fix(codegen): emit the declared cgi identity — it was searched for in a list that cannot contain it

A cgi block is a top-level declaration, so codegen_streaming classifies it via
is_top_level_decl and releases it. The identity emission then searched
toplevel_exec_stmts for that same block. Declarations are excluded from that list by
construction, so the search could never succeed. A probe printed what it actually
saw for a program whose first statement is a cgi block: [Let, Expr]. It emitted
nothing, silently, with no diagnostic on any channel.

The code documented its own assumption — 'Since cgi blocks are rare and small, they
end up in toplevel_exec_stmts' — and that assumption was false.

Capture the declared values before the release and emit from them. The search is
deleted rather than repaired, so the failure mode is removed rather than relocated.

Proven discriminating (old fails, new passes):
  minimal cgi program, old   -> 0 el_cgi_init
  minimal cgi program, fixed -> el_cgi_init with all four declared values
  neuron soul, fixed         -> principal present in the compiled binary (0 before),
                                boots in 2s, interface 110 routes in / 110 out

Consequence: a binary now carries its declared identity as a compiled constant,
which is what the identity protocol requires. Whether the runtime surfaces it to
state_get("soul_principal") is unverified and separate.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

engram/dist/engram.c

@@ -10,7 +10,6 @@ el_val_t query_param(el_val_t path, el_val_t key);
 el_val_t query_int(el_val_t path, el_val_t key, el_val_t default_val);
 el_val_t extract_id(el_val_t path, el_val_t prefix);
 el_val_t route_stats(el_val_t method, el_val_t path, el_val_t body);
-el_val_t persist_canonical(void);
 el_val_t route_create_node(el_val_t method, el_val_t path, el_val_t body);
 el_val_t route_get_node(el_val_t method, el_val_t path, el_val_t body);
 el_val_t route_scan_nodes(el_val_t method, el_val_t path, el_val_t body);
@@ -21,23 +20,18 @@ el_val_t route_create_edge(el_val_t method, el_val_t path, el_val_t body);
 el_val_t route_neighbors(el_val_t method, el_val_t path, el_val_t body);
 el_val_t route_strengthen(el_val_t method, el_val_t path, el_val_t body);
 el_val_t route_forget(el_val_t method, el_val_t path, el_val_t body);
+el_val_t route_create_ise(el_val_t method, el_val_t path, el_val_t body);
+el_val_t route_sync(el_val_t method, el_val_t path, el_val_t body);
 el_val_t route_save(el_val_t method, el_val_t path, el_val_t body);
 el_val_t route_load(el_val_t method, el_val_t path, el_val_t body);
 el_val_t route_health(el_val_t method, el_val_t path, el_val_t body);
-el_val_t route_sync(el_val_t method, el_val_t path, el_val_t body);
-el_val_t route_load_merge(el_val_t method, el_val_t path, el_val_t body);
-el_val_t route_emit_ise(el_val_t method, el_val_t path, el_val_t body);
-el_val_t route_capture_knowledge(el_val_t method, el_val_t path, el_val_t body);
 el_val_t check_auth_ok(el_val_t method, el_val_t body);
 el_val_t handle_request(el_val_t method, el_val_t path, el_val_t body);
 
-el_val_t bind_raw;
 el_val_t bind_str;
 el_val_t port;
-el_val_t data_dir_raw;
 el_val_t data_dir;
 el_val_t snapshot_path;
-el_val_t boot_snap;
 
 el_val_t parse_port(el_val_t bind) {
   el_val_t colon = str_index_of(bind, EL_STR(":"));
@@ -116,22 +110,17 @@ el_val_t route_stats(el_val_t method, el_val_t path, el_val_t body) {
   return 0;
 }
 
-el_val_t persist_canonical(void) {
-  el_val_t dir_raw = env(EL_STR("ENGRAM_DATA_DIR"));
-  el_val_t dir = ({ el_val_t _if_result_1 = 0; if (str_eq(dir_raw, EL_STR(""))) { _if_result_1 = (EL_STR("/tmp/engram")); } else { _if_result_1 = (dir_raw); } _if_result_1; });
-  engram_save(el_str_concat(dir, EL_STR("/snapshot.json")));
-  return 1;
-  return 0;
-}
-
 el_val_t route_create_node(el_val_t method, el_val_t path, el_val_t body) {
   el_val_t content = json_get_string(body, EL_STR("content"));
-  el_val_t nt_raw = json_get_string(body, EL_STR("node_type"));
-  el_val_t node_type = ({ el_val_t _if_result_2 = 0; if (str_eq(nt_raw, EL_STR(""))) { _if_result_2 = (EL_STR("Memory")); } else { _if_result_2 = (nt_raw); } _if_result_2; });
-  el_val_t sal_raw = json_get_float(body, EL_STR("salience"));
-  el_val_t salience = ({ el_val_t _if_result_3 = 0; if ((sal_raw == el_from_float(0.0))) { _if_result_3 = (el_from_float(0.5)); } else { _if_result_3 = (sal_raw); } _if_result_3; });
+  el_val_t node_type = json_get_string(body, EL_STR("node_type"));
+  if (str_eq(node_type, EL_STR(""))) {
+    node_type = EL_STR("Memory");
+  }
+  el_val_t salience = json_get_float(body, EL_STR("salience"));
+  if (salience == el_from_float(0.0)) {
+    salience = el_from_float(0.5);
+  }
   el_val_t id = engram_node(content, node_type, salience);
-  el_val_t saved = persist_canonical();
   return el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(EL_STR("{\"id\":\""), id), EL_STR("\",\"content\":\"")), content), EL_STR("\",\"node_type\":\"")), node_type), EL_STR("\"}"));
   return 0;
 }
@@ -157,9 +146,11 @@ el_val_t route_scan_nodes(el_val_t method, el_val_t path, el_val_t body) {
 }
 
 el_val_t route_scan_edges(el_val_t method, el_val_t path, el_val_t body) {
-  el_val_t dir_raw = env(EL_STR("ENGRAM_DATA_DIR"));
-  el_val_t dir = ({ el_val_t _if_result_4 = 0; if (str_eq(dir_raw, EL_STR(""))) { _if_result_4 = (EL_STR("/tmp/engram")); } else { _if_result_4 = (dir_raw); } _if_result_4; });
-  el_val_t snap_path = el_str_concat(dir, EL_STR("/.scan-export.json"));
+  el_val_t dir = env(EL_STR("ENGRAM_DATA_DIR"));
+  if (str_eq(dir, EL_STR(""))) {
+    dir = EL_STR("/tmp/engram");
+  }
+  el_val_t snap_path = el_str_concat(dir, EL_STR("/snapshot.json"));
   engram_save(snap_path);
   el_val_t snap = fs_read(snap_path);
   if (str_eq(snap, EL_STR(""))) {
@@ -174,22 +165,36 @@ el_val_t route_scan_edges(el_val_t method, el_val_t path, el_val_t body) {
 }
 
 el_val_t route_search(el_val_t method, el_val_t path, el_val_t body) {
-  el_val_t q = ({ el_val_t _if_result_5 = 0; if (str_eq(method, EL_STR("GET"))) { _if_result_5 = (query_param(path, EL_STR("q"))); } else { _if_result_5 = (json_get_string(body, EL_STR("query"))); } _if_result_5; });
-  el_val_t lim_url = query_int(path, EL_STR("limit"), 0);
-  el_val_t lim_body = json_get_int(body, EL_STR("limit"));
-  el_val_t lim_either = ({ el_val_t _if_result_6 = 0; if ((lim_url > 0)) { _if_result_6 = (lim_url); } else { _if_result_6 = (lim_body); } _if_result_6; });
-  el_val_t limit = ({ el_val_t _if_result_7 = 0; if ((lim_either > 0)) { _if_result_7 = (lim_either); } else { _if_result_7 = (20); } _if_result_7; });
+  el_val_t q = EL_STR("");
+  if (str_eq(method, EL_STR("GET"))) {
+    q = query_param(path, EL_STR("q"));
+  } else {
+    q = json_get_string(body, EL_STR("query"));
+  }
+  el_val_t limit = query_int(path, EL_STR("limit"), 20);
+  if (limit == 0) {
+    limit = json_get_int(body, EL_STR("limit"));
+  }
+  if (limit == 0) {
+    limit = 20;
+  }
   return engram_search_json(q, limit);
   return 0;
 }
 
