Merge pull request #4134 from nesquena/stage-4130

Release NE (v0.51.392): align WebUI reasoning efforts to agent set, drop max (#3505 follow-up)

CHANGELOG.md

@@ -3,6 +3,12 @@
 
 ## [Unreleased]
 
+## [v0.51.392] — 2026-06-13 — Release NE (align WebUI reasoning efforts to the agent's accepted set, drop "max")
+
+### Fixed
+
+- **The WebUI no longer offers a "Max" reasoning effort the agent never accepted.** `VALID_REASONING_EFFORTS` is now `minimal/low/medium/high/xhigh` — matching `hermes-agent`'s own `VALID_REASONING_EFFORTS` (which has no `max`) — and the reasoning picker, `/reasoning` slash-command args, and help text drop the stale `Max` option. A previously-saved `reasoning_effort: max` is migrated to `xhigh` at resolve time (instead of becoming parser-invalid and silently dropping reasoning), so existing configs keep working. This is the "only expose what the model/agent actually supports" cleanup. (#3505 review follow-up)
+
 ## [v0.51.391] — 2026-06-13 — Release ND (Stable Assistant Turn Anchors renderer snapshot adapter, inert, #3926)
 
 ### Added

api/config.py

@@ -2361,7 +2361,8 @@ def get_effective_default_model(config_data: dict | None = None) -> str:
 # Mirrors hermes_constants.parse_reasoning_effort so WebUI can validate without
 # importing from the agent tree (which may not be installed).  Any drift here
 # will show up in the shared test suite since both sides accept the same set.
-VALID_REASONING_EFFORTS = ("minimal", "low", "medium", "high", "xhigh", "max")
+# Keep this WebUI-visible set aligned with hermes-agent#29248.
+VALID_REASONING_EFFORTS = ("minimal", "low", "medium", "high", "xhigh")
 
 
 def parse_reasoning_effort(effort):
@@ -2652,6 +2653,9 @@ def coerce_reasoning_effort_for_model(
         return ""
     if raw == "none":
         return "none"
+    accepts_max_as_xhigh = raw == "max"
+    if accepts_max_as_xhigh:
+        raw = "xhigh"
     if raw not in VALID_REASONING_EFFORTS:
         return ""
     supported = resolve_model_reasoning_efforts(
@@ -2663,21 +2667,21 @@ def coerce_reasoning_effort_for_model(
     # both for models KNOWN not to support reasoning AND for models we simply
     # don't recognize (custom providers, aggregator-rewritten ids, brand-new
     # releases). Coercion exists to avoid sending a level a KNOWN-incompatible
-    # model rejects (e.g. openai-codex gpt-5 'max', o1/o3/o4 above 'high') —
+    # model rejects (e.g. openai-codex gpt-5 'max', o1/o3/o4 above 'high') -
     # those paths return a NON-empty clamped set, so the degrade ladder below
     # still applies. When the set is empty we can't tell "unsupported" from
-    # "unknown", so preserve the user's configured effort verbatim (the prior
-    # behavior) rather than silently disabling reasoning — the provider stays
-    # the final authority. Worst case is the same rejected request that master
-    # already produces, i.e. no regression. (#3505 review)
+    # "unknown", so preserve the user's configured effort verbatim where it is
+    # still valid. A stale 'max' value is no longer parser-valid on the WebUI
+    # side, so degrade that unknown-model case to xhigh instead of silently
+    # dropping reasoning later in parse_reasoning_effort(). (#3505 review)
     if not supported:
-        return raw
+        return "xhigh" if accepts_max_as_xhigh else raw
     if raw in supported:
-        return raw
+        return "xhigh" if accepts_max_as_xhigh else raw
     # Degrade to the closest *lower* supported level instead of silently
     # disabling reasoning. e.g. max -> xhigh -> high, or xhigh -> high when the
     # target model caps below the configured effort. Never escalate.
-    ladder = list(VALID_REASONING_EFFORTS)  # ascending: minimal..max
+    ladder = list(VALID_REASONING_EFFORTS)  # ascending: minimal..xhigh
     try:
         raw_idx = ladder.index(raw)
     except ValueError:
@@ -2730,7 +2734,16 @@ def get_reasoning_status(
     return {
         # Match CLI default (True if unset in config.yaml)
         "show_reasoning": bool(show_raw) if isinstance(show_raw, bool) else True,
-        "reasoning_effort": str(effort_raw or "").strip().lower(),
+        # Report the COERCED effort (not the raw config value) so boot/status/chip
+        # read paths agree with what streaming actually sends — e.g. a stale
+        # `reasoning_effort: max` surfaces as `xhigh`, not the now-unsupported `max`.
+        # (Codex review of the drop-max alignment.)
+        "reasoning_effort": coerce_reasoning_effort_for_model(
+            str(effort_raw or "").strip().lower(),
+            resolve_model,
+            provider_id=resolve_provider,
+            base_url=resolve_base_url,
+        ),
         "supported_efforts": supported_efforts,
         "supports_reasoning_effort": bool(supported_efforts),
     }

