comments, profile info, user guid, gemini conservative/creative

README.md

@@ -74,3 +74,54 @@ Since `docker` is the selected SDK, the space builds the app using the `Dockerfi
 To update the code in the space, click on `+ Contribute` button in the upper-right corner of the Files page, then click `Upload Files`. Follow the instructions to commit/upload your local files to the space.  
 
 You could add the Git repo as a remote to your local Git repository, but it would add unnecessary complexity. HuggingFace is stricter than Gitlab concerning best Git practices. You would have to configure `git-xet` and delete the `.env` file and the binary file in `rag_data` from the Git history to be able to push your changes. The `.env` file has not been added to the space. The environment variables are stored in the settings page.    
+
+
+## Load testing
+[k6](https://k6.io/open-source/) is an open-source tool for performing load testing. Test cases are defined in JavaScript files and can be run using the command `k6 run <filename>.js`.
+
+### k6 installation
+On Debian/Ubuntu:
+```
+sudo apt-get update
+sudo apt-get install k6
+```
+On Windows:
+```
+winget install k6 --source winget
+```
+On Docker:
+```
+docker pull grafana/k6
+```
+For more options, see [Install k6](https://grafana.com/docs/k6/latest/set-up/install-k6/).
+
+### Test scenarios
+The test cases are defined in the folder `/tests/stress_tests/`:
+- `chat_session.js` simulates 80 users sending three messages to one specific model.
+- `website_spike.js` simulates 80 users connecting to the application home web page.
+
+
+#### Chat session test scenario
+The chat session scenario must be run by specifying the model type and the URL of the server. For example, the following command simulates 80 users making three requests at `https://<username>-champ-bot.hf.space/chat` to the model `champ`:
+```
+k6 run chat_session.js -e MODEL_TYPE=champ -e URL=https://<username>-champ-bot.hf.space/chat
+```
+The possible values for `MODEL_TYPE` are `champ`, `google`, and `openai`.  
+
+To find your HuggingFace Space backend URL, follow these steps:
+1. Go to your space
+2. Click on the **three dots** in the top right corner
+3. Select **Embed this Space**
+4. Look for the **Direct URL** in the code snippet.
+
+Typically, the URL follows this format: `https://<username>-<space-name>.hf.space`.  
+To test locally, simply use `http://localhost:8000/chat`  
+
+The file `message_examples.txt` contains 250 pediatric medical prompts (generated by Gemini). `chat_session.js` uses this file to simulate real user messages. 
+
+#### Website spike test scenario
+The website spike scenario must be run by specifying the website URL which is simply the HuggingFace Space URL:  
+```
+k6 run website_spike.js -e URL=https://huggingface.co/spaces/<username>/champ-bot
+```
+

champ/rag.py

@@ -1,4 +1,6 @@
 # app/champ/rag.py
+import torch
+
 from pathlib import Path
 
 from langchain_community.vectorstores import FAISS as LCFAISS
@@ -12,7 +14,7 @@ def load_vector_store(
     hf_token: str = HF_TOKEN,
     rag_relpath: str = "rag_data/FAISS_ALLEN_20260129",
     embedding_model: str = "BAAI/bge-large-en-v1.5",
-    device: str = "cpu",
+    device: str = "cuda" if torch.cuda.is_available() else "cpu",
 ) -> LCFAISS:
     rag_path = base_dir / rag_relpath
 

main.py

@@ -35,10 +35,12 @@ load_dotenv()
 # -------------------- Config --------------------
 DEV = os.getenv("ENV", None) == "dev"
 
+# The "Google" models are differentiated by their temperature.
 MODEL_MAP = {
     "champ": "champ-model/placeholder",
     "openai": "gpt-5-mini-2025-08-07",
-    "google": "gemini-2.5-flash-lite",
+    "google-conservative": "gemini-2.5-flash-lite",
+    "google-creative": "gemini-2.5-flash-lite",
 }
 
 OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
@@ -67,11 +69,24 @@ class ChatRequest(BaseModel):
     user_id: str
     session_id: str
     messages: List[ChatMessage]
-    temperature: float = 0.7
     model_type: str
-    # max_new_tokens: int = 256
     consent: bool = False
     system_prompt: Optional[str] = None
+    age_group: str
+    gender: str
+    roles: List[str]
+    participant_id: str
+
+
+class CommentRequest(BaseModel):
+    user_id: str
+    session_id: str
+    comment: str
+    consent: bool = False
+    age_group: str
+    gender: str
+    roles: List[str]
+    participant_id: str
 
 
 # -------------------- Helpers --------------------
@@ -103,9 +118,7 @@ print(f"CUDA available: {torch.cuda.is_available()}")
 champ = ChampService(base_dir=BASE_DIR, hf_token=HF_TOKEN)
 