 el_val_t route_activate(el_val_t method, el_val_t path, el_val_t body) {
-  el_val_t q = ({ el_val_t _if_result_8 = 0; if (str_eq(method, EL_STR("GET"))) { _if_result_8 = (query_param(path, EL_STR("q"))); } else { _if_result_8 = (json_get_string(body, EL_STR("query"))); } _if_result_8; });
-  if (str_eq(q, EL_STR(""))) {
-    return err_json(EL_STR("missing query"));
+  el_val_t q = EL_STR("");
+  el_val_t depth = 3;
+  if (str_eq(method, EL_STR("GET"))) {
+    q = query_param(path, EL_STR("q"));
+    depth = query_int(path, EL_STR("depth"), 3);
+  } else {
+    q = json_get_string(body, EL_STR("query"));
+    el_val_t bd = json_get_int(body, EL_STR("depth"));
+    if (bd > 0) {
+      depth = bd;
+    }
   }
-  el_val_t d_raw = ({ el_val_t _if_result_9 = 0; if (str_eq(method, EL_STR("GET"))) { _if_result_9 = (query_int(path, EL_STR("depth"), 3)); } else { _if_result_9 = (json_get_int(body, EL_STR("depth"))); } _if_result_9; });
-  el_val_t depth = ({ el_val_t _if_result_10 = 0; if ((d_raw > 0)) { _if_result_10 = (d_raw); } else { _if_result_10 = (3); } _if_result_10; });
   return el_str_concat(el_str_concat(EL_STR("{\"results\":"), engram_activate_json(q, depth)), EL_STR("}"));
   return 0;
 }
@@ -197,12 +202,15 @@ el_val_t route_activate(el_val_t method, el_val_t path, el_val_t body) {
 el_val_t route_create_edge(el_val_t method, el_val_t path, el_val_t body) {
   el_val_t from_id = json_get_string(body, EL_STR("from_id"));
   el_val_t to_id = json_get_string(body, EL_STR("to_id"));
-  el_val_t rel_raw = json_get_string(body, EL_STR("relation"));
-  el_val_t relation = ({ el_val_t _if_result_11 = 0; if (str_eq(rel_raw, EL_STR(""))) { _if_result_11 = (EL_STR("associates")); } else { _if_result_11 = (rel_raw); } _if_result_11; });
-  el_val_t w_raw = json_get_float(body, EL_STR("weight"));
-  el_val_t weight = ({ el_val_t _if_result_12 = 0; if ((w_raw == el_from_float(0.0))) { _if_result_12 = (el_from_float(0.5)); } else { _if_result_12 = (w_raw); } _if_result_12; });
+  el_val_t relation = json_get_string(body, EL_STR("relation"));
+  if (str_eq(relation, EL_STR(""))) {
+    relation = EL_STR("associates");
+  }
+  el_val_t weight = json_get_float(body, EL_STR("weight"));
+  if (weight == el_from_float(0.0)) {
+    weight = el_from_float(0.5);
+  }
   engram_connect(from_id, to_id, weight, relation);
-  el_val_t saved = persist_canonical();
   return el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(EL_STR("{\"ok\":true,\"from_id\":\""), from_id), EL_STR("\",\"to_id\":\"")), to_id), EL_STR("\",\"relation\":\"")), relation), EL_STR("\"}"));
   return 0;
 }
@@ -223,7 +231,6 @@ el_val_t route_strengthen(el_val_t method, el_val_t path, el_val_t body) {
     return err_json(EL_STR("missing node_id"));
   }
   engram_strengthen(id);
-  el_val_t saved = persist_canonical();
   return ok_json();
   return 0;
 }
@@ -234,40 +241,29 @@ el_val_t route_forget(el_val_t method, el_val_t path, el_val_t body) {
     return err_json(EL_STR("missing id"));
   }
   engram_forget(id);
-  el_val_t saved = persist_canonical();
   return ok_json();
   return 0;
 }
 
-el_val_t route_save(el_val_t method, el_val_t path, el_val_t body) {
-  el_val_t p_raw = json_get_string(body, EL_STR("path"));
-  el_val_t dir_raw = env(EL_STR("ENGRAM_DATA_DIR"));
-  el_val_t dir = ({ el_val_t _if_result_13 = 0; if (str_eq(dir_raw, EL_STR(""))) { _if_result_13 = (EL_STR("/tmp/engram")); } else { _if_result_13 = (dir_raw); } _if_result_13; });
-  el_val_t p = ({ el_val_t _if_result_14 = 0; if (str_eq(p_raw, EL_STR(""))) { _if_result_14 = (el_str_concat(dir, EL_STR("/snapshot.json"))); } else { _if_result_14 = (p_raw); } _if_result_14; });
-  engram_save(p);
-  return el_str_concat(el_str_concat(EL_STR("{\"ok\":true,\"path\":\""), p), EL_STR("\"}"));
-  return 0;
-}
-
-el_val_t route_load(el_val_t method, el_val_t path, el_val_t body) {
-  el_val_t p_raw = json_get_string(body, EL_STR("path"));
-  el_val_t dir_raw = env(EL_STR("ENGRAM_DATA_DIR"));
-  el_val_t dir = ({ el_val_t _if_result_15 = 0; if (str_eq(dir_raw, EL_STR(""))) { _if_result_15 = (EL_STR("/tmp/engram")); } else { _if_result_15 = (dir_raw); } _if_result_15; });
-  el_val_t p = ({ el_val_t _if_result_16 = 0; if (str_eq(p_raw, EL_STR(""))) { _if_result_16 = (el_str_concat(dir, EL_STR("/snapshot.json"))); } else { _if_result_16 = (p_raw); } _if_result_16; });
-  engram_load(p);
-  return ok_json();
-  return 0;
-}
-
-el_val_t route_health(el_val_t method, el_val_t path, el_val_t body) {
-  return EL_STR("{\"status\":\"ok\",\"engine\":\"engram-runtime-native\"}");
+el_val_t route_create_ise(el_val_t method, el_val_t path, el_val_t body) {
+  el_val_t content = json_get_string(body, EL_STR("content"));
+  if (str_eq(content, EL_STR(""))) {
+    return err_json(EL_STR("missing content"));
+  }
+  el_val_t sal = el_from_float(0.3);
+  el_val_t imp = el_from_float(0.3);
+  el_val_t conf = el_from_float(0.8);
+  el_val_t id = engram_node_full(content, EL_STR("InternalStateEvent"), EL_STR("state-event"), sal, imp, conf, EL_STR("Episodic"), EL_STR("[\"internal-state\",\"InternalStateEvent\"]"));
+  return el_str_concat(el_str_concat(EL_STR("{\"ok\":true,\"id\":\""), id), EL_STR("\"}"));
   return 0;
 }
 
 el_val_t route_sync(el_val_t method, el_val_t path, el_val_t body) {
-  el_val_t dir_raw = env(EL_STR("ENGRAM_DATA_DIR"));
-  el_val_t dir = ({ el_val_t _if_result_17 = 0; if (str_eq(dir_raw, EL_STR(""))) { _if_result_17 = (EL_STR("/tmp/engram")); } else { _if_result_17 = (dir_raw); } _if_result_17; });
-  el_val_t snap_path = el_str_concat(dir, EL_STR("/.sync-export.json"));
+  el_val_t dir = env(EL_STR("ENGRAM_DATA_DIR"));
+  if (str_eq(dir, EL_STR(""))) {
+    dir = EL_STR("/tmp/engram");
+  }
+  el_val_t snap_path = el_str_concat(dir, EL_STR("/sync-export.json"));
   engram_save(snap_path);
   el_val_t snap = fs_read(snap_path);
   if (str_eq(snap, EL_STR(""))) {
@@ -277,68 +273,36 @@ el_val_t route_sync(el_val_t method, el_val_t path, el_val_t body) {
   return 0;
 }
 
