<span class="uc-num">01</span>
+ <div class="uc-title">XERJ AUDITS<br>ITS OWN<br><span class="accent">RUST.</span></div>
+ <p class="uc-pitch">XERJ indexed its own engine — 197 files, ~201k lines — as a queryable AST and call graph, then found an unauthenticated <code class="inline">POST /_sql</code> stack-overflow that aborts the process. A call-graph recursion-cycle query grep cannot express. Yellow is the tokens to reach the finding; blue is what it costs to read the source once.</p>
+ <div class="uc-metric">10,533 TOKENS INSTEAD OF 1.8M · BUG FIXED</div>
+ <div class="uc-persona">SECURITY ENGINEERS · AI-AGENT BUILDERS · RUST TEAMS</div>
+ <a href="/case-studies/xerj-self-audit.html" class="uc-cta">FULL CASE STUDY →</a>
+ </div>
+ <div class="uc-viz">
+ <div class="kicker"><span>MEASURED · tiktoken cl100k_base · TOKENS TO THE /_sql FINDING</span></div>
+ <div class="cmp-bars">
+ <div class="cmp-bars-head">
+ <span></span>
+ <span class="cmp-a">XERJ-ASSISTED</span>
+ <span class="cmp-b">READ EVERY FILE</span>
+ </div>
+ <div class="cmp-bars-row">
+ <span class="cmp-label">Tokens to the finding</span>
+ <span class="cmp-val cmp-a">
+ <span class="cmp-bar" style="width:0.5%;"></span>
+ <span class="cmp-num">10,533</span>
+ </span>
+ <span class="cmp-val cmp-b">
+ <span class="cmp-bar" style="width:90.0%;"></span>
+ <span class="cmp-num">1,805,277 <span class="cmp-ratio">171×</span></span>
+ </span>
+ </div>
+ </div>
+ <p style="margin-top:14px; font-size:12px; letter-spacing:.04em; color:var(--z-mute); font-family:var(--font-data);">TOKEN ACCOUNTING, NOT A LATENCY BENCHMARK · SOURCES IN <a href="https://github.com/xerj-org/xerj/tree/main/docs/case-studies/xerj-self-audit" rel="external">docs/case-studies/xerj-self-audit &rarr;</a></p>
+ </div>
+</div>
+
+<!-- 02 WORDPRESS SECURITY AUDIT -->
+<div class="uc-showcase">
+ <div class="uc-copy">
+ <span class="uc-num">02</span>
<div class="uc-title">WORDPRESS<br>SECURITY<br><span class="accent">AUDIT.</span></div>
<p class="uc-pitch">WordPress core is ~619k lines — ~26× a 200k-token context window. An agent indexed it as a queryable code graph and read only what queries surfaced. Yellow is the whole audit's token bill; blue is what it costs just to load the code once.</p>
<div class="uc-metric">~26K TOKENS INSTEAD OF ~5.2M</div>
@@ -109,10 +143,10 @@
</div>
</div>
-<!-- 02 SEMANTIC ANALYTICS -->
+<!-- 03 SEMANTIC ANALYTICS -->
<div class="uc-showcase">
<div class="uc-copy">
- <span class="uc-num">02</span>
+ <span class="uc-num">03</span>
<div class="uc-title">SEMANTIC<br>ANALYTICS IN<br><span class="accent">ONE REQUEST.</span></div>
<p class="uc-pitch">"Of calls about wifi issues, how many are 2.4GHz vs 5GHz, and what are the key issues?" — asked by calltree.ai. Half retrieval, half analytics: normally two stores and an ETL. One XERJ <code class="inline">POST /_search</code> returns the slice, the breakdown, the metrics, and the evidence.</p>
<div class="uc-metric">1 REQUEST · 1 STORE — WAS 2 + ETL</div>
@@ -161,10 +195,10 @@
</div>
</div>
-<!-- 03 POSTGRES CDC + HYBRID SEARCH -->
+<!-- 04 POSTGRES CDC + HYBRID SEARCH -->
<div class="uc-showcase">
<div class="uc-copy">
- <span class="uc-num">03</span>
+ <span class="uc-num">04</span>
<div class="uc-title">POSTGRES CDC +<br><span class="accent">HYBRID SEARCH.</span></div>
<p class="uc-pitch">daily.dev's founder asked: can XERJ replicate automatically from Postgres? A logical-replication slot streams every INSERT, UPDATE, and DELETE into XERJ — no reindex jobs — and one hybrid query fuses BM25 + vector rankings server-side, replacing the tsvector + pgvector two-query app merge.</p>
