lemonade-sdk 7.0.0__py3-none-any.whl → 7.0.1__py3-none-any.whl

lemonade/tools/server/static/styles.css

@@ -0,0 +1,313 @@
+body {
+  margin: 0;
+  font-family: 'Segoe UI', 'Arial', sans-serif;
+  background: #fffbe9;
+  color: #222;
+  min-height: 100vh;
+  display: flex;
+  flex-direction: column;
+  padding-bottom: 5rem;
+}
+
+.navbar {
+  display: flex;
+  justify-content: center;
+  gap: 2.5rem;
+  padding: 2rem 0 1.5rem 0;
+  font-size: 1.25rem;
+  font-weight: 500;
+  background: transparent;
+  letter-spacing: 0.02em;
+}
+
+.navbar a {
+  color: #444;
+  text-decoration: none;
+  transition: color 0.2s;
+}
+
+.navbar a:hover {
+  color: #e6b800;
+}
+
+.main {
+  flex: 1;
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  justify-content: flex-start;
+  min-height: 60vh;
+  margin-top: 3rem;
+}
+
+.title {
+  font-size: 3rem;
+  font-weight: 700;
+  margin-bottom: 2.5rem;
+  letter-spacing: 0.01em;
+  text-align: center;
+  color: #222;
+}
+
+.site-footer {
+  position: fixed;
+  left: 0;
+  bottom: 0;
+  width: 100%;
+  background-color: #fffbe9;
+  padding-top: 0.5rem;
+  z-index: 100;
+}
+
+.dad-joke {
+  color: #4ca64c;
+  font-size: 1.12rem;
+  text-align: center;
+  margin-bottom: 0.5rem;
+  opacity: 0.98;
+  letter-spacing: 0.01em;
+  padding: 0.2em 0;
+  width: fit-content;
+  margin-left: auto;
+  margin-right: auto;
+  background: none;
+  border-radius: 0;
+  display: block;
+}
+
+.copyright {
+  text-align: center;
+  font-size: 0.95rem;
+  color: #aaa;
+  margin-bottom: 0.5rem;
+}
+
+/* Tab and Chat UI Styling */
+.tab-container {
+  background: #fff;
+  box-shadow: 0 2px 8px rgba(0,0,0,0.1);
+  margin-bottom: 2em;
+  border-radius: 8px;
+  border: 1px solid #e0e0e0;
+  max-width: 900px;
+  width: 100%;
+}
+
+.tabs {
+  display: flex;
+  border-bottom: 1px solid #e0e0e0;
+}
+
+.tab {
+  padding: 1em 2em;
+  cursor: pointer;
+  border: none;
+  background: none;
+  font-size: 1.1em;
+  color: #666;
+  transition: color 0.2s;
+}
+
+.tab.active {
+  border-bottom: 2px solid #e6b800;
+  color: #e6b800;
+  font-weight: bold;
+  background: #fafafa;
+}
+
+.tab-content {
+  display: none;
+  padding: 2em;
+  background: #fafafa;
+  border-radius: 0 0 8px 8px;
+}
+
+.tab-content.active {
+  display: block;
+}
+
+/* Chat UI */
+.chat-container {
+  display: flex;
+  flex-direction: column;
+  height: 400px;
+  max-width: 600px;
+  margin: 0 auto;
+  border: 1px solid #e0e0e0;
+  border-radius: 8px;
+  background: #fff;
+}
+
+.chat-history {
+  flex: 1;
+  overflow-y: auto;
+  padding: 1em;
+  border-bottom: 1px solid #e0e0e0;
+  display: flex;
+  flex-direction: column;
+  gap: 0.5em;
+}
+
+.chat-message {
+  display: flex;
+  flex-direction: column;
+  align-items: flex-end;
