lemonade-sdk 8.1.6__py3-none-any.whl → 8.1.8__py3-none-any.whl

lemonade/cli.py

@@ -25,8 +25,6 @@ from lemonade.tools.humaneval import AccuracyHumaneval
 from lemonade.tools.perplexity import AccuracyPerplexity
 from lemonade.tools.accuracy import LMEvalHarness
 from lemonade.tools.prompt import LLMPrompt
-from lemonade.tools.quark.quark_load import QuarkLoad
-from lemonade.tools.quark.quark_quantize import QuarkQuantize
 from lemonade.tools.report.llm_report import LemonadeReport
 
 
@@ -45,8 +43,6 @@ def main():
         HuggingfaceBench,
         OgaLoad,
         OgaBench,
-        QuarkQuantize,
-        QuarkLoad,
         LemonadeReport,
         # Inherited from lemonade
         Cache,

lemonade/tools/report/table.py

@@ -591,7 +591,7 @@ class LemonadePerfTable(Table):
                     _wrap("Total Generated Tokens", 9),
                     Keys.RESPONSE_TOKENS,
                     "d",
-                    stat_fn=sum,
+                    stat_fn=lambda x: sum(_to_list(x)),
                 ),
                 SimpleStat(
                     _wrap("Memory Used (GB)", 8), Keys.MAX_MEMORY_USED_GBYTE, ".3f"

lemonade/tools/server/llamacpp.py

@@ -6,6 +6,7 @@ import threading
 import platform
 
 from dotenv import load_dotenv
+from fastapi import HTTPException, status
 
 from lemonade_server.pydantic_models import (
     PullConfig,
@@ -28,6 +29,20 @@ class LlamaTelemetry(WrappedServerTelemetry):
         Parse telemetry data from llama server output lines.
         """
 
+        if "vk::PhysicalDevice::createDevice: ErrorExtensionNotPresent" in line:
+            msg = (
+                "Your AMD GPU driver version is not compatible with this software.\n"
+                "Please update and try again: "
+                "https://www.amd.com/en/support/download/drivers.html"
+            )
+            logging.error(msg)
+            raise HTTPException(
+                status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+                detail=msg,
+            )
+        elif "error" in line.lower():
+            logging.error(line)
+
         # Parse Vulkan device detection
         vulkan_match = re.search(r"ggml_vulkan: Found (\d+) Vulkan devices?:", line)
         if vulkan_match:
@@ -182,7 +197,7 @@ class LlamaServer(WrappedServer):
         exe_dir = os.path.dirname(exe_path)
         env_file_path = os.path.join(exe_dir, ".env")
         if os.path.exists(env_file_path):
-            load_dotenv(env_file_path, override=True)
+            load_dotenv(env_file_path, override=False)
             env.update(os.environ)
             logging.debug(f"Loaded environment variables from {env_file_path}")
 

lemonade/tools/server/serve.py

@@ -133,6 +133,21 @@ class StopOnEvent:
         return self.stop_event.is_set()
 
 
+class NoCacheStaticFiles(StaticFiles):
+    """Custom StaticFiles class with no-cache headers"""
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+    def file_response(self, *args, **kwargs) -> Response:
+        response = super().file_response(*args, **kwargs)
+        # Add no-cache headers for all static files
+        response.headers["Cache-Control"] = "no-cache, no-store, must-revalidate"
+        response.headers["Pragma"] = "no-cache"
+        response.headers["Expires"] = "0"
+        return response
+
+
 class Server:
     """
     Open a web server that apps can use to communicate with the LLM.
@@ -198,7 +213,7 @@ class Server:
         # as the Web App
         static_dir = Path(__file__).parent / "static"
         self.app.mount(
-            "/static", StaticFiles(directory=static_dir), name="static_assets"
+            "/static", NoCacheStaticFiles(directory=static_dir), name="static_assets"
         )
 
         # Performance stats that are set during /ws and can be
@@ -1145,18 +1160,33 @@ class Server:
             )
             self.input_tokens = len(input_ids[0])
 
