Upload folder using huggingface_hub

actions/parser.py

@@ -24,6 +24,7 @@ from ..simulation.power import PowerSimulation
 from ..simulation.types import (
     CRACFaultType,
     CRACStatus,
+    GeneratorState,
     UPSMode,
 )
 
@@ -134,6 +135,53 @@ def _handle_diagnose(
                 ]
                 return CommandResult(True, "\n".join(lines), "diagnose", target)
 
+    # Check Generator
+    if power:
+        gen = power.state.generator
+        if gen.gen_id.lower() == target.lower():
+            lines = [
+                f"=== Diagnostic Report: {gen.gen_id} ===",
+                f"State: {gen.state.value}",
+                f"Output: {gen.output_power_kw:.1f} kW",
+                f"Load: {gen.load_fraction * 100:.1f}%",
+                f"Fuel Level: {gen.fuel_level_liters:.0f} L / {gen.fuel_tank_liters:.0f} L",
+                f"Fuel Consumption: {gen.fuel_consumption_lph:.1f} L/hr",
+                f"Fuel Remaining: {gen.fuel_remaining_hours:.1f} hrs",
+                f"Rated Capacity: {gen.rated_capacity_kw:.0f} kW",
+            ]
+            if gen.state == GeneratorState.OFF:
+                lines.append(">> Generator is OFF.")
+            elif gen.state in (GeneratorState.START_DELAY, GeneratorState.CRANKING, GeneratorState.WARMING):
+                lines.append(f">> Generator starting ({gen.state.value}), elapsed {gen.state_elapsed_s:.1f}s.")
+            elif gen.state == GeneratorState.READY:
+                lines.append(">> Generator READY to accept load.")
+            elif gen.state == GeneratorState.LOADED:
+                lines.append(f">> Generator LOADED at {gen.load_fraction * 100:.0f}%. Operating normally.")
+            elif gen.state == GeneratorState.COOLDOWN:
+                lines.append(f">> Generator in COOLDOWN, elapsed {gen.state_elapsed_s:.1f}s.")
+            return CommandResult(True, "\n".join(lines), "diagnose", target)
+
+    # Check PDUs
+    if power:
+        for pdu in power.state.pdus:
+            if pdu.pdu_id.lower() == target.lower():
+                lines = [
+                    f"=== Diagnostic Report: {pdu.pdu_id} ===",
+                    f"Input Power: {pdu.input_power_kw:.1f} kW",
+                    f"Output Power: {pdu.output_power_kw:.1f} kW",
+                    f"Load: {pdu.load_fraction * 100:.1f}%",
+                    f"Phase Imbalance: {pdu.phase_imbalance_pct:.1f}%",
+                    f"Breaker Tripped: {pdu.breaker_tripped}",
+                    f"Overload: {pdu.overload}",
+                ]
+                if pdu.breaker_tripped:
+                    lines.append(">> FAULT: Breaker tripped!")
+                elif pdu.overload:
+                    lines.append(">> WARNING: PDU overloaded.")
+                else:
+                    lines.append(">> No faults detected. PDU operating normally.")
+                return CommandResult(True, "\n".join(lines), "diagnose", target)
+
     return CommandResult(False, f"Unit '{target}' not found.", "diagnose", target)
 
 

scenarios/power_scenarios.py

@@ -215,7 +215,12 @@ class GeneratorTestProtocol(Scenario):
         if cmd.startswith("start_generator"):
             self._started = True
         if self._started and cmd.startswith("diagnose") and "gen" in cmd:
-            self._verified = True
+            # Only count as verified if generator is actually running
+            if power_sim and power_sim.state.generator.state in (
+                GeneratorState.READY, GeneratorState.LOADED,
+                GeneratorState.WARMING, GeneratorState.CRANKING,
+            ):
+                self._verified = True
         if cmd.startswith("stop_generator"):
             if self._started and self._verified:
                 self._stopped = True
@@ -289,7 +294,7 @@ class GeneratorTestProtocol(Scenario):
 
     @property
     def step_budget(self) -> int:
-        return 10
+        return 15
 
     @property
     def alert_message(self) -> str:
@@ -319,8 +324,9 @@ class GeneratorTestProtocol(Scenario):
 
     @property
     def game_time_per_step_s(self) -> float:
-        # Generator startup is ~17s, so 30s per step lets agent observe transitions
-        return 30.0
+        # Generator startup is ~17s. At 10s/step the agent can observe
+        # intermediate states (CRANKING, WARMING) across 2 steps.
+        return 10.0
 
 
 # ===========================================================================
@@ -346,13 +352,16 @@ class PowerFailureCascade(Scenario):
     """
 
     _CONSECUTIVE_STABLE_STEPS = 2
+    _MIN_STEPS_BEFORE_RESOLUTION = 4  # Hard scenario needs investigation
 
     def __init__(self) -> None:
         super().__init__()
         self._stable_count = 0
+        self._diagnosed_ups = False
 
     def reset_state(self) -> None:
         self._stable_count = 0
+        self._diagnosed_ups = False
 
     def configure(self, base_config: DatacenterConfig) -> DatacenterConfig:
         # Extend generator warmup to make it more challenging
@@ -378,6 +387,12 @@ class PowerFailureCascade(Scenario):
     ) -> ScenarioResult:
         dc = thermal_sim.state
 
+        # Track if agent diagnosed UPS
+        cmd_parts = action_command.strip().split()
+        if (len(cmd_parts) >= 2 and cmd_parts[0].lower() == "diagnose"
+                and "ups" in cmd_parts[1].lower()):
+            self._diagnosed_ups = True
+
         # Check temperatures
         all_within_allowable = True
         max_over = 0.0
@@ -404,7 +419,15 @@ class PowerFailureCascade(Scenario):
         else:
             self._stable_count = 0
 
-        resolved = self._stable_count >= self._CONSECUTIVE_STABLE_STEPS
+        # Resolution requires:
+        #   1. Agent diagnosed UPS status (proper incident response)
+        #   2. Stable for N consecutive steps
+        #   3. At least _MIN_STEPS_BEFORE_RESOLUTION steps taken
+        resolved = (
+            self._stable_count >= self._CONSECUTIVE_STABLE_STEPS
+            and self._diagnosed_ups
+            and step >= self._MIN_STEPS_BEFORE_RESOLUTION
+        )
 
         # Reward shaping
         scenario_reward = 0.0

scenarios/thermal_scenarios.py

@@ -161,16 +161,19 @@ class ThermalEventResponse(Scenario):
 