+  margin-bottom: 0.5em;
+}
+
+.chat-message.user {
+  align-items: flex-end;
+}
+
+.chat-message.llm {
+  align-items: flex-start;
+}
+
+.chat-bubble {
+  max-width: 70%;
+  padding: 0.7em 1.1em;
+  border-radius: 20px;
+  margin-bottom: 2px;
+  font-size: 1em;
+  word-break: break-word;
+  box-shadow: 0 1px 2px rgba(0,0,0,0.10);
+  line-height: 1.5;
+}
+
+.chat-bubble.user {
+  background: linear-gradient(135deg, #ffe066 60%, #ffd43b 100%);
+  color: #222;
+  border-bottom-right-radius: 4px;
+  align-self: flex-end;
+}
+
+.chat-bubble.llm {
+  background: #f0f0f0;
+  color: #222;
+  border-bottom-left-radius: 4px;
+  align-self: flex-start;
+}
+
+.chat-input-row {
+  display: flex;
+  gap: 0.5em;
+  padding: 1em;
+  background: #f9f9f9;
+  border-radius: 0 0 8px 8px;
+}
+
+.chat-input-row select {
+  min-width: 120px;
+  background: #fff;
+  color: #222;
+  border: 1px solid #ddd;
+  border-radius: 4px;
+  padding: 0.5em;
+}
+
+.chat-input-row input[type='text'] {
+  flex: 1;
+  padding: 0.5em;
+  border: 1px solid #ddd;
+  border-radius: 4px;
+  background: #fff;
+  color: #222;
+}
+
+.chat-input-row button {
+  padding: 0.5em 1.2em;
+  background: #e6b800;
+  color: #222;
+  border: none;
+  border-radius: 4px;
+  cursor: pointer;
+  transition: background 0.2s;
+  font-weight: 600;
+}
+
+.chat-input-row button:hover {
+  background: #d4a500;
+}
+
+.chat-input-row button:disabled {
+  background: #ccc;
+  color: #666;
+  cursor: not-allowed;
+}
+
+/* Model Management */
+.model-mgmt-container {
+  display: flex;
+  gap: 2em;
+  align-items: flex-start;
+}
+
+.model-mgmt-pane {
+  flex: 1 1 0;
+  min-width: 0;
+}
+
+.model-mgmt-pane h3 {
+  margin-top: 0;
+  color: #222;
+}
+
+.model-table {
+  width: 100%;
+  border-collapse: collapse;
+  background: #fff;
+  border: 1px solid #ddd;
+  border-radius: 4px;
+}
+
+.model-table td {
+  padding: 0.5em 0.75em;
+  vertical-align: middle;
+  border-bottom: 1px solid #eee;
+  color: #222;
+}
+
+.model-table tr:last-child td {
+  border-bottom: none;
+}
+
+.model-table button {
+  background: #e6b800;
+  color: #222;
+  border: none;
+  border-radius: 4px;
+  padding: 0.3em 0.8em;
+  cursor: pointer;
+  font-weight: 600;
+  transition: background 0.2s;
+}
+
+.model-table button:hover {
+  background: #d4a500;
+}
+
+.installing-btn {
+  background: #ccc !important;
+  color: #666 !important;
+  font-weight: bold;
+  opacity: 1 !important;
+  border: 1px solid #999;
+  cursor: wait;
+}
+
+@media (max-width: 600px) {
+  .title { 
+    font-size: 2rem; 
+  }
+  .navbar { 
+    font-size: 1rem; 
+    gap: 1.2rem; 
+  }
+  .main { 
+    margin-top: 1rem; 
+  }
+  .model-mgmt-container {
+    flex-direction: column;
+    gap: 1em;
+  }
+}

