camel-ai 0.2.77__py3-none-any.whl → 0.2.79a0__py3-none-any.whl

camel/datasets/base_generator.py

@@ -62,6 +62,7 @@ class BaseGenerator(abc.ABC, IterableDataset):
         self._buffer = buffer
         self._data: List[DataPoint] = []
         self._batch_to_save: List[DataPoint] = []
+        self._iter_position: int = 0
 
         if data_path:
             file_path = Path(data_path)
@@ -103,9 +104,9 @@ class BaseGenerator(abc.ABC, IterableDataset):
         r"""Async iterator that yields datapoints dynamically.
 
         If a `data_path` was provided during initialization, those datapoints
-        are yielded first. When self._data is empty, 20 new datapoints
-        are generated. Every 100 yields, the batch is appended to the
-        JSONL file or discarded if `cache` is None.
+        are yielded first. When self._iter_position reaches the end of _data,
+        new datapoints are generated. Every 100 yields, the batch is appended
+        to the JSONL file or discarded if `cache` is None.
 
         Yields:
             DataPoint: A single datapoint.
@@ -113,9 +114,10 @@ class BaseGenerator(abc.ABC, IterableDataset):
 
         async def generator():
             while True:
-                if not self._data:
+                if self._iter_position >= len(self._data):
                     await self.generate_new(self._buffer)
-                datapoint = self._data.pop(0)
+                datapoint = self._data[self._iter_position]
+                self._iter_position += 1
                 yield datapoint
                 self._batch_to_save.append(datapoint)
                 if len(self._batch_to_save) == 100:
@@ -132,9 +134,9 @@ class BaseGenerator(abc.ABC, IterableDataset):
         r"""Synchronous iterator for PyTorch IterableDataset compatibility.
 
         If a `data_path` was provided during initialization, those datapoints
-        are yielded first. When self._data is empty, 20 new datapoints
-        are generated. Every 100 yields, the batch is appended to the
-        JSONL file or discarded if `cache` is None.
+        are yielded first. When self._iter_position reaches the end of _data,
+        new datapoints are generated. Every 100 yields, the batch is appended
+        to the JSONL file or discarded if `cache` is None.
 
         Yields:
             DataPoint: A single datapoint.
@@ -150,9 +152,10 @@ class BaseGenerator(abc.ABC, IterableDataset):
                 raise
 
         while True:
-            if not self._data:
+            if self._iter_position >= len(self._data):
                 asyncio.run(self.generate_new(self._buffer))
-            datapoint = self._data.pop(0)
+            datapoint = self._data[self._iter_position]
+            self._iter_position += 1
             yield datapoint
             self._batch_to_save.append(datapoint)
             if len(self._batch_to_save) == 100:
@@ -248,6 +251,7 @@ class BaseGenerator(abc.ABC, IterableDataset):
 
         self.save_to_jsonl(file_path)
         self._data = []
+        self._iter_position = 0
         logger.info(f"Data flushed to {file_path} and cleared from the memory")
 
     def _init_from_jsonl(self, file_path: Path) -> List[Dict[str, Any]]:
@@ -290,3 +294,28 @@ class BaseGenerator(abc.ABC, IterableDataset):
             f"Successfully loaded {len(raw_data)} items from {file_path}"
         )
         return raw_data
+
+    def __getitem__(self, index: int) -> DataPoint:
+        r"""Get a datapoint by index without removing the datapoint from _data.
+
+        Args:
+            index (int): Index of the datapoint to retrieve.
+
+        Returns:
+            DataPoint: The datapoint at the specified index.
+
+        Raises:
+            IndexError: If the index is out of range.
+        """
+        if index < 0 or index >= len(self._data):
+            raise IndexError(f"Index {index} is out of range")
+
+        return self._data[index]
+
+    def __len__(self) -> int:
+        r"""Get the number of datapoints in the dataset.
+
+        Returns:
+            int: The number of datapoints.
+        """
+        return len(self._data)

camel/environments/single_step.py

@@ -218,9 +218,34 @@ class SingleStepEnv:
             return observations[0] if batch_size == 1 else observations
 
