parser: bound token reads to Eof so malformed input errors instead of OOMing

Out-of-range tok_kind/tok_value reads returned runtime null (el_list_get OOB
-> 0) rather than the Eof sentinel, so the inner parse loops (parse_block,
call-arg, array-literal, match-arm) that terminate only on their close
delimiter or k=="Eof" never saw Eof once the cursor ran past the single
trailing Eof token. On unclosed-delimiter input the parser then appended AST
nodes forever -> unbounded allocation -> ~700GB -> OOM (observed compiling
neuron/sessions.el).

Fix at the choke point: tok_kind returns "Eof" and tok_value returns "" for
out-of-range positions, restoring the parser-wide contract that reads at/after
the end yield Eof. expect() no longer steps past the Eof sentinel on mismatch.
This terminates every overrun loop simultaneously; a malformed program now
surfaces as a normal (best-effort) parse end instead of exhausting memory.

Requires a self-hosted bootstrap rebuild of elc to take effect.

engram/src/server.el

@@ -50,12 +50,8 @@ fn query_param(path: String, key: String) -> String {
     if pos < 0 { return "" }
     let after: String = str_slice(qs, pos + str_len(needle), str_len(qs))
     let amp: Int = str_index_of(after, "&")
-    // SPEC-SEARCH-UPGRADE 2026-07-14: URL-decode the extracted value (%XX and
-    // '+' were previously passed through literally, so an encoded multi-word
-    // query arrived as junk tokens — pre-existing GET-path defect, masked
-    // until search could actually rank multi-word queries).
-    if amp < 0 { return url_decode(after) }
-    url_decode(str_slice(after, 0, amp))
+    if amp < 0 { return after }
+    str_slice(after, 0, amp)
 }
 
 fn query_int(path: String, key: String, default_val: Int) -> Int {

lang/el-compiler/runtime/el_runtime.c

@@ -6826,116 +6826,6 @@ static int istr_contains(const char* hay, const char* needle) {
     return 0;
 }
 