<div class="uc-metric">3/3 MUTATION TYPES REPLICATED · 0 CHANGES LOST</div>
diff --git a/landing/case-studies/calltree-analytics.html b/landing/case-studies/calltree-analytics.html
index 30a91e547..ac92d14d2 100644
--- a/landing/case-studies/calltree-analytics.html
+++ b/landing/case-studies/calltree-analytics.html
@@ -52,6 +52,7 @@
<nav class="uc-subnav" aria-label="Case studies">
<a href="/case-studies.html">ALL CASE STUDIES</a>
+ <a href="/case-studies/xerj-self-audit.html">RUST SELF-AUDIT</a>
<a href="/case-studies/wordpress-security-audit.html">WORDPRESS SECURITY AUDIT</a>
<a href="/case-studies/calltree-analytics.html" class="active">SEMANTIC ANALYTICS</a>
<a href="/case-studies/daily-dev-postgres-cdc.html">POSTGRES CDC + HYBRID</a>
diff --git a/landing/case-studies/daily-dev-postgres-cdc.html b/landing/case-studies/daily-dev-postgres-cdc.html
index 31f52c8ad..b4f472168 100644
--- a/landing/case-studies/daily-dev-postgres-cdc.html
+++ b/landing/case-studies/daily-dev-postgres-cdc.html
@@ -53,6 +53,7 @@
<nav class="uc-subnav" aria-label="Case studies">
<a href="/case-studies.html">ALL CASE STUDIES</a>
+ <a href="/case-studies/xerj-self-audit.html">RUST SELF-AUDIT</a>
<a href="/case-studies/wordpress-security-audit.html">WORDPRESS SECURITY AUDIT</a>
<a href="/case-studies/calltree-analytics.html">SEMANTIC ANALYTICS</a>
<a href="/case-studies/daily-dev-postgres-cdc.html" class="active">POSTGRES CDC + HYBRID</a>
diff --git a/landing/case-studies/wordpress-security-audit.html b/landing/case-studies/wordpress-security-audit.html
index 40a64b965..890d452ec 100644
--- a/landing/case-studies/wordpress-security-audit.html
+++ b/landing/case-studies/wordpress-security-audit.html
@@ -49,6 +49,7 @@
<nav class="uc-subnav" aria-label="Case studies">
<a href="/case-studies.html">ALL CASE STUDIES</a>
+ <a href="/case-studies/xerj-self-audit.html">RUST SELF-AUDIT</a>
<a href="/case-studies/wordpress-security-audit.html" class="active">WORDPRESS SECURITY AUDIT</a>
<a href="/case-studies/calltree-analytics.html">SEMANTIC ANALYTICS</a>
<a href="/case-studies/daily-dev-postgres-cdc.html">POSTGRES CDC + HYBRID</a>
diff --git a/landing/case-studies/xerj-self-audit.html b/landing/case-studies/xerj-self-audit.html
new file mode 100644
index 000000000..c34a2ad72
--- /dev/null
+++ b/landing/case-studies/xerj-self-audit.html
@@ -0,0 +1,234 @@
+<!DOCTYPE html>
+<html lang="en" data-theme="night">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>XERJ.ai — XERJ audits its own Rust</title>
+<meta name="description" content="XERJ indexed its own Rust engine as an AST/call-graph substrate and found a real unauthenticated stack-overflow DoS — reached for 10,533 tokens instead of the 1,805,277 it costs to read the source once. A call-graph query grep cannot express; the fix is shipped.">
+<meta property="og:title" content="XERJ.ai — XERJ audits its own Rust">
+<meta property="og:description" content="197 files / ~201k lines indexed with 100% function coverage. A recursion-cycle query found an unauth /_sql process-abort DoS grep cannot express — 171× fewer tokens than read-all, proven by crashing a live server, fixed and tested.">
+<link rel="preconnect" href="https://fonts.googleapis.com">
+<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
+<link href="https://fonts.googleapis.com/css2?family=Big+Shoulders+Display:wght@400;700;800;900&family=Inter:wght@400;500;600;700&family=IBM+Plex+Sans:wght@400;500;600;700&family=JetBrains+Mono:wght@400;500;700&display=swap" rel="stylesheet">
+<link rel="icon" href="data:image/svg+xml,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 100'><rect width='100' height='100' fill='%230b0b0d'/><text x='50' y='68' font-size='60' font-family='Big Shoulders Display,sans-serif' font-weight='900' fill='%23ffc400' text-anchor='middle'>X</text></svg>">
