crewplus 0.2.29__tar.gz → 0.2.31__tar.gz

{crewplus-0.2.29 → crewplus-0.2.31}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: crewplus
-Version: 0.2.29
+Version: 0.2.31
 Summary: Base services for CrewPlus AI applications
 Author-Email: Tim Liu <tim@opsmateai.com>
 License: MIT
@@ -9,9 +9,9 @@ Project-URL: Documentation, https://crewplus.readthedocs.io
 Project-URL: Repository, https://github.com/your-org/crewplus-base
 Project-URL: Issues, https://github.com/your-org/crewplus-base/issues
 Requires-Python: <4.0,>=3.11
-Requires-Dist: langchain==0.3.25
-Requires-Dist: langchain-openai==0.3.24
-Requires-Dist: google-genai==1.21.1
+Requires-Dist: langchain>=0.3.25
+Requires-Dist: langchain-openai>=0.3.24
+Requires-Dist: google-genai>=1.21.1
 Requires-Dist: langchain-milvus<0.3.0,>=0.2.1
 Requires-Dist: langfuse<4.0.0,>=3.1.3
 Requires-Dist: langchain-mcp-adapters>=0.1.4

{crewplus-0.2.29 → crewplus-0.2.31}/crewplus/services/__init__.py

@@ -8,6 +8,5 @@ __all__ = [
     "init_load_balancer", 
     "get_model_balancer", 
     "ModelLoadBalancer", 
-    "init_services", 
     "TracedAzureChatOpenAI"
 ]

{crewplus-0.2.29 → crewplus-0.2.31}/crewplus/services/model_load_balancer.py

@@ -76,21 +76,23 @@ class ModelLoadBalancer:
             self.logger.error(f"Failed to load model configuration: {e}", exc_info=True)
             raise RuntimeError(f"Failed to load model configuration: {e}")
 
-    def get_model(self, provider: str = None, model_type: str = None, deployment_name: str = None, with_metadata: bool = False):
+    def get_model(self, provider: str = None, model_type: str = None, deployment_name: str = None, with_metadata: bool = False, selection_strategy: str = 'random'):
         """
         Get a model instance.
 
         Can fetch a model in two ways:
         1. By its specific `deployment_name`.
-        2. By `provider` and `model_type`, which will select a model using round-robin.
+        2. By `provider` and `model_type`, which will select a model using a specified strategy.
 
         Args:
             provider: The model provider (e.g., 'azure-openai', 'google-genai').
             model_type: The type of model (e.g., 'inference', 'embedding', 'embedding-large').
             deployment_name: The unique name for the model deployment.
+            with_metadata: If True, returns a tuple of (model, deployment_name).
+            selection_strategy: The selection strategy ('random', 'round_robin', or 'least_used'). Defaults to 'random'.
 
         Returns:
-            An instantiated language model object.
+            An instantiated language model object, or a tuple if with_metadata is True.
 
         Raises:
             RuntimeError: If the model configuration has not been loaded.
@@ -118,7 +120,16 @@ class ModelLoadBalancer:
                 self.logger.error(f"No models found for provider '{provider}' and type '{model_type}'")
                 raise ValueError(f"No models found for provider '{provider}' and type '{model_type}'")
 
-            selected_model_config = self._round_robin_selection(candidates)
+            if selection_strategy == 'random':
+                selected_model_config = self._random_selection(candidates)
+            elif selection_strategy == 'round_robin':
+                selected_model_config = self._round_robin_selection(candidates)
+            elif selection_strategy == 'least_used':
+                selected_model_config = self._least_used_selection(candidates)
+            else:
+                self.logger.warning(f"Unsupported selection strategy: '{selection_strategy}'. Defaulting to 'random'.")
+                selected_model_config = self._random_selection(candidates)
+                
             model_id = selected_model_config['id']
             model = self.models[model_id]
             if with_metadata:
@@ -186,6 +197,12 @@ class ModelLoadBalancer:
         self.usage_counter = defaultdict(int)
         self.current_indices = {}
 
+    def _random_selection(self, candidates: list) -> Dict:
+        """Selects a model randomly from a list of candidates."""
+        model = random.choice(candidates)
+        self.usage_counter[model['id']] += 1
+        return model
+
     def _round_robin_selection(self, candidates: list) -> Dict:
         if id(candidates) not in self.current_indices:
             self.current_indices[id(candidates)] = 0
@@ -197,8 +214,8 @@ class ModelLoadBalancer:
         return model
 
     def _least_used_selection(self, candidates: list) -> Dict:
-        min_usage = min(self.usage_counter[m['model_id']] for m in candidates)
-        least_used = [m for m in candidates if self.usage_counter[m['model_id']] == min_usage]
+        min_usage = min(self.usage_counter[m['id']] for m in candidates)
+        least_used = [m for m in candidates if self.usage_counter[m['id']] == min_usage]
         model = random.choice(least_used)
         self.usage_counter[model['id']] += 1
         return model

{crewplus-0.2.29 → crewplus-0.2.31}/crewplus/vectorstores/milvus/__init__.py

@@ -2,4 +2,4 @@ from .milvus_schema_manager import MilvusSchemaManager, ZillizSchemaManager
 from .schema_milvus import SchemaMilvus
 from .vdb_service import VDBService
 
-__all__ = ["MilvusSchemaManager", "ZillizSchemaManager", "VDBService"]
+__all__ = ["MilvusSchemaManager", "ZillizSchemaManager", "VDBService", "SchemaMilvus"]

{crewplus-0.2.29 → crewplus-0.2.31}/pyproject.toml

@@ -6,7 +6,7 @@ build-backend = "pdm.backend"
 
 [project]
 name = "crewplus"
-version = "0.2.29"
+version = "0.2.31"
 description = "Base services for CrewPlus AI applications"
 authors = [
     { name = "Tim Liu", email = "tim@opsmateai.com" },
@@ -14,9 +14,9 @@ authors = [
 readme = "README.md"
 requires-python = ">=3.11,<4.0"
 dependencies = [
-    "langchain==0.3.25",
-    "langchain-openai==0.3.24",
-    "google-genai==1.21.1",
+    "langchain>=0.3.25",
+    "langchain-openai>=0.3.24",
+    "google-genai>=1.21.1",
     "langchain-milvus (>=0.2.1,<0.3.0)",
     "langfuse (>=3.1.3,<4.0.0)",
     "langchain-mcp-adapters>=0.1.4",