lemonade/tools/server/tool_calls.py

@@ -1,10 +1,10 @@
 import re
-from typing import List, Dict
+from typing import List, Dict, Pattern, Optional
 import logging
 import json
 
 
-def extract_code_block(text):
+def extract_code_block(text: str) -> str:
     """
     Extracts the content inside triple backtick code blocks from a text.
 
@@ -54,67 +54,74 @@ def standardize_tool_call(tool_call: dict) -> dict | None:
     return standardized_tool_call
 
 
-def extract_tool_calls(
-    text: str, added_tokens_decoder: List[str]
-) -> tuple[List[Dict], str]:
+def get_tool_call_pattern(added_tokens_decoder: List[str]) -> Optional[Pattern]:
     """
-    Extracts tool calls from generated text based on tool calling identifiers.
-
-    Args:
-        text (str): The text output generated by the model.
-        added_tokens_decoder (List[str]): The list of tokens in the tokenizer.added_tokens_decoder.
-
-    Returns:
-        tuple[List[Dict], str]: A tuple containing:
-            - List[Dict]: A list of extracted tool call objects (raw JSON-like dicts)
-            - str: The original text with tool calls removed
+    Extracts tool call pattern from the added tokens decoder.
     """
-    matches = []
     special_tokens = [v.content for v in added_tokens_decoder.values()]
 
     # Pattern 1: <tool_call>...</tool_call> block
     # Sample model that uses this pattern: Qwen3-8B
     if "<tool_call>" in special_tokens and "</tool_call>" in special_tokens:
-        tool_call_pattern = re.compile(r"<tool_call>(.*?)</tool_call>", re.DOTALL)
-        matches = list(tool_call_pattern.finditer(text))
+        return re.compile(r"<tool_call>(.*?)</tool_call>", re.DOTALL)
 
     # Pattern 2: [TOOL_CALLS] [ {...} ] block
     # Sample model that uses this pattern: Mistral-7B-Instruct-v0.3
     elif "[TOOL_CALLS]" in special_tokens:
-        tool_call_pattern = re.compile(
-            r"\[TOOL_CALLS\]\s*\[(.*?)\](?=\s*<|/?eos|$)", re.DOTALL
-        )
-        matches = list(tool_call_pattern.finditer(text))
+        return re.compile(r"\[TOOL_CALLS\]\s*\[(.*?)\](?=\s*<|/?eos|$)", re.DOTALL)
 
     else:
         logging.warning(
             "Tool calling identifiers were not found for the current model."
         )
+        return None
+
+
+def extract_tool_calls(
+    text: str, tool_call_pattern: Optional[Pattern] = None
+) -> tuple[List[Dict], str]:
+    """
+    Extracts tool calls from generated text based on tool calling identifiers.
+
+    Args:
+        text (str): The text output generated by the model.
+        tool_call_pattern (Optional[Pattern]): The pattern to use to extract tool calls.
+
+    Returns:
+        tuple[List[Dict], str]: A tuple containing:
+            - List[Dict]: A list of extracted tool call objects (raw JSON-like dicts)
+            - str: The original text with tool calls removed
+    """
+
+    matches = []
+    if tool_call_pattern is not None:
+        matches = list(tool_call_pattern.finditer(text))
 
     # Some models don't use any tool calling identifiers.
     # Instead, tool calls are identified by only generating JSON content.
     # Sample model that uses this pattern: Llama-3.1-8B-Instruct
-    try:
-        # Remove the json for a code block if needed
-        parsed_text = extract_code_block(text)
-        json_tool_calls = json.loads(parsed_text)
-
-        if isinstance(json_tool_calls, dict):
-            json_tool_calls = [json_tool_calls]
-
-        extracted_tool_calls = []
-        for tool_call in json_tool_calls:
-            # Return the tool call if all calls are valid
-            standard_tool_call = standardize_tool_call(tool_call)
-            if standard_tool_call is not None:
-                extracted_tool_calls.append(standard_tool_call)
-            else:
-                return [], text
-
-        return extracted_tool_calls, ""
-
-    except json.JSONDecodeError:
-        pass
+    else:
+        try:
+            # Remove the json for a code block if needed
+            parsed_text = extract_code_block(text)
+            json_tool_calls = json.loads(parsed_text)
+
+            if isinstance(json_tool_calls, dict):
+                json_tool_calls = [json_tool_calls]
+
+            extracted_tool_calls = []
+            for tool_call in json_tool_calls:
+                # Return the tool call if all calls are valid
+                standard_tool_call = standardize_tool_call(tool_call)
+                if standard_tool_call is not None:
+                    extracted_tool_calls.append(standard_tool_call)
+                else:
+                    return [], text
+
+            return extracted_tool_calls, ""
+
+        except json.JSONDecodeError:
+            pass
 