-        # For non-llamacpp recipes, truncate inputs to ctx_size if needed
-        if self.llm_loaded.recipe != "llamacpp" and self.input_tokens > self.ctx_size:
-            # Truncate input ids
-            truncate_amount = self.input_tokens - self.ctx_size
-            input_ids = input_ids[: self.ctx_size]
+        max_prompt_length = self.ctx_size  # Default fallback
+        # For OGA models, try to read the actual max prompt length from config
+        if "oga-" in self.llm_loaded.recipe:
+            try:
+                if model.config and model.config.get("max_prompt_length"):
+                    max_prompt_length = model.config["max_prompt_length"]
+                    logging.debug(
+                        f"Using OGA model max_prompt_length: {max_prompt_length}"
+                    )
+            # pylint: disable=broad-exception-caught
+            except Exception as e:
+                logging.debug(f"Could not read OGA model config, using ctx_size: {e}")
 
+        # Apply truncation if input exceeds the limit
+        if self.input_tokens > max_prompt_length:
+            # Truncate input ids
+            truncate_amount = self.input_tokens - max_prompt_length
+            input_ids = input_ids[:max_prompt_length]
             # Update token count
-            self.input_tokens = len(input_ids)
+            if "oga-" in self.llm_loaded.recipe:
+                self.input_tokens = len(input_ids)
+            else:
+                self.input_tokens = len(input_ids[0])
 
-            # Show warning message
+            # Log warning message instead of raising exception
             truncation_message = (
-                f"Input exceeded {self.ctx_size} tokens. "
+                f"Input exceeded {max_prompt_length} tokens. "
                 f"Truncated {truncate_amount} tokens from the beginning."
             )
             logging.warning(truncation_message)

lemonade/tools/server/static/js/chat.js

@@ -1,6 +1,7 @@
 // Chat logic and functionality
 let messages = [];
 let attachedFiles = [];
+let systemMessageElement = null;
 
 // Default model configuration
 const DEFAULT_MODEL = 'Qwen2.5-0.5B-Instruct-CPU';
@@ -28,6 +29,9 @@ document.addEventListener('DOMContentLoaded', function() {
     // Update attachment button state periodically
     updateAttachmentButtonState();
     setInterval(updateAttachmentButtonState, 1000);
+    
+    // Display initial system message
+    displaySystemMessage();
 });
 
 function setupChatEventListeners() {
@@ -163,25 +167,30 @@ function updateAttachmentButtonState() {
         attachmentBtn.style.opacity = '0.5';
         attachmentBtn.style.cursor = 'not-allowed';
         attachmentBtn.title = 'Load a model first';
-        return;
-    }
-    
-    const isVision = isVisionModel(currentLoadedModel);
-    
-    if (isVision) {
-        attachmentBtn.style.opacity = '1';
-        attachmentBtn.style.cursor = 'pointer';
-        attachmentBtn.title = 'Attach images';
     } else {
-        attachmentBtn.style.opacity = '0.5';
-        attachmentBtn.style.cursor = 'not-allowed';
-        attachmentBtn.title = 'Image attachments not supported by this model';
+        const isVision = isVisionModel(currentLoadedModel);
+        
+        if (isVision) {
+            attachmentBtn.style.opacity = '1';
+            attachmentBtn.style.cursor = 'pointer';
+            attachmentBtn.title = 'Attach images';
+        } else {
+            attachmentBtn.style.opacity = '0.5';
+            attachmentBtn.style.cursor = 'not-allowed';
+            attachmentBtn.title = 'Image attachments not supported by this model';
+        }
     }
+    
+    // Update system message when model state changes
+    displaySystemMessage();
 }
 
 // Make updateAttachmentButtonState accessible globally
 window.updateAttachmentButtonState = updateAttachmentButtonState;
 
