Security hardening, async TeachChat, rate limit UX, unthrottle

Security:
- admin.py: move api_key from URL query params to POST body (litellm/models, tts/voices) — prevents key logging
- admin.py: sanitize exception messages in voice discovery — log internally, return generic errors
- teach.py: log LLM errors server-side, show friendly message to user
- nextcloud.py: normalize path with posixpath.normpath to prevent ../ traversal
- auth.py: check_rate_limit now accepts user param — admins/moderators/unthrottled always exempt

Performance:
- teach.py: make /chat endpoint async, use litellm.acompletion() — no longer blocks a uvicorn thread per request

Features:
- users: add is_unthrottled column (DB migration in setup_pgvector)
- admin.py: PUT /users/{id}/unthrottle endpoint
- AdminPage.jsx: Unlimited/Throttle toggle button per user, shows "unlimited" badge
- Rate limit messages improved with user-friendly context

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/app/main.py

@@ -228,7 +228,8 @@ def setup_pgvector():
                 UNIQUE(user_id, question_id)
             )
         """))
-        conn.commit()
+        # Unthrottle flag for users (exempt from AI/TTS rate limits)
+        conn.execute(text("ALTER TABLE users ADD COLUMN IF NOT EXISTS is_unthrottled INTEGER DEFAULT 0"))
         conn.commit()
 
 

backend/app/models/user.py

@@ -14,6 +14,7 @@ class User(Base):
     hashed_password = Column(String, nullable=False)
     name = Column(String, nullable=False)
     role = Column(String, default="user")  # admin, moderator, user
+    is_unthrottled = Column(Integer, default=0)  # 1 = exempt from rate limits
     created_at = Column(DateTime, default=datetime.utcnow)
 
     documents = relationship("PDFDocument", back_populates="user")

backend/app/routers/admin.py

@@ -46,6 +46,23 @@ def update_user_role(
     return user
 
 
+@router.put("/users/{user_id}/unthrottle", response_model=UserResponse)
+def set_user_unthrottle(
+    user_id: int,
+    data: dict,
+    db: Session = Depends(get_db),
+    admin: User = Depends(require_admin),
+):
+    """Set or clear the unthrottle flag for a user — exempt from AI/TTS rate limits."""
+    user = db.query(User).filter(User.id == user_id).first()
+    if not user:
+        raise HTTPException(status_code=404, detail="User not found")
+    user.is_unthrottled = 1 if data.get("unthrottled") else 0
+    db.commit()
+    db.refresh(user)
+    return user
+
+
 @router.post("/users", response_model=UserResponse)
 def create_user(
     user_data: UserCreate,
@@ -99,34 +116,41 @@ def list_available_models(
     return result
 
 
-@router.get("/litellm/models")
+class LiteLLMSearchRequest(BaseModel):
+    api_key: str | None = None
+    api_base: str | None = None
+
+
+@router.post("/litellm/models")
 def search_litellm_models(
-    api_key: str | None = None,
-    api_base: str | None = None,
+    data: LiteLLMSearchRequest,
     admin: User = Depends(require_admin),
 ):
     """Query available models from LiteLLM proxy or OpenAI-compatible API."""
-    base = (api_base or settings.LITELLM_API_BASE or "").rstrip("/")
-    key = api_key or settings.LITELLM_API_KEY
+    import logging
+    log = logging.getLogger(__name__)
+    base = (data.api_base or settings.LITELLM_API_BASE or "").rstrip("/")
+    key = data.api_key or settings.LITELLM_API_KEY
 
     if base:
         try:
             headers = {"Authorization": f"Bearer {key}"} if key else {}
-            resp = httpx.get(f"{base}/v1/models", headers=headers, timeout=10)
+            resp = httpx.post(f"{base}/v1/models", headers=headers, timeout=10) if False else \
+                   httpx.get(f"{base}/v1/models", headers=headers, timeout=10)
             resp.raise_for_status()
-            data = resp.json()
-            models = sorted([m["id"] for m in data.get("data", [])])
+            models = sorted([m["id"] for m in resp.json().get("data", [])])
             return {"models": models, "source": base}
         except Exception as e:
-            raise HTTPException(status_code=400, detail=f"Failed to query models API: {e}")
+            log.warning(f"LiteLLM model search failed: {e}")
+            raise HTTPException(status_code=400, detail="Failed to query models API. Check the API base URL and key.")
 
-    # Fall back to LiteLLM's built-in model list
     try:
         import litellm
         models = sorted(litellm.utils.get_valid_models())
         return {"models": models, "source": "litellm-builtin"}
     except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Failed to get LiteLLM models: {e}")
+        log.warning(f"LiteLLM builtin model list failed: {e}")
+        raise HTTPException(status_code=500, detail="Failed to retrieve LiteLLM built-in model list.")
 