-el_val_t route_load_merge(el_val_t method, el_val_t path, el_val_t body) {
+el_val_t route_save(el_val_t method, el_val_t path, el_val_t body) {
   el_val_t p = json_get_string(body, EL_STR("path"));
   if (str_eq(p, EL_STR(""))) {
-    return err_json(EL_STR("path is required"));
+    el_val_t dir = env(EL_STR("ENGRAM_DATA_DIR"));
+    if (str_eq(dir, EL_STR(""))) {
+      dir = EL_STR("/tmp/engram");
+    }
+    p = el_str_concat(dir, EL_STR("/snapshot.json"));
   }
-  if (str_eq(fs_read(p), EL_STR(""))) {
-    return err_json(EL_STR("file missing or empty"));
-  }
-  el_val_t before_n = engram_node_count();
-  el_val_t before_e = engram_edge_count();
-  engram_load_merge(p);
-  el_val_t added_n = (engram_node_count() - before_n);
-  el_val_t added_e = (engram_edge_count() - before_e);
-  el_val_t saved = persist_canonical();
-  return el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(EL_STR("{\"ok\":true,\"nodes_added\":"), int_to_str(added_n)), EL_STR(",\"edges_added\":")), int_to_str(added_e)), EL_STR(",\"node_count\":")), int_to_str(engram_node_count())), EL_STR("}"));
+  engram_save(p);
+  return el_str_concat(el_str_concat(EL_STR("{\"ok\":true,\"path\":\""), p), EL_STR("\"}"));
   return 0;
 }
 
-el_val_t route_emit_ise(el_val_t method, el_val_t path, el_val_t body) {
-  el_val_t content = json_get_string(body, EL_STR("content"));
-  if (str_eq(content, EL_STR(""))) {
-    return err_json(EL_STR("missing content"));
+el_val_t route_load(el_val_t method, el_val_t path, el_val_t body) {
+  el_val_t p = json_get_string(body, EL_STR("path"));
+  if (str_eq(p, EL_STR(""))) {
+    el_val_t dir = env(EL_STR("ENGRAM_DATA_DIR"));
+    if (str_eq(dir, EL_STR(""))) {
+      dir = EL_STR("/tmp/engram");
+    }
+    p = el_str_concat(dir, EL_STR("/snapshot.json"));
   }
-  el_val_t sal = el_from_float(0.3);
-  el_val_t imp = el_from_float(0.3);
-  el_val_t conf = el_from_float(0.8);
-  el_val_t id = engram_node_full(content, EL_STR("InternalStateEvent"), EL_STR("state-event"), sal, imp, conf, EL_STR("Episodic"), EL_STR("[\"internal-state\",\"InternalStateEvent\"]"));
-  el_val_t ret_raw = env(EL_STR("ENGRAM_ISE_RETENTION_MS"));
-  el_val_t ret_ms = ({ el_val_t _if_result_18 = 0; if (str_eq(ret_raw, EL_STR(""))) { _if_result_18 = (172800000); } else { _if_result_18 = (str_to_int(ret_raw)); } _if_result_18; });
-  el_val_t pruned = engram_prune_telemetry(ret_ms);
-  return el_str_concat(el_str_concat(el_str_concat(el_str_concat(EL_STR("{\"ok\":true,\"id\":\""), id), EL_STR("\",\"pruned\":")), int_to_str(pruned)), EL_STR("}"));
+  engram_load(p);
+  return ok_json();
   return 0;
 }
 
-el_val_t route_capture_knowledge(el_val_t method, el_val_t path, el_val_t body) {
-  el_val_t content = json_get_string(body, EL_STR("content"));
-  if (str_eq(content, EL_STR(""))) {
-    return err_json(EL_STR("missing content"));
-  }
-  el_val_t title = json_get_string(body, EL_STR("title"));
-  el_val_t label = ({ el_val_t _if_result_19 = 0; if (str_eq(title, EL_STR(""))) { _if_result_19 = (str_slice(content, 0, 60)); } else { _if_result_19 = (title); } _if_result_19; });
-  el_val_t category_raw = json_get_string(body, EL_STR("category"));
-  el_val_t category = ({ el_val_t _if_result_20 = 0; if (str_eq(category_raw, EL_STR(""))) { _if_result_20 = (EL_STR("other")); } else { _if_result_20 = (category_raw); } _if_result_20; });
-  el_val_t ktier_raw = json_get_string(body, EL_STR("tier"));
-  el_val_t ktier = ({ el_val_t _if_result_21 = 0; if (str_eq(ktier_raw, EL_STR(""))) { _if_result_21 = (EL_STR("note")); } else { _if_result_21 = (ktier_raw); } _if_result_21; });
-  el_val_t project = json_get_string(body, EL_STR("project"));
-  el_val_t tags_raw = json_get_raw(body, EL_STR("tags"));
-  el_val_t tags_base = ({ el_val_t _if_result_22 = 0; if (str_eq(tags_raw, EL_STR(""))) { _if_result_22 = (EL_STR("[]")); } else { _if_result_22 = (tags_raw); } _if_result_22; });
-  el_val_t base_len = str_len(tags_base);
-  el_val_t head = str_slice(tags_base, 0, (base_len - 1));
-  el_val_t sep = ({ el_val_t _if_result_23 = 0; if (str_eq(head, EL_STR("["))) { _if_result_23 = (EL_STR("")); } else { _if_result_23 = (EL_STR(",")); } _if_result_23; });
-  el_val_t safe_cat = str_replace(category, EL_STR("\""), EL_STR("'"));
-  el_val_t safe_tier = str_replace(ktier, EL_STR("\""), EL_STR("'"));
-  el_val_t safe_proj = str_replace(project, EL_STR("\""), EL_STR("'"));
-  el_val_t proj_tag = ({ el_val_t _if_result_24 = 0; if (str_eq(safe_proj, EL_STR(""))) { _if_result_24 = (EL_STR("")); } else { _if_result_24 = (el_str_concat(el_str_concat(EL_STR(",\"project:"), safe_proj), EL_STR("\""))); } _if_result_24; });
-  el_val_t tags = el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(el_str_concat(head, sep), EL_STR("\"category:")), safe_cat), EL_STR("\",\"tier:")), safe_tier), EL_STR("\"")), proj_tag), EL_STR("]"));
-  el_val_t sal = el_from_float(0.5);
-  el_val_t imp = el_from_float(0.5);
-  el_val_t conf = el_from_float(0.9);
-  el_val_t id = engram_node_full(content, EL_STR("Knowledge"), label, sal, imp, conf, EL_STR("Semantic"), tags);
-  el_val_t saved = persist_canonical();
-  return el_str_concat(el_str_concat(EL_STR("{\"ok\":true,\"id\":\""), id), EL_STR("\"}"));
+el_val_t route_health(el_val_t method, el_val_t path, el_val_t body) {
+  return EL_STR("{\"status\":\"ok\",\"engine\":\"engram-runtime-native\"}");
   return 0;
 }
 