 
-async def _call_openai(
-    model_id: str, msgs: list[dict], temperature: float
-) -> AsyncGenerator[str, None]:
+async def _call_openai(model_id: str, msgs: list[dict]) -> AsyncGenerator[str, None]:
     # We are streaming the output because the model answers tend to be very long and slow to generate
     stream = await openai_client.responses.create(
         model=model_id, input=msgs, stream=True
@@ -156,10 +169,13 @@ def call_llm(
     msgs = convert_messages(req.messages)
 
     if req.model_type == "openai":
-        return _call_openai(model_id, msgs, req.temperature)
+        return _call_openai(model_id, msgs)
 
-    if req.model_type == "google":
-        return _call_gemini(model_id, msgs, req.temperature), {}
+    if req.model_type == "google-conservative":
+        return _call_gemini(model_id, msgs, temperature=0.2), {}
+
+    if req.model_type == "google-creative":
+        return _call_gemini(model_id, msgs, temperature=1.0), {}
 
     # If you later add HF models via hf_client, handle here.
     raise ValueError(f"Unhandled model_type: {req.model_type}")
@@ -215,9 +231,12 @@ async def chat_endpoint(payload: ChatRequest, background_tasks: BackgroundTasks)
                     data={
                         "model_type": payload.model_type,
                         "consent": payload.consent,
-                        "temperature": payload.temperature,
                         "messages": payload.messages[-1].dict(),
                         "reply": reply,
+                        "age_group": payload.age_group,
+                        "gender": payload.gender,
+                        "roles": payload.roles,
+                        "participant_id": payload.participant_id,
                         "triage_meta": {},
                     },
                 )
@@ -235,8 +254,11 @@ async def chat_endpoint(payload: ChatRequest, background_tasks: BackgroundTasks)
                 "error": str(e),
                 "model_type": payload.model_type,
                 "consent": payload.consent,
-                "temperature": payload.temperature,
                 "messages": payload.messages[-1].dict(),
+                "age_group": payload.age_group,
+                "gender": payload.gender,
+                "roles": payload.roles,
+                "participant_id": payload.participant_id,
             },
         )
 
@@ -247,10 +269,33 @@ async def chat_endpoint(payload: ChatRequest, background_tasks: BackgroundTasks)
         data={
             "model_type": payload.model_type,
             "consent": payload.consent,
-            "temperature": payload.temperature,
             "messages": payload.messages[-1].dict(),
             "reply": reply,
+            "age_group": payload.age_group,
+            "gender": payload.gender,
+            "roles": payload.roles,
+            "participant_id": payload.participant_id,
             **(triage_meta or {}),
         },
     )
     return {"reply": reply}
+
+
+@app.post("/comment")
+def comment_endpoint(payload: CommentRequest, background_tasks: BackgroundTasks):
+    if not payload.comment:
+        return JSONResponse({"error": "No comment provided"}, status_code=400)
+
+    background_tasks.add_task(
+        log_event,
+        user_id=payload.user_id,
+        session_id=payload.session_id,
+        data={
+            "consent": payload.consent,
+            "comment": payload.comment,
+            "age_group": payload.age_group,
+            "gender": payload.gender,
+            "roles": payload.roles,
+            "participant_id": payload.participant_id,
+        },
+    )

static/app.js

@@ -2,27 +2,46 @@ const chatWindow = document.getElementById('chatWindow');
 const userInput = document.getElementById('userInput');
 const sendBtn = document.getElementById('sendBtn');
 const statusEl = document.getElementById('status');
+const statusComment = document.getElementById('commentStatus');
 
 const systemPresetSelect = document.getElementById('systemPreset');
-const tempSlider = document.getElementById('tempSlider');
-const tempValue = document.getElementById('tempValue');
-// const maxTokensSlider = document.getElementById("maxTokensSlider");
-// const maxTokensValue = document.getElementById("maxTokensValue");
 const clearBtn = document.getElementById('clearBtn');
 
-const consentOverlay = document.getElementById('consentOverlay');
+const welcomePopup = document.getElementById('welcomePopup');
+
 const consentCheckbox = document.getElementById('consentCheckbox');
 const consentBtn = document.getElementById('consentBtn');
 