         elif isinstance(self.dataset, BaseGenerator):
-            self._states = [
-                await self.dataset.async_sample() for _ in range(batch_size)
-            ]
+            # Generate more data if needed
+            if batch_size > len(self.dataset):
+                new_datapoints_needed = batch_size - len(self.dataset)
+                await self.dataset.generate_new(n=new_datapoints_needed)
+
+                # Verify that enough data was generated
+                if len(self.dataset) < batch_size:
+                    raise RuntimeError(
+                        f"Failed to generate enough datapoints. "
+                        f"Requested {batch_size}, but only "
+                        f"{len(self.dataset)} available after generation."
+                    )
+
+            # Choose sampling strategy based on whether seed is provided
+            if seed is not None:
+                # Deterministic random sampling when seed is provided
+                random_indices = rng.sample(
+                    range(len(self.dataset)), batch_size
+                )
+                self._states = [self.dataset[ind] for ind in random_indices]
+            else:
+                # Sequential sampling when no seed (backward compatible)
+                # Use async_sample to maintain sequential behavior
+                self._states = [
+                    await self.dataset.async_sample()
+                    for _ in range(batch_size)
+                ]
+
             self.current_batch_size = batch_size
             self._states_done = [False] * batch_size
 

camel/memories/__init__.py

@@ -18,7 +18,7 @@ from .agent_memories import (
     VectorDBMemory,
 )
 from .base import AgentMemory, BaseContextCreator, MemoryBlock
-from .blocks.chat_history_block import ChatHistoryBlock, EmptyMemoryWarning
+from .blocks.chat_history_block import ChatHistoryBlock
 from .blocks.vectordb_block import VectorDBBlock
 from .context_creators.score_based import ScoreBasedContextCreator
 from .records import ContextRecord, MemoryRecord
@@ -35,5 +35,4 @@ __all__ = [
     'ChatHistoryBlock',
     'VectorDBBlock',
     'LongtermAgentMemory',
-    'EmptyMemoryWarning',
 ]

camel/memories/blocks/chat_history_block.py

@@ -11,7 +11,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
-import warnings
 from typing import List, Optional
 
 from camel.memories.base import MemoryBlock
@@ -21,17 +20,6 @@ from camel.storages.key_value_storages.in_memory import InMemoryKeyValueStorage
 from camel.types import OpenAIBackendRole
 
 
-class EmptyMemoryWarning(UserWarning):
-    """Warning raised when attempting to access an empty memory.
-
-    This warning is raised when operations are performed on memory
-    that contains no records. It can be safely caught and suppressed
-    in contexts where empty memory is expected.
-    """
-
-    pass
-
-
 class ChatHistoryBlock(MemoryBlock):
     r"""An implementation of the :obj:`MemoryBlock` abstract base class for
     maintaining a record of chat histories.
@@ -81,11 +69,8 @@ class ChatHistoryBlock(MemoryBlock):
         """
         record_dicts = self.storage.load()
         if len(record_dicts) == 0:
-            warnings.warn(
-                "The `ChatHistoryMemory` is empty.",
-                EmptyMemoryWarning,
-                stacklevel=1,
-            )
+            # Empty memory is a valid state (e.g., during initialization).
+            # Users can check if memory is empty by checking the returned list.
             return list()
 
         if window_size is not None and window_size >= 0:

camel/models/aws_bedrock_model.py

@@ -13,17 +13,11 @@
 # ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
 
 import os
-from typing import Any, Dict, List, Optional, Type, Union
-
-from openai import AsyncStream
-from pydantic import BaseModel
+from typing import Any, Dict, Optional, Union
 
 from camel.configs import BedrockConfig
-from camel.messages import OpenAIMessage
 from camel.models.openai_compatible_model import OpenAICompatibleModel
 from camel.types import (
-    ChatCompletion,
-    ChatCompletionChunk,
     ModelType,
 )
 from camel.utils import BaseTokenCounter, api_keys_required
@@ -93,13 +87,3 @@ class AWSBedrockModel(OpenAICompatibleModel):
             max_retries=max_retries,
             **kwargs,
         )
-
-    async def _arun(
-        self,
-        messages: List[OpenAIMessage],
-        response_format: Optional[Type[BaseModel]] = None,
-        tools: Optional[List[Dict[str, Any]]] = None,
-    ) -> Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
-        raise NotImplementedError(
-            "AWS Bedrock does not support async inference."
-        )

camel/models/moonshot_model.py

@@ -12,6 +12,7 @@
 # limitations under the License.
 # ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
 