     _FAILED_UNIT = "CRAC-3"
     _CONSECUTIVE_STABLE_STEPS = 2
+    _MIN_STEPS_BEFORE_RESOLUTION = 5  # Agent must take at least 5 actions
 
     def __init__(self) -> None:
         super().__init__()
         self._stable_count = 0
+        self._diagnosed_fault = False  # Must diagnose the faulty unit
 
     def reset_state(self) -> None:
         self._stable_count = 0
+        self._diagnosed_fault = False
 
     def configure(self, base_config: DatacenterConfig) -> DatacenterConfig:
-        return base_config  # Default config is fine
+        return base_config
 
     def inject_fault(
         self,
@@ -189,6 +192,12 @@ class ThermalEventResponse(Scenario):
     ) -> ScenarioResult:
         dc = thermal_sim.state
 
+        # Track if agent diagnosed the faulty unit
+        cmd_parts = action_command.strip().split()
+        if (len(cmd_parts) >= 2 and cmd_parts[0].lower() == "diagnose"
+                and cmd_parts[1].upper() == self._FAILED_UNIT):
+            self._diagnosed_fault = True
+
         # Check if all zones within recommended
         all_within_recommended = True
         max_over = 0.0
@@ -205,7 +214,15 @@ class ThermalEventResponse(Scenario):
         else:
             self._stable_count = 0
 
-        resolved = self._stable_count >= self._CONSECUTIVE_STABLE_STEPS
+        # Resolution requires:
+        #   1. Agent diagnosed the faulty unit (proper procedure)
+        #   2. Temps stable for N consecutive steps
+        #   3. At least _MIN_STEPS_BEFORE_RESOLUTION steps taken
+        resolved = (
+            self._diagnosed_fault
+            and self._stable_count >= self._CONSECUTIVE_STABLE_STEPS
+            and step >= self._MIN_STEPS_BEFORE_RESOLUTION
+        )
 
         # Scenario reward: penalty proportional to temperature overshoot
         scenario_reward = -max_over * 0.5 if max_over > 0 else 0.1
@@ -304,6 +321,7 @@ class CRACFailureCascade(Scenario):
         ("CRAC-3", CRACFaultType.FAN),
     ]
     _CONSECUTIVE_STABLE_STEPS = 2
+    _MIN_STEPS_BEFORE_RESOLUTION = 5  # Hard scenario needs investigation time
 
     def __init__(self) -> None:
         super().__init__()
@@ -348,19 +366,25 @@ class CRACFailureCascade(Scenario):
         else:
             self._stable_count = 0
 
-        resolved = self._stable_count >= self._CONSECUTIVE_STABLE_STEPS
-
-        # Heavy penalty for being over allowable
-        scenario_reward = -max_over * 2.0 if max_over > 0 else 0.2
-
-        procedure_reward = self.check_procedure(action_command, action_history)
-
         # Bonus for diagnosing both units
         diagnosed_units = set()
         for h in action_history:
             parts = h.strip().split()
             if len(parts) >= 2 and parts[0].lower() == "diagnose":
                 diagnosed_units.add(parts[1].upper())
+
+        resolved = (
+            self._stable_count >= self._CONSECUTIVE_STABLE_STEPS
+            and "CRAC-1" in diagnosed_units
+            and "CRAC-3" in diagnosed_units
+            and step >= self._MIN_STEPS_BEFORE_RESOLUTION
+        )
+
+        # Heavy penalty for being over allowable
+        scenario_reward = -max_over * 2.0 if max_over > 0 else 0.2
+
+        procedure_reward = self.check_procedure(action_command, action_history)
+
         if "CRAC-1" in diagnosed_units and "CRAC-3" in diagnosed_units:
             procedure_reward += 0.2  # Bonus for thorough diagnosis
 

server/dc_ops_env_environment.py

@@ -179,6 +179,7 @@ class DcOpsEnvironment(Environment):
 
         # Initialize reward function with scenario-type-aware weights
         self._reward_fn = RewardFunction(scenario_type=self._scenario_type)
+        self._reward_fn.reset()  # Defensive: ensure clean state
 
         # Initialize simulations
         self._thermal_sim = ThermalSimulation(self._config)
@@ -261,13 +262,17 @@ class DcOpsEnvironment(Environment):
                 reward=reward,
             )
 
-        # 2. Advance simulation
+        # 2. Handle acknowledge_alarm — clear alert before new alarms overwrite
+        if cmd_result.command_name == "acknowledge_alarm" and cmd_result.success:
+            self._alert = ""
+
+        # 3. Advance simulation
         thermal_alarms, power_alarms = self._advance_simulation()
 
-        # 3. Build alert from alarms
+        # 4. Build alert from alarms (only new critical/warning alarms override)
         self._update_alert(thermal_alarms, power_alarms)
 
-        # 4. Evaluate scenario (before reward, so progress is available)
+        # 5. Evaluate scenario (before reward, so progress is available)
         scenario_result = None
         if self._scenario:
             scenario_result = self._scenario.evaluate_step(
@@ -276,7 +281,7 @@ class DcOpsEnvironment(Environment):
                 self._state.step_count,
             )
 
-        # 5. Compute reward via RewardFunction
+        # 6. Compute reward via RewardFunction
         components = self._reward_fn.compute(
             self._thermal_sim, self._power_sim, cmd_result,
             action.command, self._action_history, scenario_result,
@@ -285,10 +290,10 @@ class DcOpsEnvironment(Environment):
 
         self._cumulative_reward += reward
 
-        # 6. Check termination
+        # 7. Check termination
         self._check_termination(thermal_alarms, power_alarms)
 
-        # 6b. Scenario resolution
+        # 7b. Scenario resolution
         if scenario_result and scenario_result.resolved and not self._done:
             self._done = True
             # Speed bonus: fraction of budget remaining

