feat(extraction): add Erlang language support (.erl/.hrl) (#635, #648) (#1165)

Vendored WhatsApp/tree-sitter-erlang 0.19 (the ELP grammar, ABI 14) with an
Erlang-shaped extractor: multi-clause/multi-arity functions merged into one
symbol, -spec signatures, records with fields, -type/-opaque aliases, -define
macros, -include/-include_lib file edges, and -export-driven visibility.

Modules wrap in a namespace so remote mod:fn(...) calls resolve through the
existing qualified-name matcher as mod::fn with zero resolver changes.
-behaviour declarations link to the behaviour module — gated to namespace
targets only (bare-name fallthrough linked -behaviour(supervisor) to an
unrelated macro constant on emqx). OTP indirection with static targets is
followed: spawn/apply/proc_lib/timer/rpc MFA-argument callees, and
gen_server:call/cast(?MODULE | ?SERVER) to the module's own
handle_call/handle_cast. Var-module dispatch and message sends stay
deliberately unlinked. codegraph_explore also normalizes Erlang-native query
spelling (mod:fn/3, init/2) so named symbols resolve as typed.

Benchmarked on cowboy (189 files), ejabberd (414), emqx (2,447): extraction
PASS on all three; with-codegraph arms reached 2/2/0 file Reads vs 10/5+/19
without, fastest on the largest repo.

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

src/mcp/tools.ts

@@ -95,6 +95,36 @@ function lastQualifierPart(symbol: string): string {
   return parts[parts.length - 1] ?? symbol;
 }
 
+/**
+ * Normalize Erlang-native symbol spellings in an explore query into the shapes
+ * the rest of the pipeline already understands. Agents working Erlang code
+ * name symbols the way the language spells them — `mod:fn/3`, `init/2` — and
+ * those tokens previously died in both consumers: the flow-builder's token
+ * filter rejects `:` and `/arity` outright, and the search-side field parser
+ * eats `mod:fn` as an unknown `field:value`. Measured on cowboy: the agent
+ * named `cowboy_stream_h:request_process/3` in two queries, got no body back
+ * either time, and fell back to Read.
+ *
+ *   - `fn/3` → `fn` (arity tail after an identifier; a path segment like
+ *     `src/2fa` doesn't match because the tail must be all digits)
+ *   - `mod:fn` → `mod.fn` (exactly one colon between identifiers, so it rides
+ *     the existing Class.method qualified handling; `::`, URLs, drive letters,
+ *     and times don't match, and the query language's own field prefixes —
+ *     kind:/lang:/language:/path:/name: — are left alone)
+ *
+ * Safe cross-language: Lua's `t:m` spelling maps to the same `t.m` its
+ * qualified names use, and no other supported spelling contains a bare
+ * single-colon identifier pair.
+ */
+export function normalizeQuerySpelling(query: string): string {
+  return query
+    .replace(/\b([A-Za-z_][\w@]*)\/(\d{1,3})(?=$|[\s,()[\]/])/g, '$1')
+    .replace(
+      /(^|[\s,()[\]])(?!(?:kind|lang|language|path|name):)([a-z_][\w@]*):([A-Za-z_][\w@]*)(?=$|[\s,()[\]])/g,
+      '$1$2.$3'
+    );
+}
+
 /**
  * Calculate the recommended number of codegraph_explore calls based on project size.
  * Larger codebases need more exploration calls to cover their surface area,
@@ -1854,7 +1884,7 @@ export class ToolHandler {
       // names (Class.method / Class::method) — the agent's most precise input,
       // resolved exactly by findAllSymbols. (The old strip mangled Class.method
       // into Class, throwing the method away.)
-      const FILE_EXT = /\.(?:java|kt|kts|ts|tsx|js|jsx|mjs|cjs|cs|py|go|rb|php|swift|rs|cpp|cc|cxx|c|h|hpp|scala|lua|dart|vue|svelte|astro)$/i;
+      const FILE_EXT = /\.(?:java|kt|kts|ts|tsx|js|jsx|mjs|cjs|cs|py|go|rb|php|swift|rs|cpp|cc|cxx|c|h|hpp|scala|lua|dart|vue|svelte|astro|erl|hrl)$/i;
       const tokens = [...new Set(
         query.split(/[\s,()[\]]+/)
           .map((t) => t.replace(FILE_EXT, '').trim())
@@ -2457,8 +2487,11 @@ export class ToolHandler {
    * tax on small projects while earning its keep on large ones.
    */
   private async handleExplore(args: Record<string, unknown>): Promise<ToolResult> {
-    const query = this.validateString(args.query, 'query');
-    if (typeof query !== 'string') return query;
+    const rawQuery = this.validateString(args.query, 'query');
+    if (typeof rawQuery !== 'string') return rawQuery;
+    // One normalization point so the flow-builder, relevance search, and
+    // ranking all see the same canonical spelling (Erlang `mod:fn/arity`).
+    const query = normalizeQuerySpelling(rawQuery);
 
     const cg = this.getCodeGraph(args.projectPath as string | undefined);
     const projectRoot = cg.getProjectRoot();
@@ -2539,7 +2572,7 @@ export class ToolHandler {
     // overloads (the query also named the type) all earn it. (#1064)
     const tierSeedIds = new Set<string>();
     {
-      const FILE_EXT = /\.(?:java|kt|kts|ts|tsx|js|jsx|mjs|cjs|cs|py|go|rb|php|swift|rs|cpp|cc|cxx|c|h|hpp|scala|lua|dart|vue|svelte|astro)$/i;
+      const FILE_EXT = /\.(?:java|kt|kts|ts|tsx|js|jsx|mjs|cjs|cs|py|go|rb|php|swift|rs|cpp|cc|cxx|c|h|hpp|scala|lua|dart|vue|svelte|astro|erl|hrl)$/i;
       const CALLABLE = new Set(['method', 'function', 'component', 'constructor']);
       const isTestPath = (p: string) => /(^|\/)(tests?|specs?|__tests__|testdata|mocks?|fixtures?)\//i.test(p) || /\.(test|spec)\.[a-z]+$/i.test(p);
       const bodyLines = (n: Node) => Math.max(0, (n.endLine ?? n.startLine) - n.startLine);