+// Make displaySystemMessage accessible globally
+window.displaySystemMessage = displaySystemMessage;
+
 // Auto-load default model and send message
 async function autoLoadDefaultModelAndSend() {
     // Check if default model is available and installed
@@ -217,6 +226,7 @@ async function autoLoadDefaultModelAndSend() {
         },
         onError: (error, failedModelId) => {
             console.error('Error auto-loading default model:', error);
+            showErrorBanner('Failed to load model: ' + error.message);
         }
     });
 }
@@ -277,7 +287,7 @@ function handleChatInputKeydown(e) {
     if (e.key === 'Escape' && attachedFiles.length > 0) {
         e.preventDefault();
         clearAttachments();
-    } else if (e.key === 'Enter') {
+    } else if (e.key === 'Enter' && !e.shiftKey) {
         // Check if we have a loaded model
         if (currentLoadedModel && modelSelect.value !== '' && !modelSelect.disabled) {
             sendMessage();
@@ -438,7 +448,8 @@ function appendMessage(role, text, isMarkdown = false) {
     const bubble = document.createElement('div');
     bubble.className = 'chat-bubble ' + role;
     
-    if (role === 'llm' && isMarkdown) {
+    // Check if isMarkdown is true, regardless of role
+    if (isMarkdown) {
         bubble.innerHTML = renderMarkdownWithThinkTokens(text);
     } else {
         bubble.textContent = text;
@@ -450,6 +461,53 @@ function appendMessage(role, text, isMarkdown = false) {
     return bubble; // Return the bubble element for streaming updates
 }
 
+// Display system message based on current state
+function displaySystemMessage() {
+    // Remove existing system message if it exists
+    if (systemMessageElement) {
+        systemMessageElement.remove();
+        systemMessageElement = null;
+    }
+    
+    // Don't show system message if there are already user/LLM messages
+    if (messages.length > 0) {
+        return;
+    }
+    
+    let messageText = '';
+    
+    // Check if any models are installed
+    const hasInstalledModels = window.installedModels && window.installedModels.size > 0;
+    
+    if (!hasInstalledModels) {
+        // No models installed - show first message
+        messageText = `Welcome to Lemonade! To get started:
+1. Head over to the Model Management tab.
+2. Use the 📥Download button to download a model.
+3. Use the 🚀Load button to load the model.
+4. Come back to this tab, and you are ready to chat with the model.`;
+    } else if (!currentLoadedModel) {
+        // Models available but none loaded - show second message
+        messageText = 'Welcome to Lemonade! Choose a model from the dropdown menu below to load it and start chatting.';
+    }
+    
+    if (messageText) {
+        const div = document.createElement('div');
+        div.className = 'chat-message system';
+        div.setAttribute('data-system-message', 'true');
+        
+        const bubble = document.createElement('div');
+        bubble.className = 'chat-bubble system';
+        bubble.textContent = messageText;
+        
+        div.appendChild(bubble);
+        chatHistory.appendChild(div);
+        chatHistory.scrollTop = chatHistory.scrollHeight;
+        
+        systemMessageElement = div;
+    }
+}
+
 function updateMessageContent(bubbleElement, text, isMarkdown = false) {
     if (isMarkdown) {
         bubbleElement.innerHTML = renderMarkdownWithThinkTokens(text);
@@ -541,6 +599,12 @@ async function sendMessage() {
     const text = chatInput.value.trim();
     if (!text && attachedFiles.length === 0) return;
     
+    // Remove system message when user starts chatting
+    if (systemMessageElement) {
+        systemMessageElement.remove();
+        systemMessageElement = null;
+    }
+    
     // Check if a model is loaded, if not, automatically load the default model
     if (!currentLoadedModel) {
         const allModels = window.SERVER_MODELS || {};
@@ -624,7 +688,7 @@ async function sendMessage() {
         displayText = displayText ? `${displayText}\n[Images: ${fileNames}]` : `[Images: ${fileNames}]`;
     }
     
-    appendMessage('user', displayText);
+    appendMessage('user', displayText, true);
     
     // Add to messages array
     const userMessage = {

lemonade/tools/server/static/js/models.js

@@ -59,6 +59,11 @@ async function updateModelStatusIndicator() {
         window.initializeModelDropdown();
     }
     