static/commands.js

@@ -30,7 +30,7 @@ const COMMANDS=[
   {name:'background',desc:t('cmd_background'),fn:cmdBackground,arg:'prompt',  noEcho:true},
   {name:'status',    desc:t('cmd_status'),   fn:cmdStatus},
   {name:'voice',     desc:t('cmd_voice'),    fn:cmdVoice,     noEcho:true},
-  {name:'reasoning', desc:t('cmd_reasoning'), fn:cmdReasoning, arg:'show|hide|none|minimal|low|medium|high|xhigh|max', subArgs:['show','hide','none','minimal','low','medium','high','xhigh','max'], noEcho:true},
+  {name:'reasoning', desc:t('cmd_reasoning'), fn:cmdReasoning, arg:'show|hide|none|minimal|low|medium|high|xhigh', subArgs:['show','hide','none','minimal','low','medium','high','xhigh'], noEcho:true},
   {name:'yolo', desc:t('cmd_yolo'), fn:cmdYolo, noEcho:true},
   {name:'branch', desc:t('cmd_branch'), fn:cmdBranch, arg:'[name]', noEcho:true},
 ];
@@ -1383,13 +1383,14 @@ function cmdReasoning(args){
   const arg=(args||'').trim().toLowerCase();
   const BRAIN='\uD83E\uDDE0';
   // Matches hermes_constants.VALID_REASONING_EFFORTS + 'none' (CLI parity).
-  const EFFORTS=['none','minimal','low','medium','high','xhigh','max'];
+  // Keep this WebUI effort list in sync with hermes-agent#29248.
+  const EFFORTS=['none','minimal','low','medium','high','xhigh'];
   // Shared status renderer used by the no-args branch and as a fallback.
   function _fmtStatus(st){
     const vis=(st && st.show_reasoning===false)?'off':'on';
     const eff=(st && st.reasoning_effort)||'default';
     return BRAIN+' Reasoning effort: '+eff+' \u00B7 display: '+vis
-      +'  |  /reasoning show|hide|none|minimal|low|medium|high|xhigh|max';
+      +'  |  /reasoning show|hide|none|minimal|low|medium|high|xhigh';
   }
   if(!arg){
     // Status — read from the same config.yaml keys the CLI uses.

static/index.html

@@ -730,7 +730,6 @@
             <div class="reasoning-option" data-effort="medium">Medium</div>
             <div class="reasoning-option" data-effort="high">High</div>
             <div class="reasoning-option" data-effort="xhigh">Extra High</div>
-            <div class="reasoning-option" data-effort="max">Max</div>
           </div>
           <div class="composer-toolsets-dropdown" id="composerToolsetsDropdown">
             <div class="toolsets-dropdown-desc" id="toolsetsDropdownDesc"></div>

tests/test_issue1103_reasoning_chip_visibility.py

@@ -45,6 +45,9 @@ def test_reasoning_chip_html_starts_hidden():
         src
     )
     assert m, "composerReasoningWrap must start with style='display:none'"
+    assert 'data-effort="max"' not in src, (
+        "composer reasoning dropdown must not include Max"
+    )
 
 
 def test_ui_js_passes_model_context_to_reasoning_api():

tests/test_reasoning_effort_model_capabilities.py

@@ -83,7 +83,7 @@ def test_coerce_preserves_effort_for_unrecognized_model():
         "max",
         "brand-new-model-2099",
         provider_id="some-custom-provider",
-    ) == "max"
+    ) == "xhigh"
     # 'none' / unset still pass through unchanged for unknown models.
     assert cfg.coerce_reasoning_effort_for_model(
         "none", "some-unknown-model-xyz", provider_id="custom"
@@ -130,3 +130,30 @@ def test_get_reasoning_status_includes_supported_efforts(monkeypatch):
     )
     assert status["supported_efforts"] == ["low", "medium", "high"]
     assert status["supports_reasoning_effort"] is True
+
+
+def test_get_reasoning_status_for_reasoning_capable_model_has_no_max():
+    status = cfg.get_reasoning_status(
+        model_id="gpt-5.5",
+        provider_id="openai-codex",
+    )
+    assert status["supported_efforts"] == ["minimal", "low", "medium", "high", "xhigh"]
+    assert status["supports_reasoning_effort"] is True
+    assert "max" not in status["supported_efforts"]
+
+
+def test_get_reasoning_status_coerces_stale_max_to_xhigh(monkeypatch):
+    """A previously-saved `agent.reasoning_effort: max` (no longer a valid effort)
+    must be reported as the coerced `xhigh`, not the raw stale `max`, so the
+    boot/status/chip read paths agree with what streaming actually sends."""
+    monkeypatch.setattr(
+        cfg,
+        "_load_yaml_config_file",
+        lambda *a, **k: {"agent": {"reasoning_effort": "max"}},
+    )
+    status = cfg.get_reasoning_status(
+        model_id="gpt-5.5",
+        provider_id="openai-codex",
+    )
+    assert status["reasoning_effort"] == "xhigh"
+    assert status["reasoning_effort"] != "max"

tests/test_reasoning_show_hide.py

@@ -317,6 +317,9 @@ class TestReasoningCommand:
                 f"cmdReasoning must accept '{level}' (CLI parity with "
                 f"hermes_constants.parse_reasoning_effort)"
             )
+        assert "'max'" not in fn, (
+            "cmdReasoning should only expose efforts up to xhigh"
+        )
 
     def test_reasoning_subargs_match_cli_levels(self):
         """Autocomplete subArgs must expose every CLI effort level + show/hide."""
@@ -330,6 +333,9 @@ class TestReasoningCommand:
             assert f"'{suggestion}'" in entry, (
                 f"reasoning subArgs must include '{suggestion}' for CLI parity"
             )
+        assert "'max'" not in entry, (
+            "reasoning command subArgs must not include 'max'"
+        )
 
 
 # ── api/config.py — reasoning helpers ────────────────────────────────────────
@@ -359,7 +365,7 @@ class TestReasoningConfigHelpers:
         # Snapshot-style assertion: if hermes_constants adds a level, this
         # test will fail fast so we know to update WebUI too.
         assert VALID_REASONING_EFFORTS == (
-            'minimal', 'low', 'medium', 'high', 'xhigh', 'max'
+            'minimal', 'low', 'medium', 'high', 'xhigh'
         )
 
     def test_set_reasoning_effort_persists_to_config_yaml(self, tmp_path, monkeypatch):