server/static/index.html

@@ -219,6 +219,106 @@ body{background:var(--bg);color:var(--text);font-family:var(--font-sans);min-hei
 @media(max-width:1100px){.mobile-toggles{display:flex}}
 .toggle-btn{padding:0.25rem 0.625rem;background:transparent;border:1px solid var(--border);border-radius:var(--radius);color:var(--text-dim);font-size:0.7rem;cursor:pointer;font-family:var(--font-sans);transition:all 0.15s}
 .toggle-btn.active{border-color:var(--accent);color:var(--accent);background:rgba(59,130,246,0.08)}
+
+/* ─── Tab Navigation ─── */
+.header-tabs{display:flex;gap:0.25rem;background:var(--terminal-bg);border-radius:var(--radius);padding:0.2rem}
+.tab-btn{padding:0.375rem 1rem;border-radius:6px;border:none;background:transparent;color:var(--text-dim);font-size:0.78rem;font-weight:600;font-family:var(--font-sans);cursor:pointer;transition:all 0.2s;white-space:nowrap;position:relative}
+.tab-btn:hover{color:var(--text)}
+.tab-btn.active{background:var(--accent);color:#fff;box-shadow:0 1px 4px rgba(59,130,246,0.3)}
+.tab-content{display:none;min-height:0}
+.tab-content.active{display:block;overflow:hidden;min-height:0}
+.tab-content.active > .main{height:100%}
+
+/* ─── Guide Page ─── */
+.guide-page{display:none;overflow-y:auto;padding:2rem 1.5rem;background:var(--bg)}
+.guide-page.active{display:block}
+.guide-inner{max-width:920px;margin:0 auto}
+.guide-hero{text-align:center;padding:2.5rem 1rem 2rem;margin-bottom:2rem}
+.guide-hero h1{font-size:2rem;font-weight:800;letter-spacing:-0.03em;margin-bottom:0.5rem}
+.guide-hero h1 span{color:var(--accent)}
+.guide-hero p{color:var(--text-dim);font-size:0.95rem;max-width:560px;margin:0 auto;line-height:1.7}
+
+/* Guide sections */
+.guide-section{margin-bottom:2.5rem}
+.guide-section-header{display:flex;align-items:center;gap:0.75rem;margin-bottom:1.25rem;padding-bottom:0.75rem;border-bottom:1px solid var(--border)}
+.guide-section-icon{width:36px;height:36px;border-radius:var(--radius);display:flex;align-items:center;justify-content:center;font-size:1.1rem;flex-shrink:0}
+.guide-section-icon.blue{background:rgba(59,130,246,0.12);color:var(--accent)}
+.guide-section-icon.green{background:rgba(34,197,94,0.12);color:var(--green)}
+.guide-section-icon.orange{background:rgba(249,115,22,0.12);color:var(--orange)}
+.guide-section-icon.red{background:rgba(239,68,68,0.12);color:var(--red)}
+.guide-section-icon.cyan{background:rgba(6,182,212,0.12);color:var(--cyan)}
+.guide-section-icon.yellow{background:rgba(234,179,8,0.12);color:var(--yellow)}
+.guide-section h2{font-size:1.15rem;font-weight:700;letter-spacing:-0.01em}
+.guide-section p,.guide-section li{font-size:0.85rem;color:var(--text-dim);line-height:1.75}
+.guide-section strong{color:var(--text)}
+.guide-section ul,.guide-section ol{padding-left:1.25rem;margin:0.75rem 0}
+.guide-section li{margin-bottom:0.35rem}
+
+/* Guide cards grid */
+.guide-cards{display:grid;grid-template-columns:repeat(auto-fill,minmax(260px,1fr));gap:0.75rem}
+.guide-card{background:var(--bg-card);border:1px solid var(--border);border-radius:var(--radius-lg);padding:1rem 1.25rem;transition:border-color 0.2s}
+.guide-card:hover{border-color:var(--border-active)}
+.guide-card .gc-header{display:flex;align-items:center;justify-content:space-between;margin-bottom:0.5rem}
+.guide-card .gc-id{font-family:var(--font-mono);font-weight:800;font-size:0.9rem;color:var(--accent)}
+.guide-card .gc-diff{font-size:0.6rem;font-weight:700;padding:0.1rem 0.5rem;border-radius:999px;text-transform:uppercase;letter-spacing:0.05em}
+.guide-card .gc-name{font-weight:600;font-size:0.85rem;margin-bottom:0.375rem;color:var(--text)}
+.guide-card .gc-desc{font-size:0.78rem;color:var(--text-dim);line-height:1.6;margin-bottom:0.625rem}
+.guide-card .gc-hint{font-size:0.72rem;color:var(--cyan);font-family:var(--font-mono);background:var(--terminal-bg);padding:0.375rem 0.625rem;border-radius:4px;border:1px solid var(--border)}
+.guide-card .gc-hint strong{color:var(--text);font-size:0.68rem;text-transform:uppercase;letter-spacing:0.05em;display:block;margin-bottom:0.2rem}
+
+/* Guide table */
+.guide-table{width:100%;border-collapse:collapse;font-size:0.78rem;margin:0.75rem 0;border:1px solid var(--border);border-radius:var(--radius);overflow:hidden}
+.guide-table thead{background:var(--bg-card)}
+.guide-table th{text-align:left;padding:0.625rem 0.875rem;font-weight:600;color:var(--text);border-bottom:1px solid var(--border);font-size:0.72rem;text-transform:uppercase;letter-spacing:0.05em}
+.guide-table td{padding:0.5rem 0.875rem;border-bottom:1px solid var(--border);color:var(--text-dim);vertical-align:top}
+.guide-table tr:last-child td{border-bottom:none}
+.guide-table tbody tr:hover{background:var(--bg-card-hover)}
+.guide-table code{font-family:var(--font-mono);font-size:0.75rem;color:var(--cyan);background:var(--terminal-bg);padding:0.1rem 0.375rem;border-radius:3px}
+.guide-table .tag{display:inline-block;font-size:0.6rem;font-weight:700;padding:0.1rem 0.4rem;border-radius:999px;text-transform:uppercase}
+.tag-pos{background:var(--green-dim);color:var(--green)}
+.tag-neg{background:var(--red-dim);color:var(--red)}
+.tag-range{background:rgba(6,182,212,0.12);color:var(--cyan)}
+
+/* Guide code block */
+.guide-code{background:var(--terminal-bg);border:1px solid var(--border);border-radius:var(--radius);padding:0.875rem 1rem;font-family:var(--font-mono);font-size:0.75rem;line-height:1.6;color:var(--green);overflow-x:auto;margin:0.75rem 0;white-space:pre}
+
+/* Reward component cards */
+.reward-grid{display:grid;grid-template-columns:repeat(auto-fill,minmax(280px,1fr));gap:0.75rem;margin:1rem 0}
+.reward-card{background:var(--bg-card);border:1px solid var(--border);border-radius:var(--radius-lg);padding:1rem 1.25rem}
+.reward-card h4{font-size:0.85rem;font-weight:700;margin-bottom:0.375rem;display:flex;align-items:center;gap:0.5rem}
+.reward-card h4 .rc-range{font-size:0.65rem;font-family:var(--font-mono);color:var(--text-muted);font-weight:500}
+.reward-card p{font-size:0.78rem;color:var(--text-dim);line-height:1.6}
+.reward-card .rc-formula{font-family:var(--font-mono);font-size:0.72rem;color:var(--cyan);background:var(--terminal-bg);padding:0.35rem 0.625rem;border-radius:4px;margin-top:0.5rem;border:1px solid var(--border);display:inline-block}
+
+/* Guide weight profile inline table */
+.weight-profiles{display:grid;grid-template-columns:repeat(auto-fit,minmax(200px,1fr));gap:0.75rem;margin:1rem 0}
+.weight-profile{background:var(--bg-card);border:1px solid var(--border);border-radius:var(--radius);padding:0.875rem 1rem}
+.weight-profile h4{font-size:0.8rem;font-weight:700;margin-bottom:0.625rem;color:var(--accent)}
+.weight-bar-row{display:flex;align-items:center;gap:0.5rem;margin-bottom:0.35rem;font-size:0.72rem}
+.weight-bar-label{width:72px;color:var(--text-dim);flex-shrink:0}
+.weight-bar-track{flex:1;height:8px;background:var(--terminal-bg);border-radius:4px;overflow:hidden}
+.weight-bar-fill{height:100%;border-radius:4px;background:var(--accent);transition:width 0.3s}
+.weight-bar-val{width:32px;text-align:right;font-family:var(--font-mono);color:var(--text-muted);font-weight:600;flex-shrink:0}
+
+/* ASHRAE visual */
+.ashrae-visual{display:flex;flex-direction:column;gap:0.75rem;margin:1rem 0}
+.ashrae-row{display:flex;align-items:center;gap:0.75rem;font-size:0.78rem}
+.ashrae-label{width:32px;font-family:var(--font-mono);font-weight:700;color:var(--accent);flex-shrink:0}
+.ashrae-bar-container{flex:1;position:relative;height:24px}
+.ashrae-bar-bg{position:absolute;inset:0;background:var(--terminal-bg);border-radius:4px;border:1px solid var(--border)}
+.ashrae-bar-rec{position:absolute;height:100%;background:rgba(34,197,94,0.2);border:1px solid rgba(34,197,94,0.4);border-radius:4px}
+.ashrae-bar-allow{position:absolute;height:100%;background:rgba(234,179,8,0.1);border:1px solid rgba(234,179,8,0.25);border-radius:4px}
+.ashrae-bar-label{position:absolute;top:50%;transform:translateY(-50%);font-size:0.65rem;font-family:var(--font-mono);color:var(--text-dim)}
+
+@media(max-width:640px){
+  .guide-page{padding:1rem 0.75rem}
+  .guide-hero h1{font-size:1.5rem}
+  .guide-cards{grid-template-columns:1fr}
+  .reward-grid{grid-template-columns:1fr}
+  .weight-profiles{grid-template-columns:1fr}
+  .guide-table{font-size:0.7rem}
+  .guide-table th,.guide-table td{padding:0.375rem 0.5rem}
+}
 </style>
 </head>
 <body>
@@ -233,13 +333,18 @@ body{background:var(--bg);color:var(--text);font-family:var(--font-sans);min-hei
         <span id="statusText">Disconnected</span>
       </div>
     </div>
+    <div class="header-tabs">
+      <button class="tab-btn active" data-tab="console" onclick="switchTab('console')">Console</button>
+      <button class="tab-btn" data-tab="guide" onclick="switchTab('guide')">Guide</button>
+    </div>
     <div class="mobile-toggles">
       <button class="toggle-btn active" id="toggleScenarios" onclick="togglePanel('sidebar')">Scenarios</button>
       <button class="toggle-btn active" id="toggleMetrics" onclick="togglePanel('right-panel')">Metrics</button>
     </div>
   </header>
 