+import copy
 import os
 from typing import Any, Dict, List, Optional, Type, Union
 
@@ -19,6 +20,7 @@ from openai import AsyncStream
 from pydantic import BaseModel
 
 from camel.configs import MoonshotConfig
+from camel.logger import get_logger
 from camel.messages import OpenAIMessage
 from camel.models._utils import try_modify_message_with_format
 from camel.models.openai_compatible_model import OpenAICompatibleModel
@@ -34,6 +36,8 @@ from camel.utils import (
     update_langfuse_trace,
 )
 
+logger = get_logger(__name__)
+
 if os.environ.get("LANGFUSE_ENABLED", "False").lower() == "true":
     try:
         from langfuse.decorators import observe
@@ -84,7 +88,7 @@ class MoonshotModel(OpenAICompatibleModel):
         model_type: Union[ModelType, str],
         model_config_dict: Optional[Dict[str, Any]] = None,
         api_key: Optional[str] = None,
-        url: Optional[str] = "https://api.moonshot.ai/v1",
+        url: Optional[str] = None,
         token_counter: Optional[BaseTokenCounter] = None,
         timeout: Optional[float] = None,
         max_retries: int = 3,
@@ -93,7 +97,12 @@ class MoonshotModel(OpenAICompatibleModel):
         if model_config_dict is None:
             model_config_dict = MoonshotConfig().as_dict()
         api_key = api_key or os.environ.get("MOONSHOT_API_KEY")
-        url = url or os.environ.get("MOONSHOT_API_BASE_URL")
+        # Preserve default URL if not provided
+        if url is None:
+            url = (
+                os.environ.get("MOONSHOT_API_BASE_URL")
+                or "https://api.moonshot.ai/v1"
+            )
         timeout = timeout or float(os.environ.get("MODEL_TIMEOUT", 180))
         super().__init__(
             model_type=model_type,
@@ -125,12 +134,12 @@ class MoonshotModel(OpenAICompatibleModel):
         Returns:
             Dict[str, Any]: The prepared request configuration.
         """
-        import copy
-
         request_config = copy.deepcopy(self.model_config_dict)
 
         if tools:
-            request_config["tools"] = tools
+            # Clean tools to remove null types (Moonshot API incompatibility)
+            cleaned_tools = self._clean_tool_schemas(tools)
+            request_config["tools"] = cleaned_tools
         elif response_format:
             # Use the same approach as DeepSeek for structured output
             try_modify_message_with_format(messages[-1], response_format)
@@ -138,6 +147,94 @@ class MoonshotModel(OpenAICompatibleModel):
 