 
 @router.get("/models", response_model=list[AIModelConfigResponse])
@@ -238,14 +262,23 @@ def test_model(
         raise HTTPException(status_code=502, detail=str(e))
 
 
-@router.get("/tts/voices")
+class TTSVoiceSearchRequest(BaseModel):
+    provider: str
+    api_key: str | None = None
+    region: str | None = None
+
+
+@router.post("/tts/voices")
 def search_tts_voices(
-    provider: str,
-    api_key: str | None = None,
-    region: str | None = None,
+    data: TTSVoiceSearchRequest,
     admin: User = Depends(require_admin),
 ):
     """Discover available TTS voices from ElevenLabs, AWS Polly, or return OpenAI hardcoded list."""
+    import logging
+    log = logging.getLogger(__name__)
+    provider = data.provider
+    api_key = data.api_key
+    region = data.region
 
     if provider == "elevenlabs":
         key = api_key or settings.ELEVENLABS_API_KEY
@@ -270,7 +303,8 @@ def search_tts_voices(
         except HTTPException:
             raise
         except Exception as e:
-            raise HTTPException(status_code=400, detail=f"ElevenLabs API error: {e}")
+            log.warning(f"ElevenLabs voice discovery failed: {e}")
+            raise HTTPException(status_code=400, detail="ElevenLabs API error — check your API key and try again.")
 
     elif provider == "polly":
         access_key = api_key or settings.AWS_ACCESS_KEY_ID
@@ -302,7 +336,8 @@ def search_tts_voices(
         except HTTPException:
             raise
         except Exception as e:
-            raise HTTPException(status_code=400, detail=f"AWS Polly error: {e}")
+            log.warning(f"AWS Polly voice discovery failed: {e}")
+            raise HTTPException(status_code=400, detail="AWS Polly error — check your credentials and region.")
 
     elif provider == "openai":
         voices = []

backend/app/routers/nextcloud.py

@@ -31,8 +31,10 @@ class NCRequest(BaseModel):
 
 
 def _dav_url(server: str, username: str, path: str) -> str:
+    import posixpath
     base = server.rstrip("/")
-    p = path.lstrip("/")
+    # Normalize to collapse any ../ sequences before building URL
+    p = posixpath.normpath("/" + path).lstrip("/")
     return f"{base}/remote.php/dav/files/{username}/{p}"
 
 

backend/app/routers/teach.py

@@ -124,18 +124,19 @@ def list_teach_models(
 
 
 @router.post("/chat")
-def chat(
+async def chat(
     req: ChatRequest,
     db: Session = Depends(get_db),
     current_user: User = Depends(get_current_user),
 ):
     """Send a message to the teach AI with full question context."""
-    # Rate limit: 30 AI chat messages per user per 10 minutes
+    # Rate limit: 30 AI chat messages per user per 10 minutes (admins/unthrottled users exempt)
     check_rate_limit(
         key=f"teach_chat:{current_user.id}",
         max_calls=30,
         window_seconds=600,
-        detail="Too many AI chat messages. Please wait a few minutes before continuing.",
+        detail="You've sent too many messages to the AI tutor. Please wait a few minutes before continuing. If you need this limit raised, contact an admin.",
+        user=current_user,
     )
     model_info = _get_teach_model(db, req.model_id)
     if not model_info:
@@ -175,8 +176,10 @@ def chat(
             kwargs["api_key"] = settings.LITELLM_API_KEY
         if settings.LITELLM_API_BASE:
             kwargs["api_base"] = settings.LITELLM_API_BASE
-        response = litellm.completion(**kwargs)
+        response = await litellm.acompletion(**kwargs)
         reply = response.choices[0].message.content.strip()
         return {"reply": reply}
     except Exception as e:
-        raise HTTPException(status_code=502, detail=f"AI model error: {str(e)}")
+        import logging
+        logging.getLogger(__name__).error(f"TeachChat error for user {current_user.id} model {model_id}: {e}")
+        raise HTTPException(status_code=502, detail="The AI tutor is temporarily unavailable. Please try again in a moment.")

backend/app/routers/tts.py

@@ -51,12 +51,13 @@ def text_to_speech(
     current_user: User = Depends(get_current_user),
 ):
     """Convert text to speech using configured or user-selected TTS model."""
-    # Rate limit: 60 TTS requests per user per hour
+    # Rate limit: 60 TTS requests per user per hour (admins/unthrottled users exempt)
     check_rate_limit(
         key=f"tts_speak:{current_user.id}",
         max_calls=60,
         window_seconds=3600,
-        detail="TTS rate limit reached. You can generate up to 60 audio clips per hour.",
+        detail="You've reached the audio limit (60 clips/hour). The limit resets automatically — try again shortly. Contact an admin if you need this raised.",
+        user=current_user,
     )
     if not request.text.strip():
         raise HTTPException(status_code=400, detail="Text cannot be empty")

backend/app/schemas/auth.py

@@ -14,6 +14,7 @@ class UserResponse(BaseModel):
     email: str
     name: str
     role: str
+    is_unthrottled: int = 0
     created_at: datetime
 
     class Config:

backend/app/utils/auth.py

@@ -119,8 +119,13 @@ class TokenRefreshMiddleware(BaseHTTPMiddleware):
         return response
 
