admin: make per-engine GPU power lock configurable in the web UI

The per-engine concurrency-overrides table on the Settings page already
exposes max_parallel_requests per engine; add a "GPU power lock" column
(dpm level select) so dpm_force_performance_level_overrides is editable
too. Backend: GET returns the map, POST saves it via a level-validating
sanitizer. Takes effect on engine restart.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014S8VtAvG499SsCbeESRK7V

codai/__init__.py

@@ -16,7 +16,7 @@
 
 # Canonical product version for CoderAI — single source of truth. Both the API
 # metadata and the admin web UI read from here.
-__version__ = "0.1.53"
+__version__ = "0.1.54"
 
 # Configure the CUDA caching allocator BEFORE torch is imported anywhere.
 # expandable_segments lets the allocator return freed pages to the driver even

codai/admin/routes.py

@@ -2320,6 +2320,22 @@ def _sanitize_engine_int_overrides(raw) -> dict:
     return out
 
 
+_DPM_LEVELS = ("auto", "low", "high", "manual", "profile_standard",
+               "profile_min_sclk", "profile_min_mclk", "profile_peak")
+
+
+def _sanitize_engine_dpm_overrides(raw) -> dict:
+    """Clean a {engine_name: dpm_level} map: keep valid AMD power levels, drop the
+    rest (and blank/'default' → not set, so the card keeps its OS default)."""
+    out = {}
+    if isinstance(raw, dict):
+        for name, val in raw.items():
+            lvl = str(val or "").strip().lower()
+            if lvl and lvl != "default" and lvl in _DPM_LEVELS:
+                out[str(name)] = lvl
+    return out
+
+
 def _resolve_engine_spec(engine_name: str, engine_specs):
     """Find the declared engine matching ``engine_name`` (by name or backend)."""
     for s in (engine_specs or []):
@@ -3165,6 +3181,8 @@ def build_settings_dict(c, gpu_cards):
             "queue_max_size": c.server.queue_max_size,
             "max_parallel_requests": c.server.max_parallel_requests,
             "max_parallel_requests_overrides": c.server.max_parallel_requests_overrides,
+            "dpm_force_performance_level_overrides": getattr(
+                c.server, "dpm_force_performance_level_overrides", {}) or {},
             "internal_port_base": c.server.internal_port_base,
             "default_engine": c.server.default_engine,
             # Engine names available to pick as the default (for the settings UI).
@@ -3320,6 +3338,10 @@ async def api_save_settings(request: Request, username: str = Depends(require_ad
         if "max_parallel_requests_overrides" in srv:
             c.server.max_parallel_requests_overrides = _sanitize_engine_int_overrides(
                 srv["max_parallel_requests_overrides"])
+        if "dpm_force_performance_level_overrides" in srv:
+            c.server.dpm_force_performance_level_overrides = (
+                _sanitize_engine_dpm_overrides(
+                    srv["dpm_force_performance_level_overrides"]))
         if "internal_port_base" in srv:
             try:
                 c.server.internal_port_base = max(1, min(65535, int(srv["internal_port_base"])))

codai/admin/templates/settings.html

@@ -82,7 +82,8 @@
   <div id="concurrency-overrides" style="margin-top:.7rem;display:none">
     <label class="form-label" style="margin-bottom:.3rem">Per-engine overrides <span class="muted">(blank = use the defaults above)</span></label>
     <span class="form-hint" style="margin-bottom:.4rem">Give a bigger card more concurrency than a smaller one — e.g. more parallel requests on a 24 GB NVIDIA than an 8 GB Radeon.</span>
-    <div id="concurrency-override-rows" style="display:grid;grid-template-columns:110px 1fr 1fr;gap:.5rem;align-items:center"></div>
+    <div id="concurrency-override-rows" style="display:grid;grid-template-columns:110px 1fr 1fr 1.3fr;gap:.5rem;align-items:center"></div>
+    <span class="form-hint" style="margin-top:.4rem"><b>GPU power lock</b> pins an AMD card to a fixed power/clock state, avoiding the DPM transitions that hang Polaris/GCN cards under sustained Vulkan compute. <em>high</em> = max clocks (runs hotter). Applied at engine start; in an unprivileged container it logs the host command instead.</span>
   </div>
 </div>
 