+<link rel="stylesheet" href="/style.css?v=cs0728">
+<style>
+ /* page-local: comparison series wears the validated --z-cmp token, never gray */
+ .cmp-b .cmp-bar { background: var(--z-cmp); opacity: 1; }
+ .cmp-bars-head .cmp-b { color: var(--z-cmp); font-weight: 700; }
+ @media (max-width: 720px) {
+ .cmp-num { white-space: normal; overflow-wrap: break-word; text-align: left; flex-shrink: 1; min-width: 0; }
+ .cmp-label { white-space: normal; }
+ }
+</style>
+</head>
+<body>
+
+<!-- nav -->
+<nav class="nav" aria-label="Primary">
+ <a href="/index.html" class="brand">XERJ.AI</a>
+ <a href="/index.html">HOME</a>
+ <a href="/product.html">PRODUCT</a>
+ <span class="nav-drop">
+ <a href="/use-cases.html">USE CASES</a>
+ </span>
+ <a href="/case-studies.html" class="active">CASE STUDIES</a>
+ <span class="spacer"></span>
+ <a href="/for-agents.html">FOR AGENTS</a>
+ <a href="/docs/index.html">DOCS</a>
+ <a href="/docs/recipes/index.html">RECIPES</a>
+ <a href="/playground.html">DASHBOARDS</a>
+ <a href="https://github.com/xerj-org/xerj" rel="external">GITHUB</a>
+ <span class="theme">
+ <button type="button" data-theme="day">DAY</button>
+ <span class="dash">·</span>
+ <button type="button" data-theme="night" class="active">NIGHT</button>
+ </span>
+</nav>
+
+<nav class="uc-subnav" aria-label="Case studies">
+ <a href="/case-studies.html">ALL CASE STUDIES</a>
+ <a href="/case-studies/xerj-self-audit.html" class="active">RUST SELF-AUDIT</a>
+ <a href="/case-studies/wordpress-security-audit.html">WORDPRESS SECURITY AUDIT</a>
+ <a href="/case-studies/calltree-analytics.html">SEMANTIC ANALYTICS</a>
+ <a href="/case-studies/daily-dev-postgres-cdc.html">POSTGRES CDC + HYBRID</a>
+</nav>
+
+<div class="uc-hero">
+ <div class="uc-eyebrow">CASE STUDY · RUST SELF-AUDIT</div>
+ <h1>XERJ AUDITED ITS OWN RUST.<br><span class="accent">IT FOUND A REAL 0-DAY.</span></h1>
+ <p class="uc-lead">XERJ indexed its own engine — 197 Rust files, ~201k lines — as a queryable AST and call graph, then hunted for security bugs on top of it. A call-graph query that grep cannot express surfaced an <strong>unauthenticated <code class="inline">POST /_sql</code> stack-overflow that aborts the whole process</strong>. Reaching it cost <strong>10,533 tokens</strong> instead of the <strong>1,805,277</strong> it takes to read the source once. The bug is fixed and the fix is proven by crashing — then not crashing — a live server.</p>
+ <div class="kicker"><span>FOR</span><span class="dash">·</span><span>SECURITY ENGINEERS</span><span class="dash">·</span><span>AI-AGENT BUILDERS</span><span class="dash">·</span><span>RUST TEAMS</span></div>
+</div>
+
+<!-- Hero chart -->
+<section class="scene-section" style="border-top:0; padding-top:0;">
+ <div class="viz-block">
+ <div class="kicker"><span>MEASURED · TOKENS TO REACH THE /_sql FINDING · tiktoken cl100k_base</span></div>
+ <div class="cmp-bars">
+ <div class="cmp-bars-head">
+ <span></span>
+ <span class="cmp-a">XERJ-ASSISTED</span>
+ <span class="cmp-b">READ EVERY FILE</span>
+ </div>
+ <div class="cmp-bars-row">
+ <span class="cmp-label">Tokens to the finding</span>
+ <span class="cmp-val cmp-a">
+ <span class="cmp-bar" style="width:0.5%;"></span>
+ <span class="cmp-num">10,533</span>
+ </span>
+ <span class="cmp-val cmp-b">
+ <span class="cmp-bar" style="width:90.0%;"></span>
+ <span class="cmp-num">1,805,277 <span class="cmp-ratio">171×</span></span>
+ </span>
+ </div>
+ </div>