+const profileBtn = document.getElementById('profileBtn');
+const genderInputs = document.querySelectorAll('input[name="gender"]');
+const roleInputs = document.querySelectorAll('input[name="role"]');
+const participantInput = document.getElementById('participant-id');
+
+const popupSlider = document.getElementById('mainSlider');
+
+const leaveCommentText = document.getElementById('leave-comment');
+const commentOverlay = document.getElementById('comment-overlay');
+
+const closeCommentBtn = document.getElementById('closeCommentBtn');
+const cancelCommentBtn = document.getElementById('cancelCommentBtn');
+const sendCommentBtn = document.getElementById('sendCommentBtn');
+const commentInput = document.getElementById('commentInput');
+
 // Local in-browser chat history
 // We store for each model its chat history.
-// We store the temperature of the google model as it can change.
 const modelChats = {};
 modelChats["champ"] = {"messages": []}
 modelChats["openai"] = {"messages": []}
-modelChats["google"] = {"messages": [], "temperature": 0.2}
+modelChats["google-conservative"] = {"messages": []}
+modelChats["google-creative"] = {"messages": []}
 
 let consentGranted = false;
+
+let ageGroup = '';
+let gender = '';
+let roles = [];
+let participantId = '';
+
 let sessionId = 'session-' + crypto.randomUUID(); // Unique session ID, generated once per page load
 document.body.classList.add('no-scroll');
 
@@ -42,11 +61,6 @@ function renderMessages() {
   chatWindow.scrollTop = chatWindow.scrollHeight;
 }
 
-function updateSlidersUI() {
-  tempValue.textContent = tempSlider.value;
-  // maxTokensValue.textContent = maxTokensSlider.value;
-}
-
 function getMachineId() {
   let machineId = localStorage.getItem('MachineId');
 
@@ -73,18 +87,16 @@ async function sendMessage() {
   statusEl.textContent = 'Thinking...';
   statusEl.className = 'status status-info';
 
-  const temperature = parseFloat(tempSlider.value);
-  // const maxTokens = parseInt(maxTokensSlider.value, 10);
-  // const systemPrompt = systemPresetSelect.value;
-
   const payload = {
     user_id: getMachineId(),
     session_id: sessionId,
     messages:  modelChats[modelType]["messages"].map((m) => ({ role: m.role, content: m.content })),
-    temperature,
-    // max_new_tokens: maxTokens,
     model_type: modelType,
     consent: consentGranted,
+    age_group: ageGroup,
+    gender,
+    roles,
+    participant_id: participantId
   };
 
   try {
@@ -131,7 +143,6 @@ async function sendMessage() {
     statusEl.textContent = 'Ready';
     statusEl.className = 'status status-ok';
   } catch (err) {
-    console.error(err);
     statusEl.textContent = 'Network error.';
     statusEl.className = 'status status-error';
   }
@@ -146,12 +157,6 @@ function clearConversation() {
 
   resetSession();
   modelChats[modelType]["messages"] = [];
-  // If the model is google, we also have to clear the temperature
-  if (modelType === "google") {
-    modelChats["google"]["temperature"] = 0.2;
-    tempSlider.value = 0.2
-    updateSlidersUI();
-  }
   renderMessages();
   statusEl.textContent = 'Conversation cleared. Start a new chat!';
   statusEl.className = 'status status-ok';
@@ -168,9 +173,41 @@ consentCheckbox.addEventListener('change', () => {
 
 // Handle the consent acceptance
 consentBtn.addEventListener('click', () => {
-  consentOverlay.style.display = 'none'; // Hide overlay
-  document.body.classList.remove('no-scroll'); // NEW: re-enable scrolling
   consentGranted = true; // Mark consent as granted
+  popupSlider.style.transform = `translateX(-468px)`;
+});
+
+// When the profile is changed, enable or disable the button
+function checkProfileValidity () {
+  // 1. Check if any gender is selected
+  const genderSelected = Array.from(genderInputs).some(input => input.checked);
+
+  // 2. Check if at least one role checkbox is selected
+  const roleSelected = Array.from(roleInputs).some(input => input.checked);
+
+  // 3. Check if the participant id field has a value
+  const participantIdEntered = participantInput.value.trim().length > 0;
+
+  // 4. Enable button only if both are true
+  if (genderSelected && roleSelected && participantIdEntered) {
+      profileBtn.disabled = false;
+  } else {
+      profileBtn.disabled = true;
+  }
+}
+// Add the listener to all gender radio buttons and role checkboxes
+genderInputs.forEach(input => input.addEventListener('change', checkProfileValidity));
+roleInputs.forEach(input => input.addEventListener('change', checkProfileValidity));
+participantInput.addEventListener('input', checkProfileValidity);
+
+profileBtn.addEventListener('click', () => {
+  welcomePopup.style.display = 'none'; // Hide overlay
+  document.body.classList.remove('no-scroll'); // NEW: re-enable scrolling
+
+  ageGroup = document.getElementById('age-group').value;
+  gender = document.querySelector('input[name="gender"]:checked')?.value;
+  roles = Array.from(document.querySelectorAll('input[name="role"]:checked')).map(input => input.value);
+  participantId = participantInput.value.trim();
 });
 