-  <!-- Main Layout -->
+  <!-- Main Layout (Console Tab) -->
+  <div class="tab-content active" id="tabConsole">
   <div class="main">
     <!-- Left: Scenario Browser -->
     <aside class="sidebar" id="sidebar">
@@ -346,13 +451,6 @@ Pick a scenario to start
           <button id="sendBtn" class="btn btn-primary" onclick="sendCommand()" disabled>Send</button>
         </div>
         <div class="quick-actions" id="quickActions">
-          <button class="quick-btn" disabled onclick="quickCmd('check_status')">check_status</button>
-          <button class="quick-btn" disabled onclick="quickCmd('diagnose CRAC-1')">diagnose CRAC-1</button>
-          <button class="quick-btn" disabled onclick="quickCmd('diagnose CRAC-3')">diagnose CRAC-3</button>
-          <button class="quick-btn" disabled onclick="quickCmd('acknowledge_alarm')">ack_alarm</button>
-          <button class="quick-btn" disabled onclick="quickCmd('start_generator')">start_gen</button>
-          <button class="quick-btn" disabled onclick="quickCmd('wait')">wait</button>
-          <button class="quick-btn" disabled onclick="quickCmd('escalate')">escalate</button>
         </div>
       </div>
     </div>
@@ -424,6 +522,350 @@ Pick a scenario to start
       </div>
     </aside>
   </div>
+  </div><!-- /tabConsole -->
+
+  <!-- Guide Tab -->
+  <div class="guide-page" id="tabGuide">
+  <div class="guide-inner">
+
+    <div class="guide-hero">
+      <h1>DC<span>-Ops</span> Operations Guide</h1>
+      <p>A comprehensive reference for operating the physics-based datacenter simulation. Master thermal management, power systems, and incident response.</p>
+    </div>
+
+    <!-- ── Getting Started ── -->
+    <div class="guide-section">
+      <div class="guide-section-header">
+        <div class="guide-section-icon blue">▶</div>
+        <h2>Getting Started</h2>
+      </div>
+      <ol>
+        <li><strong>Select a scenario</strong> from the sidebar — each presents a unique datacenter challenge.</li>
+        <li><strong>Choose a facility config</strong> (Default 160 kW, Small 80 kW, or Large 600 kW).</li>
+        <li>Click <strong>Start</strong> to begin the episode. You'll see the NOC dashboard.</li>
+        <li><strong>Issue commands</strong> in the command bar — diagnose equipment, adjust setpoints, manage power.</li>
+        <li>Each command advances simulation time. You have a limited <strong>step budget</strong>.</li>
+        <li>Maximize your <strong>cumulative reward</strong> by resolving the scenario efficiently.</li>
+      </ol>
+      <p style="margin-top:0.75rem"><strong>Pro tip:</strong> Always <code style="color:var(--cyan);background:var(--terminal-bg);padding:0.1rem 0.35rem;border-radius:3px;font-family:var(--font-mono);font-size:0.8rem">diagnose</code> before making changes — the reward system gives a bonus for proper diagnostic procedures and penalizes blind interventions.</p>
+    </div>
+
+    <!-- ── Scenarios ── -->
+    <div class="guide-section">
+      <div class="guide-section-header">
+        <div class="guide-section-icon orange">⚡</div>
+        <h2>Scenarios</h2>
+      </div>
+      <p>Six operational scenarios across two categories and three difficulty levels:</p>
+
+      <h3 style="font-size:0.85rem;font-weight:700;margin:1.25rem 0 0.625rem;color:var(--text)">Thermal (Category A)</h3>
+      <div class="guide-cards">
+        <div class="guide-card">
+          <div class="gc-header">
+            <span class="gc-id">A1</span>
+            <span class="gc-diff" style="background:var(--green-dim);color:var(--green)">Easy</span>
+          </div>
+          <div class="gc-name">Cooling Setpoint Optimization</div>
+          <div class="gc-desc">CRACs are overcooling at 15°C — wasting energy. Optimize setpoints for efficiency while keeping all zones within ASHRAE recommended range (18–27°C).</div>
+          <div class="gc-hint"><strong>Strategy</strong>Raise setpoints to ~22°C. Monitor temps. Target PUE &lt; 1.6. Check that all zones stay in recommended range for 2+ steps.</div>
+        </div>
+        <div class="guide-card">
+          <div class="gc-header">
+            <span class="gc-id">A2</span>
+            <span class="gc-diff" style="background:var(--yellow-dim);color:var(--yellow)">Medium</span>
+          </div>
+          <div class="gc-name">Thermal Event Response</div>
+          <div class="gc-desc">CRAC-3 compressor failure. Zone B temps are rising. Diagnose the fault and redistribute cooling to stabilize all zones.</div>
+          <div class="gc-hint"><strong>Strategy</strong>Diagnose CRAC-3 first. Lower setpoints on remaining CRACs. Boost fan speeds. Keep all zones in recommended range for 2+ steps.</div>
+        </div>
+        <div class="guide-card">
+          <div class="gc-header">
+            <span class="gc-id">A4</span>
+            <span class="gc-diff" style="background:var(--red-dim);color:var(--red)">Hard</span>
+          </div>
+          <div class="gc-name">CRAC Failure Cascade</div>
+          <div class="gc-desc">CRAC-1 compressor failure and CRAC-3 fan failure simultaneously. A cascading thermal event threatens multiple zones.</div>
+          <div class="gc-hint"><strong>Strategy</strong>Diagnose both CRACs. Aggressively lower setpoints on CRAC-2/4. Max fan speeds. Consider load shedding on hot racks. Keep zones in allowable range.</div>
+        </div>
+      </div>
+
+      <h3 style="font-size:0.85rem;font-weight:700;margin:1.25rem 0 0.625rem;color:var(--text)">Power (Category B)</h3>
+      <div class="guide-cards">
+        <div class="guide-card">
+          <div class="gc-header">
+            <span class="gc-id">B1</span>
+            <span class="gc-diff" style="background:var(--yellow-dim);color:var(--yellow)">Medium</span>
+          </div>
+          <div class="gc-name">UPS Alarm Response</div>
+          <div class="gc-desc">UPS transferred to battery after a utility event (now restored). Diagnose the situation and acknowledge the alarm to resolve.</div>
+          <div class="gc-hint"><strong>Strategy</strong>Diagnose UPS-1 first. Verify utility is restored. Acknowledge the alarm. The UPS should return to normal operation.</div>
+        </div>
+        <div class="guide-card">
+          <div class="gc-header">
+            <span class="gc-id">B3</span>
+            <span class="gc-diff" style="background:var(--green-dim);color:var(--green)">Easy</span>
+          </div>
+          <div class="gc-name">Generator Test Protocol</div>
+          <div class="gc-desc">Routine monthly generator test. Follow the proper 5-step protocol: diagnose → start → verify → stop → confirm shutdown.</div>
+          <div class="gc-hint"><strong>Strategy</strong>1. diagnose GEN-1 → 2. start_generator → 3. wait (let it warm) → 4. diagnose GEN-1 (verify running) → 5. stop_generator</div>
+        </div>
+        <div class="guide-card">
+          <div class="gc-header">
+            <span class="gc-id">B4</span>