+ <p style="margin-top:14px; font-size:12px; letter-spacing:.04em; color:var(--z-mute); font-family:var(--font-data);">READ-ALL = EVERY .rs FILE UNDER engine/ TOKENIZED (tiktoken cl100k_base, an LLM-token proxy) · XERJ-ASSISTED = cycle-finder output + the 4 parser functions actually read + the route lookup · NOT A LATENCY BENCHMARK · SOURCES IN <a href="https://github.com/xerj-org/xerj/tree/main/docs/case-studies/xerj-self-audit" rel="external">docs/case-studies/xerj-self-audit &rarr;</a></p>
+ </div>
+</section>
+
+<div class="pain-solve">
+ <div class="pain">
+ <h3>THE CONTEXT-WINDOW WALL</h3>
+ <ul><li><strong>The code does not fit.</strong> ~201k lines is ~9× a 200k-token window — "read everything to audit it" is not an option.</li><li><strong>Chunking kills the reach.</strong> A stack-overflow in a parser is a recursion <em>cycle</em> across several functions; split them across chunks and the cycle vanishes.</li><li><strong>grep cannot say the shape.</strong> "Mutually recursive functions with no depth guard anywhere in the cycle" is not a literal string.</li></ul>
+ </div>
+ <div class="solve">
+ <h3>THE XERJ ANSWER</h3>
+ <ul><li><strong>Index once as facts.</strong> 5,095 functions, 87,137 call edges, 477 routes, a full unsafe inventory — extracted in ~3.2 s, ingested in ~1.9 s.</li><li><strong>Query the bug shape.</strong> Strongly-connected components of the call graph, filtered to cycles with no depth guard.</li><li><strong>Prove it, then fix it.</strong> Trace the cycle to its route, crash a live server, ship a bounded-depth fix, watch the same payload return HTTP 400.</li></ul>
+ </div>
+</div>
+
+<div class="proof-row">
+ <div class="stat">
+ <div class="num">100%</div>
+ <div class="label">FUNCTION COVERAGE</div>
+ <div class="ctx">5,095 / 5,095 functions · 197 files · 0 tree-sitter parse errors</div>
+ </div>
+ <div class="stat">
+ <div class="num">1</div>
+ <div class="label">UNAUTH CRITICAL, FIXED</div>
+ <div class="ctx">/_sql stack-overflow → process abort · proven live · depth-guard shipped</div>
+ </div>
+ <div class="stat">
+ <div class="num">22 / 0</div>
+ <div class="label">UNSAFE BLOCKS · UNSOUND</div>
+ <div class="ctx">every unsafe site enumerated and reviewed; all invariants hold</div>
+ </div>
+</div>
+
+<!-- Ground-truth recall -->
+<section class="scene-section">
+ <h2 class="scene" style="font-size:var(--fs-56);">DOES IT ACTUALLY <span class="accent">FIND BUGS?</span></h2>
+ <p class="wide">Before trusting the substrate on unknown bugs, we tested it on known ones. The day before, a PR fixed six real vulnerabilities in this exact engine. We indexed the <em>pre-fix</em> commit and re-ran the queries on both — the signal has to flip when the bug is fixed, or it is noise.</p>
+ <div class="laws">
+ <div class="n">F1</div><div class="law"><div class="t">QUERY-PARSER STACK OVERFLOW</div><div class="b">The recursion-cycle query flags the query_string parser cycle as <strong>UNGUARDED pre-fix</strong> and the same cycle as <strong>GUARDED</strong> on current main. The signal flips exactly on the fix.</div></div>
+ <div class="n">F7</div><div class="law"><div class="t">FIELD-LIMIT BYPASS ON INGEST</div><div class="b"><code class="inline">evolve_schema_*</code> shows <code class="inline">reads_config_limit: false</code> pre-fix and <code class="inline">true</code> after. The substrate distinguishes the vulnerable version from the patched one.</div></div>
+ </div>
+</section>
+
+<section class="scene-section">
+ <h2 class="scene" style="font-size:var(--fs-56);">THE QUERY GREP <span class="accent">CANNOT WRITE.</span></h2>