     # Process matches in reverse to avoid position shifting
     extracted_tool_calls = []

lemonade/version.py

@@ -1 +1 @@
-__version__ = "7.0.0"
+__version__ = "7.0.1"

{lemonade_sdk-7.0.0.dist-info → lemonade_sdk-7.0.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lemonade-sdk
-Version: 7.0.0
+Version: 7.0.1
 Summary: Lemonade SDK: Your LLM Aide for Validation and Deployment
 Author-email: lemonade@amd.com
 Requires-Python: >=3.10, <3.12
@@ -25,8 +25,8 @@ Requires-Dist: matplotlib
 Requires-Dist: tabulate
 Requires-Dist: huggingface-hub==0.30.2
 Provides-Extra: llm
-Requires-Dist: torch>=2.0.0; extra == "llm"
-Requires-Dist: transformers; extra == "llm"
+Requires-Dist: torch>=2.6.0; extra == "llm"
+Requires-Dist: transformers<=4.51.3; extra == "llm"
 Requires-Dist: accelerate; extra == "llm"
 Requires-Dist: py-cpuinfo; extra == "llm"
 Requires-Dist: sentencepiece; extra == "llm"
@@ -34,23 +34,20 @@ Requires-Dist: datasets; extra == "llm"
 Requires-Dist: human-eval-windows==1.0.4; extra == "llm"
 Requires-Dist: fastapi; extra == "llm"
 Requires-Dist: uvicorn[standard]; extra == "llm"
-Requires-Dist: openai>=1.66.0; extra == "llm"
+Requires-Dist: openai>=1.81.0; extra == "llm"
 Requires-Dist: lm-eval[api]; extra == "llm"
 Provides-Extra: llm-oga-cpu
 Requires-Dist: onnxruntime-genai==0.6.0; extra == "llm-oga-cpu"
 Requires-Dist: onnxruntime<1.22.0,>=1.10.1; extra == "llm-oga-cpu"
-Requires-Dist: torch<2.4,>=2.0.0; extra == "llm-oga-cpu"
 Requires-Dist: lemonade-sdk[llm]; extra == "llm-oga-cpu"
 Provides-Extra: llm-oga-igpu
 Requires-Dist: onnxruntime-genai-directml==0.6.0; extra == "llm-oga-igpu"
 Requires-Dist: onnxruntime-directml<1.22.0,>=1.19.0; extra == "llm-oga-igpu"
-Requires-Dist: torch<2.4,>=2.0.0; extra == "llm-oga-igpu"
 Requires-Dist: transformers<4.45.0; extra == "llm-oga-igpu"
 Requires-Dist: lemonade-sdk[llm]; extra == "llm-oga-igpu"
 Provides-Extra: llm-oga-cuda
 Requires-Dist: onnxruntime-genai-cuda==0.6.0; extra == "llm-oga-cuda"
 Requires-Dist: onnxruntime-gpu<1.22.0,>=1.19.1; extra == "llm-oga-cuda"
-Requires-Dist: torch<2.4,>=2.0.0; extra == "llm-oga-cuda"
 Requires-Dist: transformers<4.45.0; extra == "llm-oga-cuda"
 Requires-Dist: lemonade-sdk[llm]; extra == "llm-oga-cuda"
 Provides-Extra: llm-oga-npu