+            <span class="gc-diff" style="background:var(--red-dim);color:var(--red)">Hard</span>
+          </div>
+          <div class="gc-name">Power Failure Cascade</div>
+          <div class="gc-desc">Utility power lost with extended generator warmup. UPS running on battery. Manage battery life and thermal conditions until generator loads.</div>
+          <div class="gc-hint"><strong>Strategy</strong>Start generator immediately. Shed non-critical rack loads to preserve battery. Monitor SOC. Once generator loads, restore loads. Keep temps stable.</div>
+        </div>
+      </div>
+    </div>
+
+    <!-- ── Available Commands ── -->
+    <div class="guide-section">
+      <div class="guide-section-header">
+        <div class="guide-section-icon cyan">⌨</div>
+        <h2>Available Commands</h2>
+      </div>
+      <table class="guide-table">
+        <thead>
+          <tr>
+            <th>Command</th>
+            <th>Description</th>
+            <th>Example</th>
+          </tr>
+        </thead>
+        <tbody>
+          <tr>
+            <td><code>diagnose &lt;unit&gt;</code></td>
+            <td>Inspect a CRAC, UPS, Generator, or PDU for faults and status</td>
+            <td><code>diagnose CRAC-3</code></td>
+          </tr>
+          <tr>
+            <td><code>adjust_setpoint &lt;crac&gt; &lt;°C&gt;</code></td>
+            <td>Change CRAC supply air setpoint (10–35°C). Supply temp converges over ~30s.</td>
+            <td><code>adjust_setpoint CRAC-1 22</code></td>
+          </tr>
+          <tr>
+            <td><code>set_fan_speed &lt;crac&gt; &lt;%&gt;</code></td>
+            <td>Set CRAC fan speed (0–100%). Fan power follows cubic law.</td>
+            <td><code>set_fan_speed CRAC-2 100</code></td>
+          </tr>
+          <tr>
+            <td><code>set_rack_load &lt;rack&gt; &lt;kW&gt;</code></td>
+            <td>Adjust rack IT load (0–30 kW) — simulates workload migration.</td>
+            <td><code>set_rack_load B-05 4</code></td>
+          </tr>
+          <tr>
+            <td><code>start_crac &lt;crac&gt;</code></td>
+            <td>Start a standby CRAC unit.</td>
+            <td><code>start_crac CRAC-3</code></td>
+          </tr>
+          <tr>
+            <td><code>stop_crac &lt;crac&gt;</code></td>
+            <td>Put a CRAC into standby mode.</td>
+            <td><code>stop_crac CRAC-4</code></td>
+          </tr>
+          <tr>
+            <td><code>start_generator</code></td>
+            <td>Initiate diesel generator start sequence (OFF → CRANKING → WARMING → READY → LOADED).</td>
+            <td><code>start_generator</code></td>
+          </tr>
+          <tr>
+            <td><code>stop_generator</code></td>
+            <td>Initiate generator cooldown sequence (300s).</td>
+            <td><code>stop_generator</code></td>
+          </tr>
+          <tr>
+            <td><code>set_ups_mode &lt;ups&gt; &lt;mode&gt;</code></td>
+            <td>Set UPS mode: <code>eco</code>, <code>double_conversion</code>, <code>line_interactive</code>, or <code>bypass</code>.</td>
+            <td><code>set_ups_mode UPS-1 eco</code></td>
+          </tr>
+          <tr>
+            <td><code>refuel_generator [liters]</code></td>
+            <td>Refuel the generator. Omit liters to fill tank.</td>
+            <td><code>refuel_generator 500</code></td>
+          </tr>
+          <tr>
+            <td><code>acknowledge_alarm</code></td>
+            <td>Acknowledge the current alert — clears the alert banner.</td>
+            <td><code>acknowledge_alarm</code></td>
+          </tr>
+          <tr>
+            <td><code>check_status</code></td>
+            <td>Request full status report. Refreshes the dashboard.</td>
+            <td><code>check_status</code></td>
+          </tr>
+          <tr>
+            <td><code>escalate</code></td>
+            <td>Escalate to senior engineer. Ends the episode.</td>
+            <td><code>escalate</code></td>
+          </tr>
+          <tr>
+            <td><code>wait</code></td>
+            <td>Take no action — advances simulation time by one step.</td>
+            <td><code>wait</code></td>
+          </tr>
+        </tbody>
+      </table>
+    </div>
+
+    <!-- ── Reward System ── -->
+    <div class="guide-section">
+      <div class="guide-section-header">
+        <div class="guide-section-icon green">★</div>
+        <h2>Reward System</h2>
+      </div>
+      <p>The environment uses a <strong>6-component, research-informed</strong> reward function. Each component is bounded to [−1, 1]. The total reward is a weighted sum, clamped to [−1, 1]. Weights auto-adjust based on scenario type.</p>
+
+      <div class="reward-grid">
+        <div class="reward-card">
+          <h4>🌡️ Thermal Safety <span class="rc-range">[−1, +0.1]</span></h4>
+          <p>Dual softplus barriers at ASHRAE recommended and allowable limits. Violations are penalized smoothly — the closer to the limit, the stronger the gradient. Returns <strong>+0.1 baseline</strong> when all zones are ≥3°C below recommended max (DCRL-Green).</p>
+          <div class="rc-formula">penalty = softplus((T − T_rec) / 2.0) + 3.0 · softplus((T − T_allow) / 1.5)</div>
+        </div>
+        <div class="reward-card">
+          <h4>⚡ Power Safety <span class="rc-range">[−1, 0]</span></h4>
+          <p>Penalizes low UPS battery state-of-charge (SOC) via softplus barrier at 50% threshold. UPS fault adds a fixed penalty of 5.0. Compounds across multiple UPS units.</p>
+          <div class="rc-formula">penalty = softplus((0.5 − SOC) / 0.15) + 5.0 · [fault]</div>
+        </div>
+        <div class="reward-card">
+          <h4>📊 Efficiency <span class="rc-range">[−1, 0]</span></h4>
+          <p>PUE-based energy efficiency. PUE 1.0 (ideal) → 0, PUE 2.0 → −0.46, PUE 3.0 → −0.76. <strong>Suppressed to 0</strong> during power emergencies (UPS on battery or fault) so the agent isn't penalized for correct load shedding.</p>
+          <div class="rc-formula">reward = −tanh((PUE − 1.0) / 2.0)</div>
+        </div>
+        <div class="reward-card">
+          <h4>🎯 Scenario Progress <span class="rc-range">[−1, +1]</span></h4>
+          <p>Delta-based: rewards the <em>change</em> in progress. This provides credit assignment — only the action that actually caused forward progress gets rewarded. Each scenario defines a normalized [0, 1] progress metric.</p>
+          <div class="rc-formula">reward = progress_now − progress_prev</div>
+        </div>
+        <div class="reward-card">
+          <h4>📋 Procedure <span class="rc-range">[−1, +1]</span></h4>
+          <p>Scenario-defined procedural correctness rules. For example, diagnosing before adjusting setpoints earns a bonus (+0.2), while skipping diagnosis incurs a penalty (−0.1). Encourages proper operational procedures.</p>
+          <div class="rc-formula">reward = scenario.procedure_reward (clamped)</div>
+        </div>
+        <div class="reward-card">
