@nexus-cortex/cli 4.125.2 → 4.125.4

package/.env.defaults

@@ -296,6 +296,13 @@ CORTEX_EFFORT_TAIL=false
 # [LEDGER 2026-09-11 @4.107.0] cell-n-r4 (hard core 13×K=2, flash action model, FINAL 47/52 rows): high = NO LIFT — ah-none 6/22 (0.27), ah-low 7/25 (0.28) vs control 11/26 (0.42); +3.5–4.5 min/task latency, turns unchanged (narrow door intact). Keep unset (card default). DARK. Ledger r-cell-n-r34.
 # [LEDGER 2026-09-12 @4.107.2] PRO @high on the k=5 series COMMON (tb21-k5-ph1, full 89, K=1): 72/89 = 80.9% (Wilson 0.72-0.88), $12.77, med 737 s — +3-4 pts over the flash series (77.3 ± 2.7), CIs overlap; effort NOT separated from model (pro @default not yet run). Flash @high = no lift (cell-n-r4). Stays DARK. Ledger .cortex/bench/r-tb21-k5-ph1-2026-09-12.md
 # [LEDGER 2026-09-18 @4.119.1] TB4.0 cell e6 (29 ever-passed tasks, 3-h budgets, flash, 1 repeat): LOW = REJECT 3/29 vs high 10/28 (+2/−9, p 0.07; tokens equal, wall +60%); MAX = WASH vs high 10/29 (+4/−4) at +30% tokens/+60% wall. Wire: reasoning/iteration FLAT across levels (4.6K/5.0K/4.8K chars), only the peak call scales (15.8K/22.7K/26.1K tok) — effort caps peak reasoning, not typical. Keep HIGH for TB4.0. Ledger r-tb4-e6-effort-2026-09-18.md.
+# [LEDGER 2026-10-17 @4.125.4] `none` = the action model runs with THINKING OFF (the non-reasoning arm). Live probe on deepseek-flash (1017):
+#   {thinking:enabled + reasoning_effort:none} still reasons; {reasoning_effort:none} alone and {thinking:{type:disabled}} do not. Until 4.125.4 the
+#   harness OMITTED reasoning_effort for `none` and sent no thinking field → the provider default (thinking ON) — a `none` arm would have been a
+#   silent control. Now APIClient sends thinking.type=disabled on every DeepSeek chat request with effort none (same field the forced-tool_choice
+#   path uses). Wire proof per row: reasoningTokens == 0 / reasoning_share == 0. History: effort LOW rejected (e6 09-18: 3/29 vs 10/29); no
+#   thinking-OFF action arm has ever run — hypothesis: no reasoning walls / degenerate loops (a thinking-mode phenomenon, ~20% of failures) and
+#   forced tool calls everywhere, at the cost of raw capability on the deciding calls. First test = 14-task subset vs arm C of t4w1.
 #CORTEX_ACTION_EFFORT=
 # Trigger helper-model review on tool errors (only applies when mentorship is enabled)
 MENTORSHIP_TRIGGER_ON_ERROR=false
@@ -689,6 +696,8 @@ CORTEX_LIFT_PLAN=true
 #               lift plan budget 6000 doctrine v1; git context / slash hint / web tools off). The cells ran the ADAPTER second-attempt chain
 #               (TB2_SECOND_ATTEMPT); the library chain stands in (mechanism parity, outcome gap open — r-tb4-replicate-2026-10-09.md).
 #   bench-lite = legacy R219 set, TTY sessions only (kept for configs naming it).
+#   container  = alias of standard (4.125.3): the hosted sandbox images exported CORTEX_PROFILE=container before this lever existed; without
+#                the alias every hosted MCP / autoresearch session resolved to OFF (found 10-17 verifying 4.125.2 in a fresh sandbox).
 #   KEPT DARK on purpose (evidence undetermined/negative or bench-only): task rules + verify-scale (t4r 3/11 vs 6/11), turn status (needs a
 #   deadline), veto cap 3 (costs passes on short tasks), inventory (blocks the TUI loop), workspace clean, second-attempt chain (would redo a
 #   human's finish), output cap (any cap < 63K reaches passing calls), meets-confirm (null), orient baseline / PID1 guard / plateau stop.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nexus-cortex/cli",
-  "version": "4.125.2",
+  "version": "4.125.4",
   "description": "Nexus Cortex CLI - Terminal interface for multi-provider LLM orchestration",
   "main": "dist/index.js",
   "type": "module",
@@ -19,7 +19,7 @@
     "prepack": "node ../../scripts/copy-pkg-cortex-scaffold.mjs"
   },
   "dependencies": {
-    "@nexus-cortex/core": "4.125.2",
+    "@nexus-cortex/core": "4.125.4",
     "chalk": "^4.1.2",
     "cli-spinners": "^2.9.0",
     "commander": "^11.0.0",