 sendBtn.addEventListener('click', sendMessage);
@@ -182,47 +219,84 @@ userInput.addEventListener('keydown', (e) => {
     sendMessage();
   }
 });
-
-tempSlider.addEventListener('input', () => {
-  if (!tempSlider.disabled) {
-    updateSlidersUI();
-    const modelType = systemPresetSelect.value;
-    modelChats[modelType]["temperature"] = tempSlider.value;
+commentInput.addEventListener('keydown', (e) => {
+  if (e.key === 'Enter' && !e.shiftKey) {
+    e.preventDefault();
+    sendComment();
   }
 });
-// maxTokensSlider.addEventListener("input", updateSlidersUI);
+
+
 clearBtn.addEventListener('click', clearConversation);
 
 systemPresetSelect.addEventListener('change', () => {
-  updateTempControlForModel();   // 👈 add this
   renderMessages();
   statusEl.textContent = 'Model changed.';
   statusEl.className = 'status status-ok';
 });
 
-// initial UI state
-updateTempControlForModel();
-function updateTempControlForModel() {
-  const model = systemPresetSelect.value;
-
-  if (model === 'champ') {
-    // Fix CHAMP temperature and disable slider
-    tempSlider.disabled = true;
-    tempSlider.value = '0.2';
-    tempValue.textContent = '0.2 (fixed)';
-    tempSlider.classList.add('disabled');
-  } else if (model === 'openai' ){
-    // GPT-5 models: temperature not supported
-    tempSlider.disabled = true;
-    tempValue.textContent = 'N/A (not supported for GPT-5 models)';;
-    tempSlider.classList.add('disabled');
-  } else {
-    // Enable slider for other models
-    tempSlider.disabled = false;
-    tempSlider.classList.remove('disabled');
-    tempSlider.value = modelChats[model]["temperature"];
-    updateSlidersUI(); // refresh displayed value
-  }
+// Comments
+function openCommentOverlay(e) {
+  e.preventDefault();
+  // Let the stylesheet take over
+  commentOverlay.style.display = '';
 }
+leaveCommentText.addEventListener('click', openCommentOverlay);
+
+// Cancelling or closing the comment overlay simply hides the comment popup
+closeCommentBtn.addEventListener('click', () => {
+  commentOverlay.style.display = 'none';
+});
+cancelCommentBtn.addEventListener('click', () => {
+  commentOverlay.style.display = 'none';
+});
+
+async function sendComment() {
+  const comment = commentInput.value;
+  if (!comment) return;
+
+  const payload = {
+    user_id: getMachineId(),
+    session_id: sessionId,
+    comment,
+    consent: consentGranted,
+    age_group: ageGroup,
+    gender,
+    roles,
+    participant_id: participantId
+  };
+
+  statusComment.textContent = 'Sending...';
+  statusComment.className = 'status-info';
+
+  try {
+    const res = await fetch('/comment', {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify(payload),
+    });
+
+    if (!res.ok) {
+      statusComment.textContent = 'Error from the server';
+      statusComment.className = 'status-error';
+      return;
+    }
+
+    commentInput.value = '';
+
+    statusComment.textContent = 'Comment sent!'
+    statusComment.className = 'status-ok'
+  } catch (err) {
+    statusComment.textContent = 'Network error';
+    statusComment.className = 'status-error'
+  }
+
+};
+sendCommentBtn.addEventListener('click', sendComment);
+
 statusEl.textContent = 'Ready';
-statusEl.className = 'status status-ok';
+statusEl.className = 'status-ok';
+
+statusComment.textContent = 'Ready';
+statusComment.className = 'status-ok';
+

static/style.css

@@ -12,6 +12,10 @@ body.no-scroll {
   overflow: hidden;
 }
 
+a {
+  color: #4da6ff;
+}
+
 .chat-container {
   max-width: 900px;
   margin: 40px auto;
@@ -151,10 +155,13 @@ body.no-scroll {
   background: #3453e6;
 }
 
-/* Status text */
-.status {
+/* Status and comment text */
+.status-comment {
   margin-top: 6px;
   font-size: 0.85rem;
+
+  display: flex;
+  justify-content: space-between;
 }
 
 .status-info {
@@ -192,15 +199,15 @@ body.no-scroll {
 }
 
 /* CONSENT OVERLAY FIXED VERSION */
-.consent-overlay {
+.popup-overlay {
   position: fixed;
   top: 0;
   left: 0;
   width: 100%;
   height: 100%;
 
-  background: rgba(0, 0, 0, 0.55); /* CHANGED: darker for visibility */
-  backdrop-filter: blur(4px);
+  background: rgba(0, 0, 0, 0.8); /* CHANGED: darker for visibility */
+  /* backdrop-filter: blur(4px); */ /* removed blur for performance */
 
   display: flex;
   align-items: center;
@@ -209,8 +216,24 @@ body.no-scroll {
   z-index: 9999; /* NEW: ensure nothing covers this */
 }
 