+ <p class="wide">A stack overflow in a recursive-descent parser is rarely direct self-recursion — it is a cycle: <code class="inline">parse_or_expr → parse_and_expr → parse_condition → parse_or_expr</code>, one turn per <code class="inline">(</code>. No single file, function, or regex shows it. The free-function call graph does.</p>
+
+ <div class="kicker" style="margin-top:var(--sp-6);"><span>INDEX ONCE · ~5 S · 100% FUNCTION COVERAGE</span></div>
+ <pre class="code">$ python3 rust_ast_index.py ./engine --out ./ast-out <span class="comment"># 5,095 fns, 87,137 edges, 477 routes</span>
+$ python3 ingest.py ./ast-out --url $XERJ <span class="comment"># sent == indexed, asserted</span></pre>
+
+ <div class="kicker" style="margin-top:var(--sp-6);"><span>ONE GRAPH QUERY · 186 UNGUARDED CYCLES</span></div>
+ <pre class="code">$ python3 find_recursion_cycles.py ./ast-out
+<span class="comment"># strongly-connected components of the call graph,</span>
+<span class="comment"># flagged where NO member establishes a depth bound.</span>
+<span class="comment"># ranks the /_sql WHERE-clause parser cycle as UNGUARDED.</span></pre>
+
+ <div class="kicker" style="margin-top:var(--sp-6);"><span>TRACE TO THE ROUTE · ONE REQUEST</span></div>
+ <pre class="code">$ curl -s "$XERJ/rustroutes/_search" -H 'Content-Type: application/json' \
+ -d '{"query":{"wildcard":{"path":"*sql*"}},"_source":["method","path","handler"]}'
+<span class="comment"># POST /_sql -> sql_query (crates/xerj-api/src/router.rs:637) — unauthenticated</span></pre>
+</section>
+
+<section class="scene-section">
+ <h2 class="scene" style="font-size:var(--fs-56);">PROVEN BY <span class="accent">CRASHING IT.</span></h2>
+ <p class="wide">A finding is a claim until you trigger it. One <code class="inline">POST /_sql</code> with ~50k nested parens in the WHERE clause aborts the entire server — every tenant, every index — with a stack overflow that no <code class="inline">catch_unwind</code> can save.</p>
+ <pre class="code">$ curl -s $XERJ/_sql -d '{"query":"SELECT * FROM t WHERE ((((...50k...))))"}'
+<span class="comment"># thread 'xerj-rt' has overflowed its stack</span>
+<span class="comment"># fatal runtime error: stack overflow, aborting</span>
+<span class="comment"># ... Aborted (core dumped) — exit 134 (SIGABRT)</span></pre>
+ <p class="wide" style="margin-top:var(--sp-5);">The fix mirrors the query_string parser: a bounded recursion depth. After it, the identical payload returns a clean error in ~2 ms and the server stays up.</p>
+ <pre class="code">$ curl -s $XERJ/_sql -d '{"query":"SELECT * FROM t WHERE ((((...50k...))))"}'
+<span class="comment"># {"error":{"reason":"SQL parse error: WHERE clause nesting exceeds max depth of 64"},"status":400}</span>
+<span class="comment"># server still answering 200 afterwards.</span></pre>
+</section>
+
+<section class="scene-section">
+ <h2 class="scene" style="font-size:var(--fs-56);">WHAT THIS <span class="accent">DOES NOT SHOW.</span></h2>
+ <div class="laws">
+ <div class="n">01</div><div class="law"><div class="t">XERJ DOES NOT <span class="accent">FIND BUGS.</span></div><div class="b">The tree-sitter extractor and the call-graph analysis produce the candidates; XERJ stores and queries them at scale. The substrate narrows the search; a human or agent still reads the survivors.</div></div>
+ <div class="n">02</div><div class="law"><div class="t">THE GRAPH IS <span class="accent">NAME-BASED.</span></div><div class="b">Method-call edges are excluded on purpose — resolving <code class="inline">.len()</code> by name collapses the graph into one false component. Cycles through trait dispatch or macros are out of reach; the recursion result covers free functions.</div></div>