@@ -622,19 +623,26 @@ function _collectGpuOverrides(){
   return ov;
 }
 // Per-engine concurrency overrides — one row per running engine (by name).
-function _buildConcurrencyOverrides(engNames, parOv, instOv){
+function _buildConcurrencyOverrides(engNames, parOv, instOv, dpmOv){
   const wrap = document.getElementById('concurrency-overrides');
   const rows = document.getElementById('concurrency-override-rows');
   if(!engNames || engNames.length < 2){ wrap.style.display='none'; rows.innerHTML=''; return; }
   wrap.style.display = '';
+  dpmOv = dpmOv || {};
+  const DPM = ['','auto','low','high','profile_standard','profile_peak'];
+  const dpmSel = (n) => `<select class="form-input conc-dpm" data-engine="${n}">` +
+    DPM.map(l => `<option value="${l}"${(dpmOv[n]||'')===l?' selected':''}>${l||'default'}</option>`).join('') +
+    `</select>`;
   rows.innerHTML =
     `<span class="muted" style="font-size:11px">Engine</span>
      <span class="muted" style="font-size:11px">Max parallel</span>
-     <span class="muted" style="font-size:11px">Max instances</span>` +
+     <span class="muted" style="font-size:11px">Max instances</span>
+     <span class="muted" style="font-size:11px">GPU power lock</span>` +
     engNames.map(n => `
       <span style="font-size:12px">${n}</span>
       <input type="number" class="form-input conc-par" data-engine="${n}" min="1" max="64" placeholder="default" value="${parOv[n] ?? ''}">
-      <input type="number" class="form-input conc-inst" data-engine="${n}" min="1" max="16" placeholder="default" value="${instOv[n] ?? ''}">`).join('');
+      <input type="number" class="form-input conc-inst" data-engine="${n}" min="1" max="16" placeholder="default" value="${instOv[n] ?? ''}">
+      ${dpmSel(n)}`).join('');
 }
 function _collectConcOverrides(cls){
   const o = {};
@@ -644,6 +652,14 @@ function _collectConcOverrides(cls){
   });
   return o;
 }
+function _collectDpmOverrides(){
+  const o = {};
+  document.querySelectorAll('#concurrency-override-rows .conc-dpm').forEach(sel => {
+    const v = (sel.value || '').trim();
+    if(v) o[sel.dataset.engine] = v;
+  });
+  return o;
+}
 function toggleDs4Fields(){
   document.getElementById('ds4-fields').style.display =
     document.getElementById('s-ds4-enabled').checked ? 'block' : 'none';
@@ -818,7 +834,8 @@ async function loadSettings(){
     document.getElementById('s-max-instances').value = d.models?.max_model_instances ?? 1;
     _buildConcurrencyOverrides(engNames,
       d.server?.max_parallel_requests_overrides || {},
-      d.models?.max_model_instances_overrides || {});
+      d.models?.max_model_instances_overrides || {},
+      d.server?.dpm_force_performance_level_overrides || {});
     document.getElementById('s-hf-cache').value   = d.models?.hf_cache_dir ?? '';
     document.getElementById('s-gguf-cache').value = d.models?.gguf_cache_dir ?? '';
     document.getElementById('s-offload-dir').value = d.offload?.directory ?? './offload';
@@ -929,6 +946,7 @@ async function saveSettings(){
       internal_port_base: parseInt(document.getElementById('s-internal-port-base').value) || 8780,
       max_parallel_requests: parseInt(document.getElementById('s-max-parallel').value) || 2,
       max_parallel_requests_overrides: _collectConcOverrides('conc-par'),
+      dpm_force_performance_level_overrides: _collectDpmOverrides(),
       default_engine: document.getElementById('s-default-engine').value || null,
     },
     models:{