+    // Update system message when model status changes
+    if (window.displaySystemMessage) {
+        window.displaySystemMessage();
+    }
+    
     // Refresh model management UI if we're on the models tab
     const modelsTab = document.getElementById('content-models');
     if (modelsTab && modelsTab.classList.contains('active')) {
@@ -417,6 +422,7 @@ async function loadModel(modelId) {
         },
         onError: (error, failedModelId) => {
             console.error(`Failed to load model ${failedModelId}:`, error);
+            showErrorBanner('Failed to load model: ' + error.message);
         }
     });
 }
@@ -699,6 +705,11 @@ async function refreshModelMgmtUI() {
     if (window.initializeModelDropdown) {
         window.initializeModelDropdown();
     }
+    
+    // Update system message when installed models change
+    if (window.displaySystemMessage) {
+        window.displaySystemMessage();
+    }
 }
 
 // Make refreshModelMgmtUI globally accessible

lemonade/tools/server/static/js/shared.js

@@ -54,6 +54,14 @@ function renderMarkdown(text) {
 
 // Display an error message in the banner
 function showErrorBanner(msg) {
+    // If DOM isn't ready, wait for it
+    if (document.readyState === 'loading') {
+        document.addEventListener('DOMContentLoaded', () => {
+            showErrorBanner(msg);
+        });
+        return;
+    }
+    
     const banner = document.getElementById('error-banner');
     if (!banner) return;
     const msgEl = document.getElementById('error-banner-msg');
@@ -303,12 +311,12 @@ async function loadModelStandardized(modelId, options = {}) {
             onLoadingEnd(modelId, false);
         }
         
-        // Call error callback or show default error
+        // Call error callback and always show default error banner as fallback
         if (onError) {
             onError(error, modelId);
-        } else {
-            showErrorBanner('Failed to load model: ' + error.message);
         }
+        // Always show error banner to ensure user sees the error
+        showErrorBanner('Failed to load model: ' + error.message);
         
         return false;
     }

lemonade/tools/server/static/styles.css

@@ -157,9 +157,8 @@ body::before {
   margin-bottom: 2em;
   border-radius: 8px;
   border: 1px solid #e0e0e0;
-  max-width: 1000px;
   min-width: 320px;
-  width: calc(100% - 2rem); /* Responsive width with margin */
+  width: 100%;
   margin-left: 1rem;
   margin-right: 1rem;
 }
@@ -327,15 +326,16 @@ body::before {
 .chat-container {
   display: flex;
   flex-direction: column;
-  height: calc(100vh - 650px); /* Subtract space for navbar, title, wall of logos, etc */
   min-height: 300px;
-  max-height: 1200px;
-  max-width: 800px;
+  min-width: 300px;
+  max-width: 100%;
   width: 100%;
   margin: 0 auto;
   border: 1px solid #e0e0e0;
   border-radius: 8px;
   background: #fff;
+  resize: both;
+  overflow: auto;
 }
 
 .chat-history {
@@ -388,6 +388,21 @@ body::before {
   align-self: flex-start;
 }
 
+.chat-message.system {
+  align-items: flex-start;
+}
+
+.chat-bubble.system {
+  background: linear-gradient(135deg, #f0f8f0 0%, #e8f5e8 100%);
+  color: #2d7f47;
+  border-bottom-left-radius: 4px;
+  align-self: flex-start;
+  border: 1px solid #c8e6c9;
+  font-style: normal;
+  font-weight: 500;
+  box-shadow: 0 1px 3px rgba(45, 127, 71, 0.1);
+}
+
 /* Markdown styling within chat bubbles */
 .chat-bubble h1,
 .chat-bubble h2,
@@ -570,7 +585,7 @@ body::before {
   align-items: center;
 }
 
-.input-with-indicator input[type='text'] {
+#chat-input {
   flex: 1;
   padding: 0.5em;
   border: 1px solid #ddd;
@@ -578,6 +593,16 @@ body::before {
   background: #fff;
   color: #222;
   margin: 0;
+  resize: vertical;
+  min-height: 40px;
+  font-family: inherit;
+}
+
+/* Update placeholder style */
+#chat-input::placeholder {
+  color: #aaa;
+  opacity: 1;
+  font-style: italic;
 }
 
