llms-py 3.0.22__py3-none-any.whl → 3.0.23__py3-none-any.whl

llms/llms.json

@@ -235,17 +235,29 @@
             "enabled": true,
             "temperature": 1.0
         },
+        "lmstudio": {
+            "enabled": false,
+            "npm": "lmstudio",
+            "api": "http://127.0.0.1:1234/v1",
+            "models": {}
+        },
         "ollama": {
             "enabled": false,
             "id": "ollama",
             "npm": "ollama",
             "api": "http://localhost:11434"
         },
-        "lmstudio": {
+        "ollama-cloud": {
+            "enabled": true,
+            "env": [
+                "OLLAMA_API_KEY"
+            ]
+        },
+        "openai-local": {
             "enabled": false,
-            "npm": "lmstudio",
-            "api": "http://127.0.0.1:1234/v1",
-            "models": {}
+            "npm": "openai-local",
+            "api": "http://localhost:8000/v1",
+            "api_key": "$OPENAI_LOCAL_API_KEY"
         },
         "google": {
             "enabled": true,

llms/main.py

@@ -57,7 +57,7 @@ try:
 except ImportError:
     HAS_PIL = False
 
-VERSION = "3.0.22"
+VERSION = "3.0.23"
 _ROOT = None
 DEBUG = os.getenv("DEBUG") == "1"
 MOCK = os.getenv("MOCK") == "1"
@@ -871,8 +871,7 @@ def save_image_to_cache(base64_data, filename, image_info, ignore_info=False):
     return url, info
 
 
-async def response_json(response):
-    text = await response.text()
+def http_error_to_message(response, text):
     if response.status >= 400:
         message = "HTTP " + str(response.status) + " " + response.reason
         _dbg(f"HTTP {response.status} {response.reason}\n{dict(response.headers)}\n{text}")
@@ -885,6 +884,13 @@ async def response_json(response):
         except Exception:
             if text:
                 message += ": " + text[:100]
+        return message
+
+
+async def response_json(response):
+    text = await response.text()
+    if response.status >= 400:
+        message = http_error_to_message(response, text)
         raise Exception(message)
     response.raise_for_status()
     body = json.loads(text)
@@ -1394,6 +1400,7 @@ class OllamaProvider(OpenAiCompatible):
                 "id": k,
                 "name": v.replace(":", " "),
                 "modalities": {"input": ["text"], "output": ["text"]},
+                "tool_call": True,
                 "cost": {
                     "input": 0,
                     "output": 0,
@@ -1431,6 +1438,10 @@ class LMStudioProvider(OllamaProvider):
         return ret
 
 
+class OpenAiLocalProvider(LMStudioProvider):
+    sdk = "openai-local"
+
+
 def get_provider_model(model_name):
     for provider in g_handlers.values():
         provider_model = provider.provider_model(model_name)
@@ -2229,7 +2240,7 @@ async def get_text(url):
         async with session.get(url) as resp:
             text = await resp.text()
             if resp.status >= 400:
-                raise HTTPError(resp.status, reason=resp.reason, body=text, headers=dict(resp.headers))
+                raise Exception(http_error_to_message(resp, text))
             return text
 
 
@@ -2838,6 +2849,7 @@ class AppExtensions:
             CodestralProvider,
             OllamaProvider,
             LMStudioProvider,
+            OpenAiLocalProvider,
         ]
         self.aspect_ratios = {
             "1:1": "1024×1024",
@@ -2953,7 +2965,7 @@ class AppExtensions:
         for filter_func in self.chat_error_filters:
             try:
                 task = filter_func(e, context)
-                if asyncio.isfuture(task):
+                if inspect.iscoroutine(task):
                     await task
             except Exception as e:
                 _err("chat error filter failed", e)