@@ -365,15 +329,12 @@ el_val_t handle_request(el_val_t method, el_val_t path, el_val_t body) {
       return route_health(method, path, body);
     }
   }
-  if (str_eq(method, EL_STR("POST")) && str_eq(clean, EL_STR("/api/neuron/state-events"))) {
-    return route_emit_ise(method, path, body);
+  if (str_eq(method, EL_STR("POST")) && str_starts_with(clean, EL_STR("/api/neuron/state-events"))) {
+    return route_create_ise(method, path, body);
   }
   if (!check_auth_ok(method, body)) {
     return err_json(EL_STR("unauthorized"));
   }
-  if (str_eq(method, EL_STR("POST")) && str_eq(clean, EL_STR("/api/neuron/knowledge/capture"))) {
-    return route_capture_knowledge(method, path, body);
-  }
   if (str_eq(method, EL_STR("GET")) && (str_eq(clean, EL_STR("/api/stats")) || str_eq(clean, EL_STR("/stats")))) {
     return route_stats(method, path, body);
   }
@@ -413,40 +374,32 @@ el_val_t handle_request(el_val_t method, el_val_t path, el_val_t body) {
   if (str_eq(method, EL_STR("POST")) && (str_eq(clean, EL_STR("/api/strengthen")) || str_eq(clean, EL_STR("/strengthen")))) {
     return route_strengthen(method, path, body);
   }
+  if (str_eq(method, EL_STR("GET")) && (str_eq(clean, EL_STR("/api/sync")) || str_eq(clean, EL_STR("/sync")))) {
+    return route_sync(method, path, body);
+  }
   if (str_eq(method, EL_STR("POST")) && (str_eq(clean, EL_STR("/api/save")) || str_eq(clean, EL_STR("/save")))) {
     return route_save(method, path, body);
   }
   if (str_eq(method, EL_STR("POST")) && (str_eq(clean, EL_STR("/api/load")) || str_eq(clean, EL_STR("/load")))) {
     return route_load(method, path, body);
   }
-  if (str_eq(method, EL_STR("POST")) && (str_eq(clean, EL_STR("/api/load-merge")) || str_eq(clean, EL_STR("/load-merge")))) {
-    return route_load_merge(method, path, body);
-  }
-  if (str_eq(method, EL_STR("GET")) && str_eq(clean, EL_STR("/api/sync"))) {
-    return route_sync(method, path, body);
-  }
   return el_str_concat(el_str_concat(EL_STR("{\"error\":\"not found\",\"path\":\""), clean), EL_STR("\"}"));
   return 0;
 }
 
 int main(int _argc, char** _argv) {
   el_runtime_init_args(_argc, _argv);
-  bind_raw = env(EL_STR("ENGRAM_BIND"));
-  bind_str = ({ el_val_t _if_result_25 = 0; if (str_eq(bind_raw, EL_STR(""))) { _if_result_25 = (EL_STR(":8742")); } else { _if_result_25 = (bind_raw); } _if_result_25; });
+  bind_str = env(EL_STR("ENGRAM_BIND"));
+  if (str_eq(bind_str, EL_STR(""))) {
+    bind_str = EL_STR(":8742");
+  }
   port = parse_port(bind_str);
-  data_dir_raw = env(EL_STR("ENGRAM_DATA_DIR"));
-  data_dir = ({ el_val_t _if_result_26 = 0; if (str_eq(data_dir_raw, EL_STR(""))) { _if_result_26 = (EL_STR("/tmp/engram")); } else { _if_result_26 = (data_dir_raw); } _if_result_26; });
+  data_dir = env(EL_STR("ENGRAM_DATA_DIR"));
+  if (str_eq(data_dir, EL_STR(""))) {
+    data_dir = EL_STR("/tmp/engram");
+  }
   snapshot_path = el_str_concat(data_dir, EL_STR("/snapshot.json"));
   engram_load(snapshot_path);
-  boot_snap = fs_read(snapshot_path);
-  if (!str_eq(boot_snap, EL_STR(""))) {
-    if (engram_node_count() == 0) {
-      println(EL_STR("[engram] WARNING: snapshot.json is non-empty but load produced 0 nodes \xe2\x80\x94 preserving copy at snapshot.failed-load.json"));
-      fs_write(el_str_concat(data_dir, EL_STR("/snapshot.failed-load.json")), boot_snap);
-    } else {
-      fs_write(el_str_concat(data_dir, EL_STR("/snapshot.boot-backup.json")), boot_snap);
-    }
-  }
   println(EL_STR("[engram] runtime-native graph engine"));
   println(el_str_concat(EL_STR("[engram] data_dir="), data_dir));
   println(el_str_concat(EL_STR("[engram] node_count="), int_to_str(engram_node_count())));

engram/src/server.el

@@ -50,8 +50,12 @@ 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, "&")
-    if amp < 0 { return after }
-    str_slice(after, 0, amp)
+    // 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))
 }
 
 fn query_int(path: String, key: String, default_val: Int) -> Int {
@@ -76,43 +80,13 @@ fn route_stats(method: String, path: String, body: String) -> String {
     engram_stats_json()
 }
 
-// (2026-07-18 self-review) Scoping sweep: `let` inside an if-block creates an
-// inner scope only — it does NOT mutate the outer binding (documented with
-// evidence in awareness.el, 2026-05-25). Every default/reassignment below used
-// that broken pattern, so defaults never applied: nodes were created with
-// node_type="" and salience=0.0, /api/search and /api/activate ALWAYS ran with
-// q="" regardless of input, edges defaulted to relation=""/weight=0.0, and
-// save/load with no "path" hit engram_save(""). Rewritten to the
-// `let x = if cond { a } else { b }` expression form (the pattern the newer
-// routes route_emit_ise/route_capture_knowledge already use correctly).
-// persist_canonical — save the canonical snapshot after a durable write.
-//
-// WHY (2026-07-22 self-review): the 2026-07-21 fix correctly stopped READ
-// routes from writing the canonical snapshot.json — but nothing was left
-// that saved it on WRITE. Every mutation (node create, edge create,
-// knowledge capture, forget, merge) lived only in RAM until someone POSTed
-// /api/save manually; a process restart silently discarded everything since
-// the last manual save. Observed live: two engram restarts during the
-// 2026-07-22 review reverted the store to a ~17h-old snapshot, destroying
-// same-day writes. Reads must never write the canonical; writes must always
-// persist it. ISE telemetry is deliberately excluded (48h-pruned, loss-
-// tolerant, ~2/min — snapshotting the whole store per heartbeat is waste;
-// any durable write that follows persists the pruning too).
-fn persist_canonical() -> Int {
-    let dir_raw: String = env("ENGRAM_DATA_DIR")
-    let dir: String = if str_eq(dir_raw, "") { "/tmp/engram" } else { dir_raw }
-    engram_save(dir + "/snapshot.json")
-    return 1
-}
-
 fn route_create_node(method: String, path: String, body: String) -> String {
     let content: String = json_get_string(body, "content")
-    let nt_raw: String = json_get_string(body, "node_type")
-    let node_type: String = if str_eq(nt_raw, "") { "Memory" } else { nt_raw }
-    let sal_raw: Float = json_get_float(body, "salience")
-    let salience: Float = if sal_raw == 0.0 { 0.5 } else { sal_raw }
+    let node_type: String = json_get_string(body, "node_type")
+    if str_eq(node_type, "") { let node_type = "Memory" }
+    let salience: Float = json_get_float(body, "salience")
+    if salience == 0.0 { let salience = 0.5 }
     let id: String = engram_node(content, node_type, salience)
-    let saved: Int = persist_canonical()
     "{\"id\":\"" + id + "\",\"content\":\"" + content + "\",\"node_type\":\"" + node_type + "\"}"
 }
 
@@ -133,14 +107,13 @@ fn route_scan_nodes(method: String, path: String, body: String) -> String {
 }
 