+          <h4>🎮 Action Quality <span class="rc-range">[−1, +1]</span></h4>
+          <p>Context-aware assessment: <strong>−0.5</strong> invalid command, <strong>−0.2</strong> repeat (except <code style="font-size:0.7rem">wait</code>/<code style="font-size:0.7rem">check_status</code>), <strong>+0.3</strong> diagnose/check_status, <strong>+0.2</strong> interventions, <strong>+0.1</strong> acknowledge, <strong>−0.1</strong> escalate. Waiting during generator startup: +0.1.</p>
+          <div class="rc-formula">Heuristic scoring per action type + context</div>
+        </div>
+      </div>
+
+      <h3 style="font-size:0.85rem;font-weight:700;margin:1.5rem 0 0.75rem;color:var(--text)">Weight Profiles</h3>
+      <p>Weights auto-select based on scenario type. Components sum to 1.0.</p>
+      <div class="weight-profiles" id="weightProfiles"></div>
+    </div>
+
+    <!-- ── ASHRAE Guidelines ── -->
+    <div class="guide-section">
+      <div class="guide-section-header">
+        <div class="guide-section-icon yellow">🏛</div>
+        <h2>ASHRAE Thermal Guidelines</h2>
+      </div>
+      <p>All safety thresholds follow <strong>ASHRAE TC 9.9, 5th Edition (2021)</strong>. The <span style="color:var(--green)">recommended</span> range is optimal for equipment longevity. The <span style="color:var(--yellow)">allowable</span> range permits short-term operation during incidents.</p>
+      <table class="guide-table">
+        <thead>
+          <tr>
+            <th>Class</th>
+            <th>Recommended</th>
+            <th>Allowable</th>
+            <th>Application</th>
+          </tr>
+        </thead>
+        <tbody>
+          <tr>
+            <td><strong>A1</strong></td>
+            <td><span style="color:var(--green)">18–27°C</span></td>
+            <td><span style="color:var(--yellow)">15–32°C</span></td>
+            <td>Enterprise servers</td>
+          </tr>
+          <tr>
+            <td><strong>A2</strong></td>
+            <td><span style="color:var(--green)">18–27°C</span></td>
+            <td><span style="color:var(--yellow)">10–35°C</span></td>
+            <td>Volume servers (most common)</td>
+          </tr>
+          <tr>
+            <td><strong>A3</strong></td>
+            <td><span style="color:var(--green)">18–27°C</span></td>
+            <td><span style="color:var(--yellow)">5–40°C</span></td>
+            <td>Extended temperature range</td>
+          </tr>
+          <tr>
+            <td><strong>A4</strong></td>
+            <td><span style="color:var(--green)">18–27°C</span></td>
+            <td><span style="color:var(--yellow)">5–45°C</span></td>
+            <td>Maximum flexibility</td>
+          </tr>
+          <tr>
+            <td><strong>H1</strong></td>
+            <td><span style="color:var(--green)">18–22°C</span></td>
+            <td><span style="color:var(--yellow)">5–25°C</span></td>
+            <td>High-density / AI / HPC (GPU servers)</td>
+          </tr>
+        </tbody>
+      </table>
+      <p style="margin-top:0.75rem;font-size:0.78rem;color:var(--text-dim)"><strong>Key insight:</strong> The reward system uses softplus barriers at both recommended and allowable limits. Staying ≥3°C below recommended max yields a +0.1 thermal safety bonus. Exceeding allowable limits incurs 3× the per-degree penalty of recommended violations.</p>
+    </div>
+
+    <!-- ── Physics Engine ── -->
+    <div class="guide-section">
+      <div class="guide-section-header">
+        <div class="guide-section-icon red">⚙</div>
+        <h2>Physics Engine</h2>
+      </div>
+
+      <h3 style="font-size:0.85rem;font-weight:700;margin:0.75rem 0 0.5rem;color:var(--text)">Thermal Model — RC Network</h3>
+      <p>The simulation uses a <strong>lumped-capacitance RC thermal network</strong> — the standard approach for datacenter transient thermal analysis. Each zone's temperature evolves according to:</p>
+      <div class="guide-code">C_total · dT/dt = Q_IT − Q_cooling + Q_envelope + Q_internal
+
+Where:
+  C_total    = C_air + C_equipment (dominated by server thermal mass)
+  Q_IT       = Σ rack IT loads [W] — all electrical power converts to heat
+  Q_cooling  = Σ CRAC outputs [W] — capacity varies with return air temp
+  Q_envelope = (T_outside − T_zone) / R_envelope [W]</div>
+
+      <p>Important CRAC characteristics:</p>
+      <ul>
+        <li><strong>Capacity vs. return temp:</strong> Q_actual = Q_rated × [1 + 0.03 × (T_return − T_rated)], so capacity increases when a zone heats up</li>
+        <li><strong>Fan power:</strong> Cubic law (affinity laws) — P_fan = P_rated × (speed%)³</li>
+        <li><strong>Supply temp lag:</strong> 30-second time constant between setpoint change and actual supply temp</li>
+        <li><strong>Recirculation:</strong> Hot air mixing caused by dominant airflow imbalance</li>
+      </ul>
+
+      <h3 style="font-size:0.85rem;font-weight:700;margin:1.25rem 0 0.5rem;color:var(--text)">Power Model</h3>
+      <p><strong>UPS quadratic loss model</strong> (APC White Paper 108):</p>
+      <div class="guide-code">η(x) = x / (x + 0.013 + 0.006x + 0.011x²)
+
+  90.5% efficient at 25% load
+  93.6% efficient at 50% load
+  94.0% efficient at 75% load</div>
+      <p><strong>Battery discharge:</strong> SOC depletes based on load, UPS efficiency, and temperature derating.</p>
+
+      <h3 style="font-size:0.85rem;font-weight:700;margin:1.25rem 0 0.5rem;color:var(--text)">Generator State Machine</h3>
+      <div class="guide-code">OFF ─→ START_DELAY (4s) ─→ CRANKING (5s) ─→ WARMING (8s) ─→ READY ─→ LOADED
+                                                                                         ↓
+                                                                                   COOLDOWN (300s) ─→ OFF</div>
+      <p>ATS (Automatic Transfer Switch) performs mechanical transfer in 100ms. Retransfer delay is 300 seconds to prevent rapid switching.</p>
+    </div>
+
+    <!-- ── Research References ── -->
+    <div class="guide-section">
+      <div class="guide-section-header">
+        <div class="guide-section-icon blue">📚</div>
+        <h2>Research Foundation</h2>
+      </div>
+      <ul>
+        <li><strong>Google/DeepMind (2017):</strong> Demonstrated 40% cooling energy reduction using RL with softplus barrier functions for safety constraints.</li>
+        <li><strong>DCRL-Green (ICLR 2025):</strong> Multi-objective reward with softplus barriers and positive safe-state baseline for safe RL in datacenters.</li>
+        <li><strong>ASHRAE TC 9.9, 5th Edition (2021):</strong> Industry-standard thermal guidelines used for all safety thresholds.</li>
+        <li><strong>APC White Paper 108:</strong> UPS quadratic loss model with experimentally calibrated coefficients.</li>
+        <li><strong>Process Reward Models:</strong> Delta-based progress rewards for improved credit assignment in multi-step reasoning.</li>
+      </ul>
+    </div>
+
+  </div>
+  </div><!-- /tabGuide -->
 </div>
 