+.slider {
+  display: flex;
+  width: 840px;
+  transition: transform 0.5s cubic-bezier(0.25, 1, 0.5, 1);
+
+  /* Performance Boosters */
+  will-change: transform;
+  perspective: 1000px;
+}
+
+.popup-window {
+  width: 468px;
+  overflow: hidden;
+}
+
 /* Dark theme overlay box */
-.consent-box {
+.popup-step {
+  flex-shrink: 0;
   background: #141b2f; /* CHANGED: match theme */
   color: #f5f5f5; /* NEW: readable on dark bg */
   padding: 24px;
@@ -225,8 +248,129 @@ body.no-scroll {
   margin: 16px 0;
   gap: 10px;
 }
+
 /* Disable look for CHAMP fixed temperature */
 input[type='range'].disabled {
   opacity: 0.6;
   cursor: not-allowed;
+}
+
+/* Comment area */
+.comment-area {
+  position: relative;
+  display: flex;
+  flex-direction: column;
+  gap: 16px;
+  background: #1a2238;
+  padding: 24px;
+  border-radius: 15px;
+  width: 90%;
+  max-width: 450px;
+  border: 1px solid #2c3554;
+  box-shadow: 0 4px 12px rgba(0, 0, 0, 0.4);
+}
+
+.comment-area h2 {
+  margin: 0 0 8px 0;
+  font-size: 1.5rem;
+  color: #f5f5f5;
+  font-weight: 600;
+}
+
+.comment-area textarea {
+  width: 425px;
+  min-height: 120px;
+  border-radius: 10px;
+  border: 1px solid #2c3554;
+  background: #0d1324;
+  color: #f5f5f5;
+  padding: 12px;
+  resize: vertical;
+  font-size: 1rem;
+  font-family: inherit;
+  transition: border-color 0.2s ease;
+}
+
+.comment-area textarea:focus {
+  outline: none;
+  border-color: #4a5f8f;
+}
+
+.comment-area textarea::placeholder {
+  color: #6b7280;
+}
+
+/* Button container */
+.comment-area .button-group {
+  display: flex;
+  gap: 12px;
+  justify-content: flex-end;
+}
+
+/* Cancel button */
+.comment-area .cancelBtn {
+  background: transparent;
+  color: #9ca3af;
+  border: 1px solid #2c3554;
+  padding: 10px 24px;
+  border-radius: 8px;
+  font-size: 0.95rem;
+  font-weight: 500;
+  cursor: pointer;
+  transition: all 0.2s ease;
+}
+
+.comment-area .cancelBtn:hover {
+  background: #0d1324;
+  color: #f5f5f5;
+  border-color: #4a5f8f;
+}
+
+/* Send button */
+.comment-area .sendBtn {
+  background: #3b82f6;
+  color: #ffffff;
+  padding: 10px 24px;
+  border-radius: 8px;
+  border: none;
+  font-size: 0.95rem;
+  font-weight: 500;
+  cursor: pointer;
+  transition: all 0.2s ease;
+}
+
+.comment-area .sendBtn:hover {
+  background: #2563eb;
+}
+
+.comment-area .sendBtn:active {
+  transform: scale(0.98);
+}
+
+/* Close button (X) */
+.comment-area .closeBtn {
+  position: absolute;
+  top: 15px;
+  right: 15px;
+  width: 28px;        /* Explicit small width */
+  height: 28px;       /* Explicit small height */
+  padding: 0;
+  background: transparent;
+  color: #6b7280;
+  border: none;
+  border-radius: 4px;
+  font-size: 20px;
+  line-height: 28px;  /* Center the × vertically */
+  text-align: center; /* Center the × horizontally */
+  cursor: pointer;
+  transition: all 0.2s ease;
+}
+
+.comment-area .close-btn:hover {
+  background: #0d1324;
+  color: #f5f5f5;
+}
+
+.comment-area .close-btn:active {
+  transform: scale(0.95);
 }

templates/index.html

@@ -19,6 +19,11 @@
           remember to avoid sharing any sensitive or private details during the
           conversation.
         </p>
+        <p class="subtitle">User guide: 
+          <a href="https://docs.google.com/document/d/1-2UIpKbh1BdAmgCaF4QdcaZ4H5fwkQkKRigHz47EejY/edit?usp=sharing" target="_blank">
+            CHAMP Model Comparison – Participant Testing Guide
+          </a>
+        </p>
       </header>
 
       <!-- Controls bar -->
@@ -29,45 +34,79 @@
             <option value="champ" selected>CHAMP</option>
             <!-- champ is our model -->
             <option value="openai">GPT-5.2</option>
-            <option value="google">Gemini-3</option>
+            <option value="google-conservative">Gemini-3 (Conservative)</option>
+            <option value="google-creative">Gemini-3 (Creative)</option>
           </select>
         </div>
 