 // route_scan_edges — bulk export of all edges as a JSON array. Implemented
-// via engram_save → fs_read of a SCRATCH export path. (2026-07-21 self-review:
-// previously this saved over the canonical snapshot.json on every GET — if the
-// process ever booted with a partial/empty store, the first read request
-// clobbered the good snapshot. Read routes must never write the canonical path.)
+// via engram_save → fs_read of the canonical on-disk snapshot, which the
+// runtime keeps in lockstep with the in-memory graph. Live against the
+// running graph, not a stale export.
 fn route_scan_edges(method: String, path: String, body: String) -> String {
-    let dir_raw: String = env("ENGRAM_DATA_DIR")
-    let dir: String = if str_eq(dir_raw, "") { "/tmp/engram" } else { dir_raw }
-    let snap_path: String = dir + "/.scan-export.json"
+    let dir: String = env("ENGRAM_DATA_DIR")
+    if str_eq(dir, "") { let dir = "/tmp/engram" }
+    let snap_path: String = dir + "/snapshot.json"
     engram_save(snap_path)
     let snap: String = fs_read(snap_path)
     if str_eq(snap, "") { return "[]" }
@@ -153,34 +126,40 @@ fn route_scan_edges(method: String, path: String, body: String) -> String {
 }
 
 fn route_search(method: String, path: String, body: String) -> String {
-    let q: String = if str_eq(method, "GET") { query_param(path, "q") } else { json_get_string(body, "query") }
-    let lim_url: Int = query_int(path, "limit", 0)
-    let lim_body: Int = json_get_int(body, "limit")
-    let lim_either: Int = if lim_url > 0 { lim_url } else { lim_body }
-    let limit: Int = if lim_either > 0 { lim_either } else { 20 }
+    let q: String = ""
+    if str_eq(method, "GET") {
+        let q = query_param(path, "q")
+    } else {
+        let q = json_get_string(body, "query")
+    }
+    let limit: Int = query_int(path, "limit", 20)
+    if limit == 0 { let limit = json_get_int(body, "limit") }
+    if limit == 0 { let limit = 20 }
     return engram_search_json(q, limit)
 }
 
 fn route_activate(method: String, path: String, body: String) -> String {
-    let q: String = if str_eq(method, "GET") { query_param(path, "q") } else { json_get_string(body, "query") }
-    // Guard: engram_activate with an empty query matches zero seeds, which
-    // zeroes ALL carried working-memory weights (documented in awareness.el
-    // perceive()). Never let an empty activation through to wipe WM.
-    if str_eq(q, "") { return err_json("missing query") }
-    let d_raw: Int = if str_eq(method, "GET") { query_int(path, "depth", 3) } else { json_get_int(body, "depth") }
-    let depth: Int = if d_raw > 0 { d_raw } else { 3 }
+    let q: String = ""
+    let depth: Int = 3
+    if str_eq(method, "GET") {
+        let q = query_param(path, "q")
+        let depth = query_int(path, "depth", 3)
+    } else {
+        let q = json_get_string(body, "query")
+        let bd: Int = json_get_int(body, "depth")
+        if bd > 0 { let depth = bd }
+    }
     return "{\"results\":" + engram_activate_json(q, depth) + "}"
 }
 
 fn route_create_edge(method: String, path: String, body: String) -> String {
     let from_id: String = json_get_string(body, "from_id")
     let to_id: String = json_get_string(body, "to_id")
-    let rel_raw: String = json_get_string(body, "relation")
-    let relation: String = if str_eq(rel_raw, "") { "associates" } else { rel_raw }
-    let w_raw: Float = json_get_float(body, "weight")
-    let weight: Float = if w_raw == 0.0 { 0.5 } else { w_raw }
+    let relation: String = json_get_string(body, "relation")
+    if str_eq(relation, "") { let relation = "associates" }
+    let weight: Float = json_get_float(body, "weight")
+    if weight == 0.0 { let weight = 0.5 }
     engram_connect(from_id, to_id, weight, relation)
-    let saved: Int = persist_canonical()
     "{\"ok\":true,\"from_id\":\"" + from_id + "\",\"to_id\":\"" + to_id + "\",\"relation\":\"" + relation + "\"}"
 }
 
@@ -195,7 +174,6 @@ fn route_strengthen(method: String, path: String, body: String) -> String {
     let id: String = json_get_string(body, "node_id")
     if str_eq(id, "") { return err_json("missing node_id") }
     engram_strengthen(id)
-    let saved: Int = persist_canonical()
     ok_json()
 }
 
@@ -203,24 +181,27 @@ fn route_forget(method: String, path: String, body: String) -> String {
     let id: String = extract_id(path, "/api/nodes/")
     if str_eq(id, "") { return err_json("missing id") }
     engram_forget(id)
-    let saved: Int = persist_canonical()
     ok_json()
 }
 
 fn route_save(method: String, path: String, body: String) -> String {
-    let p_raw: String = json_get_string(body, "path")
-    let dir_raw: String = env("ENGRAM_DATA_DIR")
-    let dir: String = if str_eq(dir_raw, "") { "/tmp/engram" } else { dir_raw }
-    let p: String = if str_eq(p_raw, "") { dir + "/snapshot.json" } else { p_raw }
+    let p: String = json_get_string(body, "path")
+    if str_eq(p, "") {
+        let dir: String = env("ENGRAM_DATA_DIR")
+        if str_eq(dir, "") { let dir = "/tmp/engram" }
+        let p = dir + "/snapshot.json"
+    }
     engram_save(p)
     "{\"ok\":true,\"path\":\"" + p + "\"}"
 }
 
 fn route_load(method: String, path: String, body: String) -> String {
-    let p_raw: String = json_get_string(body, "path")
-    let dir_raw: String = env("ENGRAM_DATA_DIR")
-    let dir: String = if str_eq(dir_raw, "") { "/tmp/engram" } else { dir_raw }
-    let p: String = if str_eq(p_raw, "") { dir + "/snapshot.json" } else { p_raw }
+    let p: String = json_get_string(body, "path")
+    if str_eq(p, "") {
+        let dir: String = env("ENGRAM_DATA_DIR")
+        if str_eq(dir, "") { let dir = "/tmp/engram" }
+        let p = dir + "/snapshot.json"
+    }
     engram_load(p)
     ok_json()
 }
@@ -242,36 +223,15 @@ fn route_health(method: String, path: String, body: String) -> String {
 // (it skips nodes already present by ID). Auth-exempt: same-host internal call.
 // (2026-06-27 self-review: added this route to fix silent 10-min sync failures)
 fn route_sync(method: String, path: String, body: String) -> String {
-    let dir_raw: String = env("ENGRAM_DATA_DIR")
-    let dir: String = if str_eq(dir_raw, "") { "/tmp/engram" } else { dir_raw }
-    // 2026-07-21 self-review: export to a scratch path, never the canonical
-    // snapshot.json — read routes must not be able to clobber the good snapshot.
-    let snap_path: String = dir + "/.sync-export.json"
+    let dir: String = env("ENGRAM_DATA_DIR")
+    if str_eq(dir, "") { let dir = "/tmp/engram" }
+    let snap_path: String = dir + "/snapshot.json"
     engram_save(snap_path)
     let snap: String = fs_read(snap_path)
     if str_eq(snap, "") { return "{\"nodes\":[],\"edges\":[]}" }
     return snap
 }
 