 #attachment-indicator {
@@ -1897,7 +1922,6 @@ body::before {
 
 .category-content {
   display: none;
-  padding: 0;
 }
 
 .category-content.expanded {

lemonade/tools/server/static/webapp.html

@@ -52,7 +52,7 @@
                             <option value="">Pick a model</option>
                         </select>
                         <div class="input-with-indicator">
-                            <input type="text" id="chat-input" placeholder="Type your message..." /> 
+                            <textarea id="chat-input" placeholder="Type your message..." rows="1"></textarea> 
                         </div>
                         <input type="file" id="file-attachment" style="display: none;" multiple accept="image/*">
                         <button id="attachment-btn" title="Attach files">&#x1F4CE;</button>
@@ -92,9 +92,9 @@
                             <img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/lm_eval.png" alt="LM-Eval" class="app-logo-img">
                             <span class="app-name">LM-Eval</span>
                         </a>
-                        <a href="https://lemonade-server.ai/docs/server/apps/codeGPT/" target="_blank" class="app-logo-item" title="CodeGPT">
-                            <img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/codegpt.jpg" alt="CodeGPT" class="app-logo-img">
-                            <span class="app-name">CodeGPT</span>
+                        <a href="https://github.com/lemonade-sdk/lemonade-arcade" target="_blank" class="app-logo-item" title="Lemonade Arcade">
+                            <img src="https://raw.githubusercontent.com/lemonade-sdk/lemonade-arcade/refs/heads/main/docs/assets/favicon.ico" alt="Lemonade Arcade" class="app-logo-img">
+                            <span class="app-name">Lemonade Arcade</span>
                         </a>
                     <a href="https://github.com/lemonade-sdk/lemonade/blob/main/docs/server/apps/ai-toolkit.md" target="_blank" class="app-logo-item" title="AI Toolkit">
                         <img src="https://raw.githubusercontent.com/lemonade-sdk/assets/refs/heads/main/partner_logos/ai_toolkit.png" alt="AI Toolkit" class="app-logo-img">

lemonade/tools/server/wrapped_server.py

@@ -81,6 +81,7 @@ class WrappedServer(ABC):
         self.process: subprocess.Popen = None
         self.server_name: str = server_name
         self.telemetry: WrappedServerTelemetry = telemetry
+        self.log_thread_exception = None
 
     def choose_port(self):
         """
@@ -192,6 +193,8 @@ class WrappedServer(ABC):
 
                     if self.process.poll() is not None:
                         break
+            except HTTPException as e:
+                self.log_thread_exception = e
             except UnicodeDecodeError as e:
                 logging.debug(
                     "Unicode decode error reading subprocess output: %s", str(e)
@@ -217,6 +220,11 @@ class WrappedServer(ABC):
                 )
             time.sleep(1)
 
+        if self.log_thread_exception:
+            e = self.log_thread_exception
+            self.log_thread_exception = None
+            raise e
+
     @abstractmethod
     def _launch_server_subprocess(
         self,

lemonade/version.py

@@ -1 +1 @@
-__version__ = "8.1.6"
+__version__ = "8.1.8"

lemonade_install/install.py

@@ -48,10 +48,6 @@ NPU_DRIVER_DOWNLOAD_URL = (
 REQUIRED_NPU_DRIVER_VERSION = "32.0.203.280"
 
 lemonade_install_dir = Path(__file__).parent.parent.parent
-DEFAULT_QUARK_VERSION = "quark-0.6.0"
-DEFAULT_QUARK_DIR = os.path.join(
-    lemonade_install_dir, "install", "quark", DEFAULT_QUARK_VERSION
-)
 