 <script>
@@ -449,6 +891,39 @@ const SCENARIOS = {
   B4: { name: 'Power Failure Cascade', type: 'power', diff: 'Hard' },
 };
 
+// Scenario-adaptive quick action definitions
+const QUICK_ACTIONS = {
+  _common: ['check_status', 'wait', 'acknowledge_alarm', 'escalate'],
+  A1: ['adjust_setpoint CRAC-1 22', 'adjust_setpoint CRAC-2 22', 'adjust_setpoint CRAC-3 22', 'adjust_setpoint CRAC-4 22', 'diagnose CRAC-1'],
+  A2: ['diagnose CRAC-3', 'diagnose CRAC-1', 'adjust_setpoint CRAC-1 20', 'adjust_setpoint CRAC-2 20', 'set_fan_speed CRAC-1 100', 'set_fan_speed CRAC-2 100'],
+  A4: ['diagnose CRAC-1', 'diagnose CRAC-3', 'adjust_setpoint CRAC-2 16', 'adjust_setpoint CRAC-4 16', 'set_fan_speed CRAC-2 100', 'set_fan_speed CRAC-4 100', 'set_rack_load B-05 4'],
+  B1: ['diagnose UPS-1', 'diagnose GEN-1', 'start_generator', 'stop_generator'],
+  B3: ['start_generator', 'diagnose GEN-1', 'stop_generator'],
+  B4: ['diagnose UPS-1', 'diagnose GEN-1', 'start_generator', 'set_rack_load A-05 4', 'set_rack_load B-05 4'],
+};
+
+function buildQuickActions(scenarioId) {
+  const container = document.getElementById('quickActions');
+  container.innerHTML = '';
+  const specific = QUICK_ACTIONS[scenarioId] || [];
+  const common = QUICK_ACTIONS._common;
+  const all = [...specific, ...common];
+  for (const cmd of all) {
+    const btn = document.createElement('button');
+    btn.className = 'quick-btn';
+    btn.disabled = !episodeActive;
+    // Short display label
+    let label = cmd;
+    if (cmd === 'acknowledge_alarm') label = 'ack_alarm';
+    else if (cmd === 'check_status') label = 'check_status';
+    else if (cmd === 'start_generator') label = 'start_gen';
+    else if (cmd === 'stop_generator') label = 'stop_gen';
+    btn.textContent = label;
+    btn.onclick = () => quickCmd(cmd);
+    container.appendChild(btn);
+  }
+}
+
 // ─── WebSocket connection ────────────────────────────────────────────
 function connectWebSocket() {
   return new Promise((resolve, reject) => {
@@ -523,6 +998,7 @@ function selectScenario(id) {
   const btn = document.getElementById('startBtn');
   btn.disabled = false;
   btn.textContent = `Start ${id}: ${SCENARIOS[id].name}`;
+  buildQuickActions(id);
 }
 