-// route_load_merge — POST /api/load-merge {"path": "..."} — merge a snapshot
-// file into the live store WITHOUT resetting it (engram_load_merge skips nodes
-// already present by id). Added 2026-07-21 self-review to restore the 244 kn-
-// identity Knowledge nodes lost from the snapshot lineage between 05-13 and
-// 07-13. Requires an explicit path: refuses to run without one so it can never
-// be triggered accidentally against a default.
-fn route_load_merge(method: String, path: String, body: String) -> String {
-    let p: String = json_get_string(body, "path")
-    if str_eq(p, "") { return err_json("path is required") }
-    if str_eq(fs_read(p), "") { return err_json("file missing or empty") }
-    let before_n: Int = engram_node_count()
-    let before_e: Int = engram_edge_count()
-    engram_load_merge(p)
-    let added_n: Int = engram_node_count() - before_n
-    let added_e: Int = engram_edge_count() - before_e
-    let saved: Int = persist_canonical()
-    "{\"ok\":true,\"nodes_added\":" + int_to_str(added_n) + ",\"edges_added\":" + int_to_str(added_e) + ",\"node_count\":" + int_to_str(engram_node_count()) + "}"
-}
-
 // route_emit_ise — write an InternalStateEvent node from the soul daemon.
 //
 // Endpoint: POST /api/neuron/state-events
@@ -285,20 +245,10 @@ fn route_load_merge(method: String, path: String, body: String) -> String {
 //
 // Salience/importance set to match engram_node_full ISE defaults used by the
 // in-process fallback path in awareness.el (salience=0.3, importance=0.3,
-// confidence=0.8, tier=Episodic).
+// confidence=0.8, tier=Episodic). High temporal_decay_rate (1.617) — ISEs
+// are inherently transient; they should decay faster than structural knowledge.
 // (2026-06-26 self-review: added this route after discovering ise_post was
 // silently failing — the soul posts here but the endpoint didn't exist.)
-//
-// Retention (2026-07-16 self-review): an earlier comment here claimed ISEs
-// got temporal_decay_rate=1.617 — that was never implemented (engram_node_full
-// hardcodes 0.0), and per-node decay only dampens activation anyway; it never
-// removes nodes. By 2026-07-16 ISEs were 75% of the store (10,175 of 13,522
-// nodes, ~4,300/day, unbounded). ISEs are already WM-excluded in
-// engram_activate, so the fix is retention, not decay: every insert calls
-// engram_prune_telemetry(), a single O(nodes+edges) compaction pass that
-// removes ISEs older than ENGRAM_ISE_RETENTION_MS (default 48h), protecting
-// "session-start" labels and self_review events as durable history. At
-// ~3 ISEs/min this bounds telemetry at ~8.6k nodes instead of growing forever.
 fn route_emit_ise(method: String, path: String, body: String) -> String {
     let content: String = json_get_string(body, "content")
     if str_eq(content, "") { return err_json("missing content") }
@@ -310,64 +260,6 @@ fn route_emit_ise(method: String, path: String, body: String) -> String {
         sal, imp, conf,
         "Episodic", "[\"internal-state\",\"InternalStateEvent\"]"
     )
-    let ret_raw: String = env("ENGRAM_ISE_RETENTION_MS")
-    let ret_ms: Int = if str_eq(ret_raw, "") { 172800000 } else { str_to_int(ret_raw) }
-    let pruned: Int = engram_prune_telemetry(ret_ms)
-    "{\"ok\":true,\"id\":\"" + id + "\",\"pruned\":" + int_to_str(pruned) + "}"
-}
-
-// ── Knowledge capture ─────────────────────────────────────────────────────────
-//
-// route_capture_knowledge — direct Knowledge-node capture over HTTP.
-//
-// Endpoint: POST /api/neuron/knowledge/capture   (auth required: "_auth" in body)
-// Body:     {"content": "...", "title": "...", "category": "...",
-//            "tier": "note|lesson|canonical", "tags": [...], "project": "...",
-//            "_auth": "<key>"}
-//
-// WHY (2026-07-15 self-review): the world-ingestor integrator was designed
-// against this endpoint (its MCP-unavailable fallback), but the route never
-// existed — every direct push 404'd, and because the auth gate ran before
-// routing, the failure surfaced as {"error":"unauthorized"} and was
-// misdiagnosed for two weeks while world knowledge silently dropped.
-// POST /api/nodes was no substitute: it discards label/tags/tier, which
-// makes captured knowledge invisible to tag-scoped search and curiosity.
-//
-// The incoming knowledge tier (note/lesson/canonical) is preserved as a
-// "tier:<x>" tag rather than mapped onto Engram's cognitive tiers — Knowledge
-// nodes land in Semantic (stable reference), and the epistemic tier stays
-// queryable without inventing a lossy mapping.
-fn route_capture_knowledge(method: String, path: String, body: String) -> String {
-    let content: String = json_get_string(body, "content")
-    if str_eq(content, "") { return err_json("missing content") }
-    let title: String = json_get_string(body, "title")
-    let label: String = if str_eq(title, "") { str_slice(content, 0, 60) } else { title }
-    let category_raw: String = json_get_string(body, "category")
-    let category: String = if str_eq(category_raw, "") { "other" } else { category_raw }
-    let ktier_raw: String = json_get_string(body, "tier")
-    let ktier: String = if str_eq(ktier_raw, "") { "note" } else { ktier_raw }
-    let project: String = json_get_string(body, "project")
-    let tags_raw: String = json_get_raw(body, "tags")
-    let tags_base: String = if str_eq(tags_raw, "") { "[]" } else { tags_raw }
-    // Merge category/tier/project markers into the tag array. Search matches
-    // against the tags string, so these make captures findable by facet.
-    let base_len: Int = str_len(tags_base)
-    let head: String = str_slice(tags_base, 0, base_len - 1)
-    let sep: String = if str_eq(head, "[") { "" } else { "," }
-    let safe_cat: String = str_replace(category, "\"", "'")
-    let safe_tier: String = str_replace(ktier, "\"", "'")
-    let safe_proj: String = str_replace(project, "\"", "'")
-    let proj_tag: String = if str_eq(safe_proj, "") { "" } else { ",\"project:" + safe_proj + "\"" }
-    let tags: String = head + sep + "\"category:" + safe_cat + "\",\"tier:" + safe_tier + "\"" + proj_tag + "]"
-    let sal: Float = 0.5
-    let imp: Float = 0.5
-    let conf: Float = 0.9
-    let id: String = engram_node_full(
-        content, "Knowledge", label,
-        sal, imp, conf,
-        "Semantic", tags
-    )
-    let saved: Int = persist_canonical()
     "{\"ok\":true,\"id\":\"" + id + "\"}"
 }
 
@@ -407,12 +299,6 @@ fn handle_request(method: String, path: String, body: String) -> String {
         return err_json("unauthorized")
     }
 
-    // Knowledge capture (auth enforced above; the world-ingestor integrator
-    // and any headless session without MCP push knowledge through this)
-    if str_eq(method, "POST") && str_eq(clean, "/api/neuron/knowledge/capture") {
-        return route_capture_knowledge(method, path, body)
-    }
-
     // Stats
     if str_eq(method, "GET") && (str_eq(clean, "/api/stats") || str_eq(clean, "/stats")) {
         return route_stats(method, path, body)
@@ -469,9 +355,6 @@ fn handle_request(method: String, path: String, body: String) -> String {
     if str_eq(method, "POST") && (str_eq(clean, "/api/load") || str_eq(clean, "/load")) {
         return route_load(method, path, body)
     }
-    if str_eq(method, "POST") && (str_eq(clean, "/api/load-merge") || str_eq(clean, "/load-merge")) {
-        return route_load_merge(method, path, body)
-    }
 
     // Sync — soul daemon periodic pull of non-ISE knowledge into in-process graph
     if str_eq(method, "GET") && str_eq(clean, "/api/sync") {
@@ -483,31 +366,16 @@ fn handle_request(method: String, path: String, body: String) -> String {
 
 // ── Entry ─────────────────────────────────────────────────────────────────────
 
-let bind_raw: String = env("ENGRAM_BIND")
-let bind_str: String = if str_eq(bind_raw, "") { ":8742" } else { bind_raw }
+let bind_str: String = env("ENGRAM_BIND")
+if str_eq(bind_str, "") { let bind_str = ":8742" }
 let port: Int = parse_port(bind_str)
 
 // On startup, try to load any existing snapshot (best effort).
-let data_dir_raw: String = env("ENGRAM_DATA_DIR")
-let data_dir: String = if str_eq(data_dir_raw, "") { "/tmp/engram" } else { data_dir_raw }
+let data_dir: String = env("ENGRAM_DATA_DIR")
+if str_eq(data_dir, "") { let data_dir = "/tmp/engram" }
 let snapshot_path: String = data_dir + "/snapshot.json"
 engram_load(snapshot_path)
 