 # List of supported Ryzen AI processor series (can be extended in the future)
 SUPPORTED_RYZEN_AI_SERIES = ["300"]
@@ -177,7 +173,7 @@ def _get_ryzenai_version_info(device=None):
             f"{e}\n Please install lemonade-sdk with "
             "one of the oga extras, for example:\n"
             "pip install lemonade-sdk[dev,oga-cpu]\n"
-            "See https://lemonade_server.ai/install_options.html for details"
+            "See https://lemonade-server.ai/install_options.html for details"
         ) from e
 
 
@@ -445,18 +441,18 @@ class Install:
             "variable (e.g., Ryzen AI uses environment variable OGA_TOKEN).",
         )
 
-        parser.add_argument(
-            "--quark",
-            help="Install Quark Quantization tool for LLMs",
-            choices=["0.6.0"],
-        )
-
         parser.add_argument(
             "--llamacpp",
             help="Install llama.cpp binaries with specified backend",
             choices=["rocm", "vulkan"],
         )
 
+        parser.add_argument(
+            "--override",
+            action="store_true",
+            help="Override the deprecation error to use legacy tools.",
+        )
+
         return parser
 
     @staticmethod
@@ -637,7 +633,7 @@ class Install:
         return file
 
     @staticmethod
-    def _install_ryzenai(ryzenai, build_model, yes, token):
+    def _install_ryzenai(ryzenai, build_model, yes, token, override=False):
         # Check if the processor is supported before proceeding
         check_ryzen_ai_processor()
 
@@ -654,6 +650,9 @@ class Install:
             + "=" * 80
             + "\n"
         )
+        if not override:
+            raise ValueError(warning_msg)
+
         print(warning_msg)
 
         # Delete any previous Ryzen AI installation in this environment
@@ -715,36 +714,6 @@ class Install:
         except IOError as e:
             print(f"An error occurred while writing {version_info_path}: {e}")
 
-    @staticmethod
-    def _install_quark(quark):
-        quark_install_dir = os.path.join(lemonade_install_dir, "install", "quark")
-        os.makedirs(quark_install_dir, exist_ok=True)
-
-        # Install Quark utilities
-        quark_url = (
-            f"https://www.xilinx.com/bin/public/openDownload?filename=quark-{quark}.zip"
-        )
-        quark_path = download_and_extract_package(
-            url=quark_url,
-            version=quark,
-            install_dir=quark_install_dir,
-            package_name="quark",
-        )
-        # Install Quark wheel
-        wheel_url = (
-            "https://www.xilinx.com/bin/public/openDownload?"
-            f"filename=quark-{quark}-py3-none-any.whl"
-        )
-        wheel_path = os.path.join(quark_install_dir, f"quark-{quark}-py3-none-any.whl")
-        print(f"\nInstalling Quark wheel from {wheel_url}")
-        download_file(wheel_url, wheel_path, "wheel file")
-
-        install_cmd = f"{sys.executable} -m pip install --no-deps {wheel_path}"
-        subprocess.run(install_cmd, check=True, shell=True)
-        os.remove(wheel_path)
-
-        print(f"\nQuark installed successfully at: {quark_path}")
-
     @staticmethod
     def _install_llamacpp(backend):
         """
@@ -762,22 +731,19 @@ class Install:
         self,
         ryzenai: Optional[str] = None,
         build_model: Optional[str] = None,
-        quark: Optional[str] = None,
         llamacpp: Optional[str] = None,
         yes: bool = False,
         token: Optional[str] = None,
+        override: bool = False,
     ):
-        if ryzenai is None and quark is None and llamacpp is None:
+        if ryzenai is None and llamacpp is None:
             raise ValueError(
                 "You must select something to install, "
-                "for example `--ryzenai`, `--quark`, or `--llamacpp`"
+                "for example `--ryzenai` or `--llamacpp`"
             )
 
         if ryzenai is not None:
-            self._install_ryzenai(ryzenai, build_model, yes, token)
-
-        if quark is not None:
-            self._install_quark(quark)
+            self._install_ryzenai(ryzenai, build_model, yes, token, override)
 
         if llamacpp is not None:
             self._install_llamacpp(llamacpp)