+ <div class="n">03</div><div class="law"><div class="t">EVERY CANDIDATE STILL <span class="accent">NEEDS A READ.</span></div><div class="b">Blunt category filters over-return (87 for one alloc query); argument-provenance signals sharpen them, and every published finding was read and, where possible, executed. Two unproven traversal leads were <em>not</em> published.</div></div>
+ <div class="n">04</div><div class="law"><div class="t">ONE PASS, HIGHEST-VALUE <span class="accent">LENSES.</span></div><div class="b">This publishes what was proven or fully enumerated — the DoS and the unsafe inventory. The broader auth-flow and business-logic lenses are scoped in the walkthrough, not exhausted here.</div></div>
+ </div>
+</section>
+
+<section class="scene-section">
+ <h2 class="scene" style="font-size:var(--fs-56);">READ THE <span class="accent">WALKTHROUGH.</span></h2>
+ <p class="wide">The extractor, the ingest script, the recursion-cycle finder, the query cookbook with live hit counts, the ground-truth recall test, and the coverage accounting are all published. Every number reproduces from scratch.</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/xerj-self-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 WALKTHROUGH</span>
+ <span class="arrow">→</span>
+ </a>
+ <a href="https://github.com/xerj-org/xerj/tree/main/docs/examples/rust-ast-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>GET THE TOOLING</span>
+ <span class="arrow">→</span>
+ </a>
+ </div>
+</section>
+
+<section class="lead-capture" id="demo">
+ <div class="kicker"><span>READY?</span><span class="dash">·</span><span>REQUEST ACCESS</span></div>
+ <h3>AUDIT YOUR<br><span class="accent">OWN RUST.</span></h3>
+ <p>The same extractor + queries index your crate into XERJ and hunt the same bug shapes — recursion cycles, unbounded allocation, unsafe soundness, path traversal. Want a walkthrough with your team? Leave an email.</p>
+ <form class="lead-form" id="leadForm" action="#" method="post" novalidate>
+ <div class="lead-form-input">
+ <input type="email" name="email" placeholder="[email protected]" required aria-label="Work email" autocomplete="email" />
+ </div>
+ <button type="submit" class="demo">SEE THE DEMO <span class="arrow">→</span></button>
+ <span class="note">We only use this email to send you the binary. Ever.</span>
+ <span class="ok">✓ THANKS. CHECK YOUR INBOX WITHIN 24 HOURS.</span>
+ </form>
+</section>
+
+<footer class="footer">
+ <span>XERJ.AI · CASE STUDIES · V0.1</span>
+ <span>STORAGE FOR THE AGENT ERA</span>
+ <span><a href="/index.html">HOME</a> · <a href="/case-studies.html">ALL CASE STUDIES</a> · <a href="/docs/index.html">DOCS</a> · <a href="/playground.html">DASHBOARDS</a> &middot; <a href="/llms.txt">LLMS.TXT</a></span>
+</footer>
+
+<script>
+ (function () {
+ const root = document.documentElement;
+ for (const btn of document.querySelectorAll('[data-theme]')) {
+ btn.addEventListener('click', () => {
+ const t = btn.getAttribute('data-theme');
+ root.setAttribute('data-theme', t);
+ for (const b of document.querySelectorAll('[data-theme]')) {
+ b.classList.toggle('active', b.getAttribute('data-theme') === t);
+ }
+ });
+ }
+ const form = document.getElementById('leadForm');
+ if (form) {
+ form.addEventListener('submit', (e) => {
+ e.preventDefault();
+ const email = form.querySelector('input[type=email]').value.trim();
+ if (!email || !email.includes('@')) return;
+ console.log('[xerj lead]', email);
+ try { fetch('/lead', { method: 'POST', headers: { 'content-type': 'application/json' }, body: JSON.stringify({ email: email, source: 'xerj-self-audit' }), keepalive: true }).catch(function () {}); } catch (_) {}
+ form.classList.add('sent');
+ });
+ }
+ })();
+</script>
+
+</body>
+</html>