-// 2026-07-21 self-review boot guard: if the snapshot file has content but the
-// load produced 0 nodes, something is wrong (corrupt file / parse failure).
-// Preserve the evidence and warn loudly — and since read routes no longer write
-// the canonical path, a bad boot can no longer clobber the good snapshot.
-let boot_snap: String = fs_read(snapshot_path)
-if !str_eq(boot_snap, "") {
-    if engram_node_count() == 0 {
-        println("[engram] WARNING: snapshot.json is non-empty but load produced 0 nodes — preserving copy at snapshot.failed-load.json")
-        fs_write(data_dir + "/snapshot.failed-load.json", boot_snap)
-    } else {
-        // Good load: keep a boot-time backup of the snapshot as loaded.
-        fs_write(data_dir + "/snapshot.boot-backup.json", boot_snap)
-    }
-}
-
 println("[engram] runtime-native graph engine")
 println("[engram] data_dir=" + data_dir)
 println("[engram] node_count=" + int_to_str(engram_node_count()))

lang/el-compiler/runtime/el_runtime.c

@@ -6826,75 +6826,116 @@ static int istr_contains(const char* hay, const char* needle) {
     return 0;
 }
 
-/* ── Tokenized query matching ───────────────────────────────────────────
- * The engram query surface (search / activate / goal-bias) historically
- * matched the ENTIRE raw query string as a single case-insensitive
- * substring via istr_contains(field, q). That is Ctrl-F, not search:
- * a multi-word query like "windows msi signing" only matched a node whose
- * text contained that exact contiguous run, so real multi-word queries
- * returned zero. istr_contains stays as the per-TOKEN primitive; these
- * helpers split the query on whitespace and match ANY token, then rank by
- * how many DISTINCT tokens a node covers. Single-token queries are a strict
- * special case (score is 0 or 1) so single-word callers never regress. */
-#define ENGRAM_MAX_QTOKENS 32
-#define ENGRAM_QTOK_LEN    256
+/* ---- 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. */
 
-/* Split q on whitespace into up to ENGRAM_MAX_QTOKENS distinct
- * (case-insensitive) tokens. Returns the token count. Over-long tokens are
- * truncated to ENGRAM_QTOK_LEN-1; over-count tokens are ignored. */
-static int engram_tokenize_query(const char* q,
-                                 char toks[][ENGRAM_QTOK_LEN], int maxtok) {
+#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;
-    if (!q) return 0;
-    const char* p = q;
-    while (*p && n < maxtok) {
-        while (*p && isspace((unsigned char)*p)) p++;
-        if (!*p) break;
-        char buf[ENGRAM_QTOK_LEN];
-        size_t tl = 0;
-        while (*p && !isspace((unsigned char)*p)) {
-            if (tl < sizeof(buf) - 1) buf[tl++] = *p;
-            p++;
-        }
-        buf[tl] = '\0';
-        if (tl == 0) continue;
-        int dup = 0;
-        for (int s = 0; s < n; s++) {
-            if (strcasecmp(toks[s], buf) == 0) { dup = 1; break; }
-        }
-        if (dup) continue;
-        memcpy(toks[n], buf, tl + 1);
-        n++;
+    while (*s && isalnum((unsigned char)*s)) {
+        if (n < cap - 1) out[n++] = (char)tolower((unsigned char)*s);
+        s++;
     }
-    return n;
+    out[n] = 0; *ps = s; return 1;
 }
 
-/* Count how many of the ntok distinct query tokens appear (case-insensitive)
- * in the node's content, label, or tags. 0 == no match. */
-static int engram_node_match_score(const EngramNode* n,
-                                   char toks[][ENGRAM_QTOK_LEN], int ntok) {
-    int score = 0;
-    for (int t = 0; t < ntok; t++) {
-        if (istr_contains(n->content, toks[t]) ||
-            istr_contains(n->label,   toks[t]) ||
-            istr_contains(n->tags,    toks[t]))
-            score++;
+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]++;
     }
-    return score;
 }
 
-/* Rank entry: distinct-token match count (primary, desc) then salience
- * (tiebreak, desc). */
-typedef struct { int64_t idx; int score; double salience; } EngramRankEntry;
-static int engram_rank_cmp(const void* a, const void* b) {
-    const EngramRankEntry* ea = (const EngramRankEntry*)a;
-    const EngramRankEntry* eb = (const EngramRankEntry*)b;
-    if (ea->score != eb->score) return eb->score - ea->score;      /* desc */
-    if (ea->salience < eb->salience) return 1;
-    if (ea->salience > eb->salience) return -1;
+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);
@@ -6902,33 +6943,12 @@ 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;
-    char toks[ENGRAM_MAX_QTOKENS][ENGRAM_QTOK_LEN];
-    int ntok = engram_tokenize_query(q, toks, ENGRAM_MAX_QTOKENS);
-    if (ntok == 0) return lst;
-    EngramRankEntry* hits = malloc((size_t)g->node_count * sizeof(EngramRankEntry));
+    if (g->node_count == 0) return lst;
+    EngramHit* hits = (EngramHit*)malloc((size_t)g->node_count * sizeof(EngramHit));
     if (!hits) return lst;
-    int64_t nhits = 0;
-    for (int64_t i = 0; i < g->node_count; 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;
-        int sc = engram_node_match_score(n, toks, ntok);
-        if (sc > 0) {
-            hits[nhits].idx      = i;
-            hits[nhits].score    = sc;
-            hits[nhits].salience = n->salience;
-            nhits++;
-        }
-    }
-    /* Rank by distinct tokens matched (desc) then salience (desc), then cap. */
-    qsort(hits, (size_t)nhits, sizeof(EngramRankEntry), engram_rank_cmp);
-    int64_t end = nhits < lim ? nhits : lim;
-    for (int64_t k = 0; k < end; k++) {
-        lst = el_list_append(lst, engram_node_to_map(&g->nodes[hits[k].idx]));
-    }
+    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);
     return lst;
 }
@@ -7206,14 +7226,10 @@ static double engram_temporal_proximity_bonus(int64_t node_created,
 static double engram_goal_bias(const EngramNode* n, const char* query) {
     if (!query || !*query) return 1.0;
     double bias = 1.0;
-    /* Direct lexical overlap, graded by token coverage: a node covering all
-     * query tokens gets the full +0.5; partial coverage gets a proportional
-     * share. Single-token queries → full +0.5 on match, identical to before. */
-    {
-        char toks[ENGRAM_MAX_QTOKENS][ENGRAM_QTOK_LEN];
-        int ntok = engram_tokenize_query(query, toks, ENGRAM_MAX_QTOKENS);
-        int sc   = engram_node_match_score(n, toks, ntok);
-        if (sc > 0 && ntok > 0) bias += 0.5 * ((double)sc / (double)ntok);
+    /* Direct lexical overlap: node content/label/tags share text with query. */
+    if (istr_contains(n->content, query) || istr_contains(n->label, query) ||
+        istr_contains(n->tags, query)) {
+        bias += 0.5;
     }
     /* Node-type resonance with query intent. */
     int technical_query = istr_contains(query, "code")         ||
@@ -7279,21 +7295,14 @@ el_val_t engram_activate(el_val_t query, el_val_t depth) {
     if (!seeds) {
         free(best_bg); free(best_hops); free(reached); return out;
     }
-    /* Tokenize once: a node seeds if it matches ANY query token, and its seed
-     * activation is scaled by token coverage (fraction of distinct query
-     * tokens it contains) so a node matching all words seeds more strongly
-     * than one matching a single word. Single-word queries → coverage 1.0,
-     * identical to the prior whole-query behavior. */
-    char toks[ENGRAM_MAX_QTOKENS][ENGRAM_QTOK_LEN];
-    int ntok = engram_tokenize_query(q, toks, ENGRAM_MAX_QTOKENS);
     for (int64_t i = 0; i < g->node_count; i++) {
         EngramNode* n = &g->nodes[i];
-        int sc = engram_node_match_score(n, toks, ntok);
-        if (sc > 0) {
+        if (istr_contains(n->content, q) ||
+            istr_contains(n->label,   q) ||
+            istr_contains(n->tags,    q)) {
             double tdecay = engram_temporal_decay(n, now_ms);
             double dampen = engram_activation_dampen(n);
-            double cover  = ntok > 0 ? (double)sc / (double)ntok : 1.0;
-            double act    = n->salience * tdecay * dampen * cover;
+            double act    = n->salience * tdecay * dampen;
             seeds[seed_count].idx        = i;
             seeds[seed_count].act        = act;
             seeds[seed_count].created_at = n->created_at;
@@ -7855,44 +7864,23 @@ 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, '[');
-    int first = 1;
-    if (q && *q) {
-        char toks[ENGRAM_MAX_QTOKENS][ENGRAM_QTOK_LEN];
-        int ntok = engram_tokenize_query(q, toks, ENGRAM_MAX_QTOKENS);
-        if (ntok > 0) {
-            EngramRankEntry* hits =
-                malloc((size_t)g->node_count * sizeof(EngramRankEntry));
-            if (hits) {
-                int64_t nhits = 0;
-                for (int64_t i = 0; i < g->node_count; i++) {
-                    EngramNode* n = &g->nodes[i];
-                    /* Filter transparent layers — same as engram_search. */
-                    if (engram_layer_is_transparent(n->layer_id)) continue;
-                    int sc = engram_node_match_score(n, toks, ntok);
-                    if (sc > 0) {
-                        hits[nhits].idx      = i;
-                        hits[nhits].score    = sc;
-                        hits[nhits].salience = n->salience;
-                        nhits++;
-                    }
-                }
-                /* Rank by distinct tokens matched (desc) then salience (desc). */
-                qsort(hits, (size_t)nhits, sizeof(EngramRankEntry),
-                      engram_rank_cmp);
-                int64_t end = nhits < lim ? nhits : lim;
-                for (int64_t k = 0; k < end; k++) {
-                    if (!first) jb_putc(&b, ',');
-                    engram_emit_node_json(&b, &g->nodes[hits[k].idx]);
-                    first = 0;
-                }
-                free(hits);
+    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]);
             }
+            free(hits);
         }
     }
     jb_putc(&b, ']');