{lemonade_sdk-7.0.0.dist-info → lemonade_sdk-7.0.1.dist-info}/RECORD

@@ -4,7 +4,7 @@ lemonade/cache.py,sha256=djr2qgyUUAWlQv8FehU9qlNtCwK0IZqo82hcBDyZ3-A,2850
 lemonade/cli.py,sha256=_s-LWpaVIhOmaP0Q1qirXxNiBhdumAZ-5ub5-lRNccs,4351
 lemonade/sequence.py,sha256=KSH7BPsiyDKsOsg_ziQKEGsDwMmuO_YbgPRBxkZd0pw,13267
 lemonade/state.py,sha256=sdSezla7Cd7KYL90xY3p9kcNV4ndSyN6UvNLOr3vBMA,5261
-lemonade/version.py,sha256=VgMOOqsYbyb60I1RmlZpqwqQ0C0IyT3R0c8_xX4pRGM,22
+lemonade/version.py,sha256=co6LyaBArt-ahHXYZSdSER8TFZ2vVTb86CNG6X8Pxwc,22
 lemonade/common/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 lemonade/common/analyze_model.py,sha256=sYWDznEUEWjx_Qekg7f1hHY4Pfe87IQ77lmsWqePgE0,803
 lemonade/common/build.py,sha256=Pk86mCr6fyBIx2zXDpq0BkdahlCmWRnwSTpShA_gwZw,7849
@@ -45,17 +45,21 @@ lemonade/tools/report/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3h
 lemonade/tools/report/llm_report.py,sha256=bVHhwCINA-Ok2EdSwAsLubsc83N3KWOVuwTguw7jDcE,6676
 lemonade/tools/report/table.py,sha256=a0TXo1X84RxCSu0un_XM3ANOlhLtPDuqtGwR7eomf2s,24853
 lemonade/tools/server/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-lemonade/tools/server/serve.py,sha256=VCBeu-aVYC5BjEo5mzKIjsN48YKCNoxXgDggXxmRgy8,49941
-lemonade/tools/server/tool_calls.py,sha256=8l8K94Hfh2Q1421kUNECDFzeE7CNtvX9MKsdLVUdIng,5005
+lemonade/tools/server/instructions.py,sha256=Lvm-tRZaYgHkyt3zQkmMChkXO6rUiLoIAunudmMr_D8,13388
+lemonade/tools/server/llamacpp.py,sha256=PeHg1DbMGcf68txFgC1CJJN5HRHEnIJ4_4EDhvqAFUI,9255
+lemonade/tools/server/pydantic_models.py,sha256=z1RAs9hkAFkOfMiTPtmUiC3CD2P6OMI2N0J2ztNs0d4,2179
+lemonade/tools/server/serve.py,sha256=7meKOKVHaODHBYD_3dDJyaiwoC_m4z_FWniZfsZ9cCI,50655
+lemonade/tools/server/tool_calls.py,sha256=xrAlQwKG-nv2xLlf8f9CDSaUbyMn8ZtHkds9iZLG9K8,5230
+lemonade/tools/server/static/styles.css,sha256=8U1EejQaqRLQ6QTCF5UG_dLPtLjRwT1menUHMDhaq2M,5045
 lemonade_install/__init__.py,sha256=26zohKg2jgr_5y7tObduWMYQg8zCTWMZHL8lfi2zZVQ,40
 lemonade_install/install.py,sha256=61qUO7kWCLcdjK0_IQZ46-rKP_AWkyznh4YpDclPKyM,28036
-lemonade_sdk-7.0.0.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-lemonade_sdk-7.0.0.dist-info/licenses/NOTICE.md,sha256=B8lEqi4QE41J9ljz4Riv2JgHD1v8GCZE6nNBHO3KIA0,2135
-lemonade_server/cli.py,sha256=PPW6Rvmqp4Nzadsn49mrW532XlqPYQQGApnubgF-xVg,7270
-lemonade_server/model_manager.py,sha256=Z0FmP7DVru1Rv0MNPDfRHLB8rORyGBzB1-ty9uLvGYA,3477
-lemonade_server/server_models.json,sha256=ys9oOlwBo0TmWjcNU1JsvGOAcxVCbt9U0tLzP7vGIN8,4997
-lemonade_sdk-7.0.0.dist-info/METADATA,sha256=9DvnEjMCXsNHrPw5EenEIgy4jlj8sDTiWB3HsdfN_wA,5608
-lemonade_sdk-7.0.0.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
-lemonade_sdk-7.0.0.dist-info/entry_points.txt,sha256=gJppn0ETtXXR6ceKWEIRdk42kMC7ps59EmU3NCPyPUk,144
-lemonade_sdk-7.0.0.dist-info/top_level.txt,sha256=10ap5GNiPhalO4V50LRoxA1FqRT9g3Xkia6BITu880k,42
-lemonade_sdk-7.0.0.dist-info/RECORD,,
+lemonade_sdk-7.0.1.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+lemonade_sdk-7.0.1.dist-info/licenses/NOTICE.md,sha256=B8lEqi4QE41J9ljz4Riv2JgHD1v8GCZE6nNBHO3KIA0,2135
+lemonade_server/cli.py,sha256=DR6sIt66K1sZZG3ascEw_6HUgz3UhU9KGUyzxf4nO_A,7351
+lemonade_server/model_manager.py,sha256=WDGDxrKjq-u2GkGWLNUsRk0d74J-RG2yCYEnH8WMnDw,4010
+lemonade_server/server_models.json,sha256=ZSg1R555bLVW4U7BPaYX5ZgwaJVNAP3z1C62dzMRqAM,6198
+lemonade_sdk-7.0.1.dist-info/METADATA,sha256=bvg9-Tzg_v8sTKjkAJtLahpDq_GmLDMDKA9PTisaNGw,5443
+lemonade_sdk-7.0.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+lemonade_sdk-7.0.1.dist-info/entry_points.txt,sha256=gJppn0ETtXXR6ceKWEIRdk42kMC7ps59EmU3NCPyPUk,144
+lemonade_sdk-7.0.1.dist-info/top_level.txt,sha256=10ap5GNiPhalO4V50LRoxA1FqRT9g3Xkia6BITu880k,42
+lemonade_sdk-7.0.1.dist-info/RECORD,,

