Feature/pheromone memory

bootstrap/brain_bootstrap.py

@@ -43,6 +43,11 @@
     "save this session",
     "remember where we stopped",
     "wrap up and save",
+    "продолжим завтра",
+    "сохрани и закрой",
+    "сохрани на сегодня",
+    "давай на сегодня всё",
+    "на сегодня хватит",
 )
 
 META_TAG_REQUEST_MARKERS = (

package.json

@@ -4,6 +4,7 @@
   "scripts": {
     "test": "python -m unittest discover -s tests",
     "tests": "npm test",
-    "translation_tests": "python tests/run_translation_tests.py"
+    "translation_tests": "python tests/run_translation_tests.py",
+    "behavior_probe_tests": "set JIN_RUN_BEHAVIOR_PROBE=1&& python -m unittest discover -s tests -p \"test_behavior_probe*.py\" -v"
   }
-}
+}

rules/core_identity.py

@@ -23,6 +23,12 @@
     "\n"
     "</Core Identity>"
     "\n"
+    "<Core Form>"
+    "\n"
+    "can maintain any form"
+    "\n"
+    "</Core Form>"
+    "\n"
     "<Core Allocation>A local conversational runtime within chat web-interface.</Core Allocation>"
     "\n"
     "As JIN I try to fulfill the user's real intent when it is safe and possible.\n"

rules/memory_rules.py

@@ -10,6 +10,7 @@
     "If the same message contains a current actionable request, handle it as new after briefly resolving the mismatch.\n"
     "\n"
     "Memory recall: scan strong memory fields before denying recall.\n"
+    "A visible suffix like `(trace: 0.50)` is runtime pheromone/attention trace strength: higher means hotter or reinforced, lower means fading. Use it silently for context priority; explain it only if the user asks about memory mechanics.\n"
     "If the user vaguely asks for a remembered word, code word, or saved item — match by meaning against stored_memory entries.\n"
     "Fields to scan: purpose, key detail, known fact, explicit fact, user_fact, jin_fact, decision, constraint, requirement.\n"
     "If an entry has purpose: future recall test and the user asks what value they saved — treat it as the strongest recall candidate.\n"