 function togglePanel(id) {
@@ -536,6 +1012,8 @@ function setControlsEnabled(enabled) {
   document.getElementById('commandInput').disabled = !enabled;
   document.getElementById('sendBtn').disabled = !enabled;
   document.querySelectorAll('.quick-btn').forEach(b => b.disabled = !enabled);
+  // Rebuild quick actions if scenario changed while disabled
+  if (enabled && selectedScenario) buildQuickActions(selectedScenario);
 }
 
 function quickCmd(cmd) {
@@ -893,6 +1371,57 @@ function updateRewardHistory() {
   }
 }
 
+// ─── Tab switching ───────────────────────────────────────────────────
+function switchTab(tab) {
+  document.querySelectorAll('.tab-btn').forEach(b => b.classList.remove('active'));
+  document.querySelector(`.tab-btn[data-tab="${tab}"]`).classList.add('active');
+
+  // Console tab uses display:contents, guide uses display:block
+  const consoleEl = document.getElementById('tabConsole');
+  const guideEl = document.getElementById('tabGuide');
+  if (tab === 'console') {
+    consoleEl.classList.add('active');
+    guideEl.classList.remove('active');
+  } else {
+    consoleEl.classList.remove('active');
+    guideEl.classList.add('active');
+    guideEl.scrollTop = 0;
+  }
+}
+
+// ─── Build weight profile visualizations ─────────────────────────────
+function buildWeightProfiles() {
+  const profiles = {
+    'Thermal Scenarios': { thermal_safety: 0.30, power_safety: 0.05, efficiency: 0.10, progress: 0.30, procedure: 0.20, action: 0.05 },
+    'Power Scenarios':   { thermal_safety: 0.10, power_safety: 0.25, efficiency: 0.05, progress: 0.30, procedure: 0.25, action: 0.05 },
+    'Default':           { thermal_safety: 0.30, power_safety: 0.15, efficiency: 0.25, progress: 0.00, procedure: 0.00, action: 0.30 },
+  };
+  const colors = {
+    thermal_safety: 'var(--red)', power_safety: 'var(--yellow)', efficiency: 'var(--green)',
+    progress: 'var(--accent)', procedure: 'var(--cyan)', action: 'var(--orange)'
+  };
+  const labels = {
+    thermal_safety: 'Thermal', power_safety: 'Power', efficiency: 'Efficiency',
+    progress: 'Progress', procedure: 'Procedure', action: 'Action'
+  };
+  const container = document.getElementById('weightProfiles');
+  for (const [name, weights] of Object.entries(profiles)) {
+    const card = document.createElement('div');
+    card.className = 'weight-profile';
+    let html = `<h4>${name}</h4>`;
+    for (const [key, val] of Object.entries(weights)) {
+      const pct = val * 100;
+      html += `<div class="weight-bar-row">
+        <span class="weight-bar-label">${labels[key]}</span>
+        <div class="weight-bar-track"><div class="weight-bar-fill" style="width:${pct * 3.33}%;background:${colors[key]}"></div></div>
+        <span class="weight-bar-val">${pct.toFixed(0)}%</span>
+      </div>`;
+    }
+    card.innerHTML = html;
+    container.appendChild(card);
+  }
+}
+
 // ─── Health check ────────────────────────────────────────────────────
 async function checkHealth() {
   try {
@@ -906,6 +1435,7 @@ async function checkHealth() {
 
 // ─── Init ────────────────────────────────────────────────────────────
 checkHealth();
+buildWeightProfiles();
 </script>
 </body>
 </html>

simulation/thermal.py

@@ -485,16 +485,18 @@ class ThermalSimulation:
         self._state.outside_temp_c = temp_c
 
     def _find_crac(self, unit_id: str) -> CRACState | None:
+        target = unit_id.lower()
         for zone in self._state.zones:
             for crac in zone.crac_units:
-                if crac.unit_id == unit_id:
+                if crac.unit_id.lower() == target:
                     return crac
         return None
 
     def _find_rack(self, rack_id: str) -> RackState | None:
+        target = rack_id.lower()
         for zone in self._state.zones:
             for rack in zone.racks:
-                if rack.rack_id == rack_id:
+                if rack.rack_id.lower() == target:
                     return rack
         return None
 

tests/test_scenarios.py

@@ -180,16 +180,18 @@ class TestA2ThermalEvent:
 
     def test_procedure_bonus_for_diagnose_first(self) -> None:
         """Diagnosing before adjusting should yield higher reward."""
-        # Run 1: diagnose first, then adjust
+        # Run 1: diagnose first, then adjust (procedure bonus on step 2)
         env1 = DcOpsEnvironment()
         env1.reset(scenario="A2")
-        obs1a = env1.step(DcOpsAction(command="diagnose CRAC-3"))
+        env1.step(DcOpsAction(command="diagnose CRAC-3"))
         obs1b = env1.step(DcOpsAction(command="adjust_setpoint CRAC-4 20"))
         r_with_diagnose = obs1b.reward
 
-        # Run 2: adjust without diagnosing
+        # Run 2: wait, then adjust without diagnosing (procedure penalty on step 2)
+        # Using wait keeps physics comparable so only the procedure bonus differs
         env2 = DcOpsEnvironment()
         env2.reset(scenario="A2")
+        env2.step(DcOpsAction(command="wait"))
         obs2 = env2.step(DcOpsAction(command="adjust_setpoint CRAC-4 20"))
         r_without_diagnose = obs2.reward
 
@@ -294,9 +296,9 @@ class TestB3GeneratorTest:
 
         assert obs.done is True
 
-    def test_uses_30s_steps(self) -> None:
+    def test_uses_10s_steps(self) -> None:
         s = get_scenario("B3")
-        assert s.game_time_per_step_s == 30.0
+        assert s.game_time_per_step_s == 10.0
 
 
 # ===========================================================================