{lemonade_sdk-7.0.0.dist-info → lemonade_sdk-7.0.1.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.7.1)
+Generator: setuptools (80.9.0)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

lemonade_server/cli.py

@@ -92,7 +92,7 @@ def pull(model_names: List[str]):
     if server_running:
         import requests
 
-        base_url = f"http://localhost:{port}/api/v0"
+        base_url = f"http://localhost:{port}/api/v1"
 
         for model_name in model_names:
             # Install the model
@@ -143,7 +143,7 @@ def status(verbose: bool = True) -> Tuple[bool, int]:
 
 def is_lemonade_server(pid):
     """
-    Check wether or not a given PID corresponds to a Lemonade server
+    Check whether or not a given PID corresponds to a Lemonade server
     """
     try:
         process = psutil.Process(pid)
@@ -158,6 +158,8 @@ def is_lemonade_server(pid):
                 "lemonade",
             ]:
                 return True
+            elif "llama-server" in process.name():
+                return False
             if not process.parent():
                 return False
             process = process.parent()

lemonade_server/model_manager.py

@@ -50,7 +50,7 @@ class ModelManager:
         downloaded_models = {}
         for model in self.supported_models:
             if (
-                self.supported_models[model]["checkpoint"]
+                self.supported_models[model]["checkpoint"].split(":")[0]
                 in self.downloaded_hf_checkpoints
             ):
                 downloaded_models[model] = self.supported_models[model]
@@ -62,22 +62,17 @@ class ModelManager:
         Returns a dictionary of locally available models that are enabled by
         the current installation.
         """
-        hybrid_installed = (
-            "onnxruntime-vitisai" in pkg_resources.working_set.by_key
-            and "onnxruntime-genai-directml-ryzenai" in pkg_resources.working_set.by_key
-        )
+        return self.filter_models_by_backend(self.downloaded_models)
 