-        <div class="control-group">
-          <label for="tempSlider">
-            Temperature:
-            <span id="tempValue">0.7</span>
-          </label>
-          <input
-            type="range"
-            id="tempSlider"
-            min="0.0"
-            max="1.0"
-            step="0.1"
-            value="0.7"
-          />
-        </div>
-
         <button id="clearBtn" class="secondary-button">Clear</button>
       </div>
 
-      <div id="consentOverlay" class="consent-overlay">
-        <div class="consent-box">
-          <h2>Before you continue</h2>
-          <p>
-            By using this demo you agree that your messages will be shared with
-            us for processing. Do not provide sensitive or private details.
-          </p>
-
-          <label class="consent-check">
-            <input type="checkbox" id="consentCheckbox" />
-            I understand and agree
-          </label>
-
-          <button id="consentBtn" class="primary-button" disabled>
-            Agree and Continue
-          </button>
+      <div id="welcomePopup" class="popup-overlay">
+        <div class="popup-window">
+        <div class="slider" id="mainSlider">
+          <div class="consent-box popup-step">
+            <h2>Before you continue</h2>
+            <p>
+              By using this demo you agree that your messages will be shared with
+              us for processing. Do not provide sensitive or private details.
+            </p>
+
+            <label class="consent-check">
+              <input type="checkbox" id="consentCheckbox" />
+              I understand and agree
+            </label>
+
+            <button id="consentBtn" class="primary-button" disabled>
+              Agree and Continue
+            </button>
+          </div>
+
+          <div class="profile popup-step">
+            <h2>Profile information</h2>
+            <p>We collect this information to help us understand how different groups of users interact with the system.
+              This data allows us to improve the experience and ensure the tool is effective for everyone.</p>
+            <label for="age-group">Age group</label>
+            <select id="age-group">
+              <option value="0-18" selected>0-18</option>
+              <option value="18-24">18-24</option>
+              <option value="25-34">25-34</option>
+              <option value="35-44">35-44</option>
+              <option value="45-54">45-54</option>
+              <option value="55-64">55-64</option>
+              <option value="65+">65+</option>
+            </select>
+            
+            <p>Gender</p>
+            <input type="radio" id="male" name="gender" value="M">
+            <label for="male">Male</label>
+
+            <input type="radio" id="female" name="gender" value="F">
+            <label for="female">Female</label>
+
+            <p>Role</p>
+
+            <input type="checkbox" id="patient" name="role" value="patient">
+            <label for="patient">Patient</label><br>
+            <input type="checkbox" id="clinician" name="role" value="clinician">
+            <label for="clinician">Clinician</label><br>
+            <input type="checkbox" id="computer-scientist" name="role" value="computer-scientist">
+            <label for="computer-scientist">Computer scientist</label><br>
+            <input type="checkbox" id="researcher" name="role" value="researcher">
+            <label for="researcher">Researcher</label><br>
+            <input type="checkbox" id="other" name="role" value="other">
+            <label for="other">Other</label><br>
+            <br>
+            <label for="participant-id">Participant Id</label>
+            <input type="text" id="participant-id" name="participant-id"><br>
+            <br>
+            <button id="profileBtn" class="primary-button" disabled>
+              Continue
+            </button>
+          </div>
+        </div>
         </div>
+
       </div>
 
       <!-- Chat window -->
@@ -85,8 +124,27 @@
         <button id="sendBtn">Send</button>
       </footer>
 
-      <!-- Status line -->
-      <div id="status" class="status"></div>
+      <!-- Status/Comment line -->
+      <div class="status-comment">
+        <span id="status"></span>
+        <span id="leave-comment"><a href="#">Leave a comment</a></span>
+      </div>
+
+      <div id="comment-overlay" class="popup-overlay" style="display:none">
+        <div class="popup-step comment-area">
+          <button id="closeCommentBtn" class="closeBtn" aria-label="Close">×</button>
+          <h2>Leave a comment</h2>
+          <textarea
+            id="commentInput"
+            rows="2"
+            placeholder="Type your comment and click Send..."
+          ></textarea>
+          <div id="commentStatus" class="comment-status"></div>
+          <button id="cancelCommentBtn" class="cancelBtn">Cancel</button>
+          <button id="sendCommentBtn" class="sendBtn">Send</button>
+        </div>
+      </div>
+
     </div>
 
     <script src="/static/app.js"></script>

tests/stress_tests/chat_session.js

@@ -1,5 +1,4 @@
-// This file aims to test a case where 80 users send 3 messages to ONE model (very pessimistic).
-// This test aims to help us identify the most appropriate plan for each model.
+// This file aims to test a scenario where 80 users send 3 messages to ONE model (pessimistic).
 