 
-def check_rate_limit(key: str, max_calls: int, window_seconds: int, detail: str):
-    """Generic Redis-backed rate limiter. Raises 429 if limit exceeded. Degrades gracefully if Redis is down."""
+def check_rate_limit(key: str, max_calls: int, window_seconds: int, detail: str, user=None):
+    """Generic Redis-backed rate limiter. Admins and unthrottled users are always exempt.
+    Degrades gracefully if Redis is down."""
+    # Admins, moderators, and explicitly unthrottled users bypass all rate limits
+    if user is not None:
+        if getattr(user, 'role', '') in ('admin', 'moderator') or getattr(user, 'is_unthrottled', 0):
+            return
     try:
         import redis as redis_lib
         r = redis_lib.from_url(settings.REDIS_URL, decode_responses=True, socket_connect_timeout=1)

frontend/src/pages/AdminPage.jsx

@@ -87,6 +87,16 @@ export default function AdminPage() {
     }
   }
 
+  const toggleUnthrottle = async (userId, currentVal) => {
+    try {
+      await api.put(`/admin/users/${userId}/unthrottle`, { unthrottled: !currentVal })
+      setSuccess(currentVal ? 'Rate limits restored for user' : 'Rate limits removed for user')
+      loadData()
+    } catch (err) {
+      setError(err.response?.data?.detail || 'Failed to update throttle setting')
+    }
+  }
+
   const createModel = async (e) => {
     e.preventDefault()
     setError('')
@@ -151,10 +161,10 @@ export default function AdminPage() {
     setSearchResults([])
     setSearchLoading(true)
     try {
-      const params = new URLSearchParams()
-      if (searchApiKey) params.set('api_key', searchApiKey)
-      if (searchApiBase) params.set('api_base', searchApiBase)
-      const res = await api.get(`/admin/litellm/models?${params}`)
+      const res = await api.post('/admin/litellm/models', {
+        api_key: searchApiKey || null,
+        api_base: searchApiBase || null,
+      })
       setSearchResults(res.data.models)
     } catch (err) {
       setSearchError(err.response?.data?.detail || 'Failed to query models')
@@ -168,7 +178,7 @@ export default function AdminPage() {
     setEmbedSearchResults([])
     setEmbedSearchLoading(true)
     try {
-      const res = await api.get('/admin/litellm/models')
+      const res = await api.post('/admin/litellm/models', {})
       const all = res.data.models || []
       // Filter to likely embedding models
       setEmbedSearchResults(all)
@@ -231,9 +241,10 @@ export default function AdminPage() {
     setTtsVoices([])
     setTtsVoicesLoading(true)
     try {
-      const params = new URLSearchParams({ provider: ttsProvider })
-      if (ttsSearchKey) params.set('api_key', ttsSearchKey)
-      const res = await api.get(`/admin/tts/voices?${params}`)
+      const res = await api.post('/admin/tts/voices', {
+        provider: ttsProvider,
+        api_key: ttsSearchKey || null,
+      })
       setTtsVoices(res.data.voices)
     } catch (err) {
       setTtsVoicesError(err.response?.data?.detail || 'Failed to fetch voices')
@@ -578,10 +589,15 @@ export default function AdminPage() {
                 <strong>{u.name}</strong>
                 <div style={{ fontSize: '0.85rem', color: '#64748b' }}>{u.email} · joined {new Date(u.created_at).toLocaleDateString()}</div>
               </div>
-              <div style={{ display: 'flex', gap: 6, alignItems: 'center' }}>
+              <div style={{ display: 'flex', gap: 6, alignItems: 'center', flexWrap: 'wrap' }}>
                 <span className={`badge ${u.role === 'admin' ? 'badge-error' : u.role === 'moderator' ? 'badge-processing' : 'badge-ready'}`}>
                   {u.role}
                 </span>
+                {u.is_unthrottled ? (
+                  <span style={{ fontSize: '0.72rem', padding: '2px 7px', borderRadius: 10, background: '#fef9c3', color: '#92400e', fontWeight: 600 }}>
+                    unlimited
+                  </span>
+                ) : null}
                 <select
                   value={u.role}
                   onChange={e => updateRole(u.id, e.target.value)}
@@ -591,6 +607,14 @@ export default function AdminPage() {
                   <option value="moderator">moderator</option>
                   <option value="admin">admin</option>
                 </select>
+                <button
+                  className="btn btn-secondary btn-sm"
+                  title={u.is_unthrottled ? 'Restore rate limits' : 'Remove rate limits for this user'}
+                  onClick={() => toggleUnthrottle(u.id, u.is_unthrottled)}
+                  style={{ fontSize: '0.75rem', padding: '3px 8px' }}
+                >
+                  {u.is_unthrottled ? '🔒 Throttle' : '🚀 Unlimited'}
+                </button>
               </div>
             </div>
           ))}