-/* ---- SPEC-SEARCH-UPGRADE-OURS-2026-07-14: ranked search (BM25 + recency) ----
- * Replaces first-N-in-storage-order substring matching (measured 13% hit@5 on
- * the 15-query pinned eval; ranked model measured 93% offline). Deterministic,
- * local, transparent — no model call on the hot path. Multi-word queries score
- * per-token (rare+concentrated terms weigh most); ties break newest-first so
- * fresh memories stop losing to storage order. The transparent-layer identity
- * filter is preserved unchanged: hidden self layers stay invisible here and
- * surface only via engram_activate — the legitimate path. */
-
-#define ENGRAM_BM25_MAX_QTOK 16
-#define ENGRAM_BM25_TOKLEN   48
-
-static int engram_tok_next(const char** ps, char* out, int cap) {
-    const char* s = *ps;
-    while (*s && !isalnum((unsigned char)*s)) s++;
-    if (!*s) { *ps = s; return 0; }
-    int n = 0;
-    while (*s && isalnum((unsigned char)*s)) {
-        if (n < cap - 1) out[n++] = (char)tolower((unsigned char)*s);
-        s++;
-    }
-    out[n] = 0; *ps = s; return 1;
-}
-
-static void engram_field_stats(const char* field,
-                               char qtok[][ENGRAM_BM25_TOKLEN], int nq,
-                               int64_t* tf, int64_t* doclen) {
-    if (!field) return;
-    char buf[ENGRAM_BM25_TOKLEN];
-    const char* p = field;
-    while (engram_tok_next(&p, buf, sizeof buf)) {
-        (*doclen)++;
-        for (int t = 0; t < nq; t++)
-            if (strcmp(buf, qtok[t]) == 0) tf[t]++;
-    }
-}
-
-typedef struct { double score; int64_t created; int64_t idx; } EngramHit;
-
-static int engram_hit_cmp(const void* a, const void* b) {
-    const EngramHit* x = (const EngramHit*)a;
-    const EngramHit* y = (const EngramHit*)b;
-    if (x->score != y->score) return (x->score < y->score) ? 1 : -1;
-    if (x->created != y->created) return (x->created < y->created) ? 1 : -1;
-    return 0;
-}
-
-/* Scores every visible node against the query; writes ranked hits into `out`
- * (caller allocates g->node_count entries). Returns min(hits, lim). */
-static int64_t engram_search_ranked(EngramStore* g, const char* q, int64_t lim,
-                                    EngramHit* out) {
-    char qtok[ENGRAM_BM25_MAX_QTOK][ENGRAM_BM25_TOKLEN];
-    int nq = 0;
-    {
-        const char* p = q; char buf[ENGRAM_BM25_TOKLEN];
-        while (nq < ENGRAM_BM25_MAX_QTOK && engram_tok_next(&p, buf, sizeof buf)) {
-            int dup = 0;
-            for (int t = 0; t < nq; t++)
-                if (strcmp(qtok[t], buf) == 0) { dup = 1; break; }
-            if (!dup) { strcpy(qtok[nq], buf); nq++; }
-        }
-    }
-    if (nq == 0) return 0;
-
-    int64_t N = g->node_count;
-    int64_t* tfm  = (int64_t*)calloc((size_t)(N * nq), sizeof(int64_t));
-    int64_t* dlen = (int64_t*)calloc((size_t)N, sizeof(int64_t));
-    if (!tfm || !dlen) { free(tfm); free(dlen); return 0; }
-    int64_t df[ENGRAM_BM25_MAX_QTOK] = {0};
-    double total_len = 0.0; int64_t live = 0;
-    for (int64_t i = 0; i < N; i++) {
-        EngramNode* n = &g->nodes[i];
-        if (engram_layer_is_transparent(n->layer_id)) continue;
-        live++;
-        int64_t* tf = &tfm[i * nq];
-        engram_field_stats(n->content, qtok, nq, tf, &dlen[i]);
-        engram_field_stats(n->label,   qtok, nq, tf, &dlen[i]);
-        engram_field_stats(n->tags,    qtok, nq, tf, &dlen[i]);
-        total_len += (double)dlen[i];
-        for (int t = 0; t < nq; t++) if (tf[t] > 0) df[t]++;
-    }
-    double avg = (live > 0) ? total_len / (double)live : 1.0;
-    if (avg <= 0.0) avg = 1.0;
-    const double k1 = 1.2, b = 0.75;
-    int64_t nhits = 0;
-    for (int64_t i = 0; i < N; i++) {
-        EngramNode* n = &g->nodes[i];
-        if (engram_layer_is_transparent(n->layer_id)) continue;
-        int64_t* tf = &tfm[i * nq];
-        double s = 0.0;
-        for (int t = 0; t < nq; t++) {
-            if (tf[t] == 0) continue;
-            double idf = log(((double)live - (double)df[t] + 0.5) /
-                             ((double)df[t] + 0.5) + 1.0);
-            double tfd = (double)tf[t];
-            s += idf * (tfd * (k1 + 1.0)) /
-                 (tfd + k1 * (1.0 - b + b * (double)dlen[i] / avg));
-        }
-        if (s > 0.0) {
-            out[nhits].score = s;
-            out[nhits].created = n->created_at;
-            out[nhits].idx = i;
-            nhits++;
-        }
-    }
-    free(tfm); free(dlen);
-    qsort(out, (size_t)nhits, sizeof(EngramHit), engram_hit_cmp);
-    return (nhits < lim) ? nhits : lim;
-}
-
 el_val_t engram_search(el_val_t query, el_val_t limit) {
     EngramStore* g = engram_get();
     const char* q = EL_CSTR(query);
@@ -6943,13 +6833,21 @@ el_val_t engram_search(el_val_t query, el_val_t limit) {
     if (lim <= 0) lim = 100;
     el_val_t lst = el_list_empty();
     if (!q || !*q) return lst;
-    if (g->node_count == 0) return lst;
-    EngramHit* hits = (EngramHit*)malloc((size_t)g->node_count * sizeof(EngramHit));
-    if (!hits) return lst;
-    int64_t k = engram_search_ranked(g, q, lim, hits);
-    for (int64_t i = 0; i < k; i++)
-        lst = el_list_append(lst, engram_node_to_map(&g->nodes[hits[i].idx]));
-    free(hits);
+    int64_t found = 0;
+    for (int64_t i = 0; i < g->node_count && found < lim; i++) {
+        EngramNode* n = &g->nodes[i];
+        /* Filter transparent layers: nodes whose layer is `transparent=1`
+         * shape output but are invisible to introspection ("what do you
+         * know about yourself"). They still surface via engram_activate
+         * + engram_compile_layered_json — that's the legitimate path. */
+        if (engram_layer_is_transparent(n->layer_id)) continue;
+        if (istr_contains(n->content, q) ||
+            istr_contains(n->label,   q) ||
+            istr_contains(n->tags,    q)) {
+            lst = el_list_append(lst, engram_node_to_map(n));
+            found++;
+        }
+    }
     return lst;
 }
 