 import http from 'k6/http';
 import { sleep, check } from 'k6';
@@ -16,8 +15,8 @@ export const options = {
   scenarios: {
     my_spike_test: {
       executor: 'per-vu-iterations',
-      vus: 80,               // 80 total users
-      iterations: 1,        // Each user runs the script exactly once
+      vus: 80,              // 80 total users
+      iterations: 1,        // Each user runs the function exactly once.
     },
   },
 };
@@ -27,8 +26,7 @@ export default function () {
   // sending their messages at the exact same time.
   sleep(Math.random() * 10);
 
-  // const url = 'https://marvin-cusm-chatbot-champ-chatbot.hf.space/chat'
-  const url = 'http://localhost:8000/chat'
+  const url = __ENV.URL
 
   const conversation = [];
   // Each VU sends 3 messages
@@ -48,24 +46,30 @@ export default function () {
     const params = { headers: { 'Content-Type': 'application/json' } };
 
     const res = http.post(url, JSON.stringify(payload), params);
+
     check(res, {'status is 200': (r) => r.status === 200})
 
     let reply = '';
     if (res.status === 200) {
       // k6 does not support streaming response bodies. It waits for the entire response until
       // the stream is 'done'. Therefore, we do not need to read the chunks one by one.
-      const data = res.json();
+      let data = "";
+      try {
+        data = res.json();
+      } catch (error) {
+        // However, if the response contains streamed data, it is not in JSON format.
+        // We would have to read the body to access that data.
+        data = res.body;
+      }
       reply = data.reply || '(No reply)';
       conversation.push({ role: 'assistant', content: reply });
     } else {
-      console.log(res.status);
-      console.log(res.body);
+      console.error(res.status);
+      console.error(res.body);
       conversation.push({ role: 'assistant', content: '(No reply)' });
     }
 
-    // Users usually wait before sending another message.
-    // It takes time to read the response and write another message.
-
+    // Simulating reading time, thinking time and writing time.
     // Simulate reading speed: ~200ms per word in the reply + 2s thinking time
     const readingTime = (reply.split(' ').length * 0.2) + 2;
     // Cap it so it doesn't wait forever, but add some randomness (jitter)
@@ -75,8 +79,31 @@ export default function () {
   }
 }
 
+// TEST RESULT ANALYSIS
+
+// CHAMP
+// The bottleneck associated with CHAMP's performance is the HuggingFace Space hardware.
+// CHAMP actually requires a lot of computing power, because of the FAISS retrieval system.
+// Using a T4 GPU for accelerated retrieval has halved the average request duration, from 11.65s
+// to 5.56s. 
+// The performance of CHAMP could be improved by selecting a more powerful GPU (such as L4 or A10G)
+// or by running a simpler vector search algorithm. The current algorithm (maximal marginal relevance)
+// optimizes for similarity to query and diversity among selected documents. A basic similarity search
+// should be faster to run.
+
+// Google (Gemini)
+// The average request duration is 1.13s with a maximum of 3.44s. This level of performance is acceptable.
 
-// CHAMP (GROK) - BASIC CPU - 80VUs (~$1.20)
+// OpenAI (GPT-5-mini)
+// Previous tests used GPT-5-nano. However, this model had strict rate limits and tests would exceed them
+// which caused them to fail. The model was switched to GPT-5-mini.
+// The performance is quite poor, with an average request duration of 22.73s and a maximum duration of 59.62s.
+// The bottleneck lies within the OpenAI API, which leaves us with little room for optimization.
+
+
+// RAW TEST RESULTS
+
+// CHAMP (GROK) - BASIC CPU - 80VUs
 //   █ TOTAL RESULTS
 
 //     checks_total.......: 240     2.408279/s
@@ -101,8 +128,95 @@ export default function () {
 //     data_received..................: 547 kB 5.5 kB/s
 //     data_sent......................: 345 kB 3.5 kB/s
 
+// running (01m39.7s), 00/80 VUs, 80 complete and 0 interrupted iterations
+// my_spike_test ✓ [======================================] 80 VUs  01m39.7s/10m0s  80/80 iters, 1 per VU
+
 