-        downloaded_models_enabled = {}
-        for model, value in self.downloaded_models.items():
-            if value["recipe"] == "oga-hybrid" and hybrid_installed:
-                downloaded_models_enabled[model] = value
-            else:
-                # All other models are CPU models right now
-                # This logic will get more sophisticated when we
-                # start to support more backends
-                downloaded_models_enabled[model] = value
-
-        return downloaded_models_enabled
+    def download_gguf(self, checkpoint) -> str:
+        # The colon after the checkpoint name indicates which
+        # specific GGUF to download
+        repo_id = checkpoint.split(":")[0]
+        pattern_to_match = f'*{checkpoint.split(":")[1]}.gguf'
+        return huggingface_hub.snapshot_download(
+            repo_id=repo_id,
+            allow_patterns=[pattern_to_match],
+        )
 
     def download_models(self, models: list[str]):
         """
@@ -91,7 +86,29 @@ class ModelManager:
                 )
             checkpoint = self.supported_models[model]["checkpoint"]
             print(f"Downloading {model} ({checkpoint})")
-            huggingface_hub.snapshot_download(repo_id=checkpoint)
+
+            if "gguf" in checkpoint.lower():
+                self.download_gguf(checkpoint)
+            else:
+                huggingface_hub.snapshot_download(repo_id=checkpoint)
+
+    def filter_models_by_backend(self, models: dict) -> dict:
+        """
+        Returns a filtered dict of models that are enabled by the
+        current environment.
+        """
+        hybrid_installed = (
+            "onnxruntime-vitisai" in pkg_resources.working_set.by_key
+            and "onnxruntime-genai-directml-ryzenai" in pkg_resources.working_set.by_key
+        )
+        filtered = {}
+        for model, value in models.items():
+            if value.get("recipe") == "oga-hybrid":
+                if hybrid_installed:
+                    filtered[model] = value
+            else:
+                filtered[model] = value
+        return filtered
 
 
 # This file was originally licensed under Apache 2.0. It has been modified.

lemonade_server/server_models.json

@@ -138,5 +138,47 @@
         "recipe": "oga-igpu",
         "reasoning": false,
         "suggested": false
+    },
+    "Qwen3-0.6B-GGUF": {
+        "checkpoint": "unsloth/Qwen3-0.6B-GGUF:Q4_0",
+        "recipe": "llamacpp",
+        "reasoning": true,
+        "suggested": true
+    },
+    "Qwen3-1.7B-GGUF": {
+        "checkpoint": "unsloth/Qwen3-1.7B-GGUF:Q4_0",
+        "recipe": "llamacpp",
+        "reasoning": true,
+        "suggested": true
+    },
+    "Qwen3-4B-GGUF": {
+        "checkpoint": "unsloth/Qwen3-4B-GGUF:Q4_0",
+        "recipe": "llamacpp",
+        "reasoning": true,
+        "suggested": true
+    },
+    "Qwen3-8B-GGUF": {
+        "checkpoint": "unsloth/Qwen3-8B-GGUF:Q4_0",
+        "recipe": "llamacpp",
+        "reasoning": true,
+        "suggested": true
+    },
+    "DeepSeek-Qwen3-8B-GGUF": {
+        "checkpoint": "unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:Q4_1",
+        "recipe": "llamacpp",
+        "reasoning": true,
+        "suggested": true
+    },
+    "Qwen3-14B-GGUF": {
+        "checkpoint": "unsloth/Qwen3-14B-GGUF:Q4_0",
+        "recipe": "llamacpp",
+        "reasoning": true,
+        "suggested": true
+    },
+    "Qwen3-30B-A3B-GGUF": {
+        "checkpoint": "unsloth/Qwen3-30B-A3B-GGUF:Q4_0",
+        "recipe": "llamacpp",
+        "reasoning": true,
+        "suggested": true
     }
 }