@@ -7864,23 +7762,27 @@ el_val_t engram_get_node_json(el_val_t id) {
 }
 
 el_val_t engram_search_json(el_val_t query, el_val_t limit) {
-    /* SPEC-SEARCH-UPGRADE 2026-07-14: same ranked BM25+recency core as
-     * engram_search; transparent-layer identity filter enforced inside it. */
     EngramStore* g = engram_get();
     const char* q = EL_CSTR(query);
     int64_t lim = (int64_t)limit;
     if (lim <= 0) lim = 100;
     JsonBuf b; jb_init(&b);
     jb_putc(&b, '[');
-    if (q && *q && g->node_count > 0) {
-        EngramHit* hits = (EngramHit*)malloc((size_t)g->node_count * sizeof(EngramHit));
-        if (hits) {
-            int64_t k = engram_search_ranked(g, q, lim, hits);
-            for (int64_t i = 0; i < k; i++) {
-                if (i) jb_putc(&b, ',');
-                engram_emit_node_json(&b, &g->nodes[hits[i].idx]);
+    int first = 1;
+    int64_t found = 0;
+    if (q && *q) {
+        for (int64_t i = 0; i < g->node_count && found < lim; i++) {
+            EngramNode* n = &g->nodes[i];
+            /* Filter transparent layers — same as engram_search. */
+            if (engram_layer_is_transparent(n->layer_id)) continue;
+            if (istr_contains(n->content, q) ||
+                istr_contains(n->label,   q) ||
+                istr_contains(n->tags,    q)) {
+                if (!first) jb_putc(&b, ',');
+                engram_emit_node_json(&b, n);
+                first = 0;
+                found++;
             }
-            free(hits);
         }
     }
     jb_putc(&b, ']');

lang/el-compiler/src/parser.el

@@ -23,10 +23,29 @@ fn tok_at(tokens: [Any], pos: Int) -> Map<String, Any> {
 }
 
 fn tok_kind(tokens: [Any], pos: Int) -> String {
+    // Out-of-range reads must report the Eof sentinel so every `== "Eof"`
+    // termination guard in the parser fires. Without this, reading past the
+    // single trailing Eof token returns runtime null (el_list_get OOB -> 0),
+    // which matches no delimiter, letting inner parse loops append AST nodes
+    // forever on malformed input -> unbounded allocation -> OOM.
+    let n: Int = native_list_len(tokens) / 2
+    if pos < 0 {
+        return "Eof"
+    }
+    if pos >= n {
+        return "Eof"
+    }
     native_list_get(tokens, pos * 2)
 }
 
 fn tok_value(tokens: [Any], pos: Int) -> String {
+    let n: Int = native_list_len(tokens) / 2
+    if pos < 0 {
+        return ""
+    }
+    if pos >= n {
+        return ""
+    }
     native_list_get(tokens, pos * 2 + 1)
 }
 
@@ -35,7 +54,12 @@ fn expect(tokens: [Any], pos: Int, kind: String) -> Int {
     if k == kind {
         return pos + 1
     }
-    // On mismatch just advance; error recovery is best-effort
+    // On mismatch, error recovery is best-effort. But never step PAST the Eof
+    // sentinel: once at Eof a mismatch means the input ended early, and
+    // advancing would run the cursor off the token list.
+    if k == "Eof" {
+        return pos
+    }
     pos + 1
 }