 
+// CHAMP (GROK) - T4 small - 80VUs
+//   █ TOTAL RESULTS
 
-// running (01m39.7s), 00/80 VUs, 80 complete and 0 interrupted iterations
-// my_spike_test ✓ [======================================] 80 VUs  01m39.7s/10m0s  80/80 iters, 1 per VU
+//     checks_total.......: 240     3.02291/s
+//     checks_succeeded...: 100.00% 240 out of 240
+//     checks_failed......: 0.00%   0 out of 240
+
+//     ✓ status is 200
+
+//     HTTP
+//     http_req_duration..............: avg=5.56s min=31.95ms med=5.74s max=9.78s p(90)=7.92s p(95)=8.92s
+//       { expected_response:true }...: avg=5.56s min=31.95ms med=5.74s max=9.78s p(90)=7.92s p(95)=8.92s
+//     http_req_failed................: 0.00%  0 out of 240
+//     http_reqs......................: 240    3.02291/s
+
+//     EXECUTION
+//     iteration_duration.............: avg=1m5s  min=35.54s  med=1m7s  max=1m19s p(90)=1m16s p(95)=1m17s
+//     iterations.....................: 80     1.007637/s
+//     vus............................: 1      min=1        max=80
+//     vus_max........................: 80     min=80       max=80
+
+//     NETWORK
+//     data_received..................: 539 kB 6.8 kB/s
+//     data_sent......................: 335 kB 4.2 kB/s
+
+// running (01m19.4s), 00/80 VUs, 80 complete and 0 interrupted iterations
+// my_spike_test ✓ [======================================] 80 VUs  01m19.4s/10m0s  80/80 iters, 1 per VU
+
+
+
+// GEMINI - T4 small - 80VUs
+//   █ TOTAL RESULTS
+
+//     checks_total.......: 240     3.611159/s
+//     checks_succeeded...: 100.00% 240 out of 240
+//     checks_failed......: 0.00%   0 out of 240
+
+//     ✓ status is 200
+
+//     HTTP
+//     http_req_duration..............: avg=1.13s  min=266.67ms med=991.5ms max=3.44s p(90)=2.08s p(95)=2.44s
+//       { expected_response:true }...: avg=1.13s  min=266.67ms med=991.5ms max=3.44s p(90)=2.08s p(95)=2.44s
+//     http_req_failed................: 0.00%  0 out of 240
+//     http_reqs......................: 240    3.611159/s
+
+//     EXECUTION
+//     iteration_duration.............: avg=48.54s min=15.63s   med=53.61s  max=1m6s  p(90)=1m1s  p(95)=1m3s
+//     iterations.....................: 80     1.20372/s
+//     vus............................: 1      min=1        max=80
+//     vus_max........................: 80     min=80       max=80
+
+//     NETWORK
+//     data_received..................: 621 kB 9.3 kB/s
+//     data_sent......................: 416 kB 6.3 kB/s
+
+// running (01m06.5s), 00/80 VUs, 80 complete and 0 interrupted iterdations
+// my_spike_test ✓ [======================================] 80 VUs  01m06.5s/10m0s  80/80 iters, 1 per VU
+
+
+
+// GPT-5-mini - T4 small - 80VUs
+//   █ TOTAL RESULTS
+
+//     checks_total.......: 240     2.010998/s
+//     checks_succeeded...: 100.00% 240 out of 240
+//     checks_failed......: 0.00%   0 out of 240
+
+//     ✓ status is 200
+
+//     HTTP
+//     http_req_duration..............: avg=22.73s min=7.66s  med=20.73s max=59.62s p(90)=32.1s p(95)=38.59s
+//       { expected_response:true }...: avg=22.73s min=7.66s  med=20.73s max=59.62s p(90)=32.1s p(95)=38.59s
+//     http_req_failed................: 0.00%  0 out of 240
+//     http_reqs......................: 240    2.010998/s
+
+//     EXECUTION
+//     iteration_duration.............: avg=1m20s  min=51.07s med=1m18s  max=1m59s  p(90)=1m35s p(95)=1m49s
+//     iterations.....................: 80     0.670333/s
+//     vus............................: 1      min=1        max=80
+//     vus_max........................: 80     min=80       max=80
+
+//     NETWORK
+//     data_received..................: 4.9 MB 41 kB/s
+//     data_sent......................: 247 kB 2.1 kB/s
+
+// running (01m59.3s), 00/80 VUs, 80 complete and 0 interrupted iterations
+// my_spike_test ✓ [======================================] 80 VUs  01m59.3s/10m0s  80/80 iters, 1 per VU

tests/stress_tests/website_spike.js

@@ -1,5 +1,4 @@
 // This file aims to test a case where 80 users connect to the website at the same time.
-// This test aims to help us identify the most appropriate HuggingFace hardware.
 import http from 'k6/http';
 import { check } from 'k6';
 
@@ -15,12 +14,11 @@ export const options = {
 };
 
 export default function () {
+  const url = __ENV.URL;
   // User lands on the site
-  const res = http.get('https://huggingface.co/spaces/marvin-cusm-chatbot/champ-chatbot');
+  const res = http.get(url);
   
   check(res, { 'status is 200': (r) => r.status === 200 });
 }
 
-// iteration_duration.............: avg=4.91s min=3.01s med=5.11s max=7.36s p(90)=5.37s p(95)=5.87s
-// It might take up to 7 seconds for a page to load
-// We might want to upgrade
+// http_req_duration..............: avg=1.28s min=628.16ms med=1.35s max=2.21s p(90)=1.7s  p(95)=1.77s