lang/el-compiler/src/codegen.el

@@ -3626,6 +3626,24 @@ fn codegen_streaming(tokens: [Any], sigs: [Map<String, Any>], source: String) ->
     let pos: Int = 0
     let el_main_body: [Map<String, Any>] = native_list_empty()
     let toplevel_exec_stmts: [Map<String, Any>] = native_list_empty()
+    // CGI IDENTITY CAPTURE (2026-08-09). A cgi block is a top-level DECLARATION, so
+    // the classifier below correctly excludes it from toplevel_exec_stmts and calls
+    // el_release on it. The identity emission further down then searched
+    // toplevel_exec_stmts for it — a list that structurally can never contain it —
+    // found nothing, and emitted nothing, silently. Measured: that search sees only
+    // [Let, Expr] for a program whose first statement is a cgi block.
+    // Fix: copy the values out BEFORE the release (strings, so no dangling reference)
+    // and emit from these. No search, so the failure mode is removed rather than moved.
+    let cgi_have: Bool = false
+    let cgi_name_v: String = ""
+    let cgi_did_v: String = ""
+    let cgi_prin_v: String = ""
+    let cgi_net_v: String = ""
+    let cgi_eng_v: String = ""
+    let cgi_has_did: Bool = false
+    let cgi_has_prin: Bool = false
+    let cgi_has_net: Bool = false
+    let cgi_has_eng: Bool = false
     let has_toplevel_exec: Bool = false
 
     let stream_running: Bool = true
@@ -3736,6 +3754,20 @@ fn codegen_streaming(tokens: [Any], sigs: [Map<String, Any>], source: String) ->
                         if is_top_level_decl(stmt) {
                             // Import, TypeDef, EnumDef, CgiBlock, ServiceBlock, ExternFn
                             // These are no-ops in codegen (forward decls already emitted)
+                            // — except a CgiBlock, whose declared identity must survive
+                            // this release to be emitted as a compiled constant.
+                            if str_eq(sk, "CgiBlock") {
+                                let cgi_have = true
+                                let cgi_name_v = stmt["name"]
+                                let cgi_did_v = stmt["dharma_id"]
+                                let cgi_prin_v = stmt["principal"]
+                                let cgi_net_v = stmt["network"]
+                                let cgi_eng_v = stmt["engram"]
+                                let cgi_has_did = stmt["has_dharma_id"]
+                                let cgi_has_prin = stmt["has_principal"]
+                                let cgi_has_net = stmt["has_network"]
+                                let cgi_has_eng = stmt["has_engram"]
+                            }
                             el_release(stmt)
                         } else {
                             if str_eq(sk, "Let") {
@@ -3814,33 +3846,17 @@ fn codegen_streaming(tokens: [Any], sigs: [Map<String, Any>], source: String) ->
         let sig2 = native_list_get(sigs, si2)
         let sk3: String = sig2["kind"]
         if str_eq(sk3, "cgi_block") {
-            // We need the full cgi_block data — it was parsed by scan_fn_sigs
-            // but scan only stored the name. For cgi_init we need dharma_id etc.
-            // Since cgi blocks are rare and small, they end up in toplevel_exec_stmts.
-            // Find the CgiBlock in toplevel_exec_stmts.
-            let tes_n: Int = native_list_len(toplevel_exec_stmts)
-            let tes_i: Int = 0
-            while tes_i < tes_n {
-                let tes = native_list_get(toplevel_exec_stmts, tes_i)
-                let tes_k: String = tes["stmt"]
-                if str_eq(tes_k, "CgiBlock") {
-                    let cname2: String = tes["name"]
-                    let cdid2: String = tes["dharma_id"]
-                    let cprin2: String = tes["principal"]
-                    let cnet2: String = tes["network"]
-                    let ceng2: String = tes["engram"]
-                    let has_did2: Bool = tes["has_dharma_id"]
-                    let has_prin2: Bool = tes["has_principal"]
-                    let has_net2: Bool = tes["has_network"]
-                    let has_eng2: Bool = tes["has_engram"]
-                    let arg_name2: String = "EL_STR(" + c_str_lit(cname2) + ")"
-                    let arg_did2: String = cgi_arg(cdid2, has_did2)
-                    let arg_prin2: String = cgi_arg(cprin2, has_prin2)
-                    let arg_net2: String = cgi_arg(cnet2, has_net2)
-                    let arg_eng2: String = cgi_arg(ceng2, has_eng2)
-                    emit_line("  el_cgi_init(" + arg_name2 + ", " + arg_did2 + ", " + arg_prin2 + ", " + arg_net2 + ", " + arg_eng2 + ");")
-                }
-                let tes_i = tes_i + 1
+            // Emit from the values captured before the declaration was released.
+            // The previous implementation searched toplevel_exec_stmts, which by
+            // construction never contains a declaration — so it emitted nothing and
+            // said nothing. See the capture block near toplevel_exec_stmts init.
+            if cgi_have {
+                let arg_name2: String = "EL_STR(" + c_str_lit(cgi_name_v) + ")"
+                let arg_did2: String = cgi_arg(cgi_did_v, cgi_has_did)
+                let arg_prin2: String = cgi_arg(cgi_prin_v, cgi_has_prin)
+                let arg_net2: String = cgi_arg(cgi_net_v, cgi_has_net)
+                let arg_eng2: String = cgi_arg(cgi_eng_v, cgi_has_eng)
+                emit_line("  el_cgi_init(" + arg_name2 + ", " + arg_did2 + ", " + arg_prin2 + ", " + arg_net2 + ", " + arg_eng2 + ");")
             }
         }
         let si2 = si2 + 1

lang/el-compiler/src/parser.el

@@ -23,29 +23,10 @@ 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)
 }
 
@@ -54,12 +35,7 @@ fn expect(tokens: [Any], pos: Int, kind: String) -> Int {
     if k == kind {
         return pos + 1
     }
-    // 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
-    }
+    // On mismatch just advance; error recovery is best-effort
     pos + 1
 }