         return request_config
 
+    def _clean_tool_schemas(
+        self, tools: List[Dict[str, Any]]
+    ) -> List[Dict[str, Any]]:
+        r"""Clean tool schemas to remove null types for Moonshot compatibility.
+
+        Moonshot API doesn't accept {"type": "null"} in anyOf schemas.
+        This method removes null type definitions from parameters.
+
+        Args:
+            tools (List[Dict[str, Any]]): Original tool schemas.
+
+        Returns:
+            List[Dict[str, Any]]: Cleaned tool schemas.
+        """
+
+        def remove_null_from_schema(schema: Any) -> Any:
+            """Recursively remove null types from schema."""
+            if isinstance(schema, dict):
+                # Create a copy to avoid modifying the original
+                result = {}
+
+                for key, value in schema.items():
+                    if key == 'type' and isinstance(value, list):
+                        # Handle type arrays like ["string", "null"]
+                        filtered_types = [t for t in value if t != 'null']
+                        if len(filtered_types) == 1:
+                            # Single type remains, convert to string
+                            result[key] = filtered_types[0]
+                        elif len(filtered_types) > 1:
+                            # Multiple types remain, keep as array
+                            result[key] = filtered_types
+                        else:
+                            # All were null, use string as fallback
+                            logger.warning(
+                                "All types in tool schema type array "
+                                "were null, falling back to 'string' "
+                                "type for Moonshot API compatibility. "
+                                "Original tool schema may need review."
+                            )
+                            result[key] = 'string'
+                    elif key == 'anyOf':
+                        # Handle anyOf with null types
+                        filtered = [
+                            item
+                            for item in value
+                            if not (
+                                isinstance(item, dict)
+                                and item.get('type') == 'null'
+                            )
+                        ]
+                        if len(filtered) == 1:
+                            # If only one type remains, flatten it
+                            return remove_null_from_schema(filtered[0])
+                        elif len(filtered) > 1:
+                            result[key] = [
+                                remove_null_from_schema(item)
+                                for item in filtered
+                            ]
+                        else:
+                            # All were null, return string type as fallback
+                            logger.warning(
+                                "All types in tool schema anyOf were null, "
+                                "falling back to 'string' type for "
+                                "Moonshot API compatibility. Original "
+                                "tool schema may need review."
+                            )
+                            return {"type": "string"}
+                    else:
+                        # Recursively process other values
+                        result[key] = remove_null_from_schema(value)
+
+                return result
+            elif isinstance(schema, list):
+                return [remove_null_from_schema(item) for item in schema]
+            else:
+                return schema
+
+        cleaned_tools = copy.deepcopy(tools)
+        for tool in cleaned_tools:
+            if 'function' in tool and 'parameters' in tool['function']:
+                params = tool['function']['parameters']
+                if 'properties' in params:
+                    params['properties'] = remove_null_from_schema(
+                        params['properties']
+                    )
+
+        return cleaned_tools
+
     @observe()
     async def _arun(
         self,

camel/societies/workforce/events.py

@@ -0,0 +1,122 @@
+# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
+from __future__ import annotations
+
+from datetime import datetime, timezone
+from typing import Any, Dict, List, Literal, Optional, Union
+
+from pydantic import BaseModel, ConfigDict, Field
+
+
+class WorkforceEventBase(BaseModel):
+    model_config = ConfigDict(frozen=True, extra='forbid')
+    event_type: Literal[
+        "task_decomposed",
+        "task_created",
+        "task_assigned",
+        "task_started",
+        "task_completed",
+        "task_failed",
+        "worker_created",
+        "worker_deleted",
+        "queue_status",
+        "all_tasks_completed",
+    ]
+    metadata: Optional[Dict[str, Any]] = None
+    timestamp: datetime = Field(
+        default_factory=lambda: datetime.now(timezone.utc)
+    )
+
+
+class WorkerCreatedEvent(WorkforceEventBase):
+    event_type: Literal["worker_created"] = "worker_created"
+    worker_id: str
+    worker_type: str
+    role: str
+
+
+class WorkerDeletedEvent(WorkforceEventBase):
+    event_type: Literal["worker_deleted"] = "worker_deleted"
+    worker_id: str
+    reason: Optional[str] = None
+
+
+class TaskDecomposedEvent(WorkforceEventBase):
+    event_type: Literal["task_decomposed"] = "task_decomposed"
+    parent_task_id: str
+    subtask_ids: List[str]
+
+
+class TaskCreatedEvent(WorkforceEventBase):
+    event_type: Literal["task_created"] = "task_created"
+    task_id: str
+    description: str
+    parent_task_id: Optional[str] = None
+    task_type: Optional[str] = None
+
+
+class TaskAssignedEvent(WorkforceEventBase):
+    event_type: Literal["task_assigned"] = "task_assigned"
+    task_id: str
+    worker_id: str
+    queue_time_seconds: Optional[float] = None
+    dependencies: Optional[List[str]] = None
+
+
+class TaskStartedEvent(WorkforceEventBase):
+    event_type: Literal["task_started"] = "task_started"
+    task_id: str
+    worker_id: str
+
+
+class TaskCompletedEvent(WorkforceEventBase):
+    event_type: Literal["task_completed"] = "task_completed"
+    task_id: str
+    worker_id: str
+    result_summary: Optional[str] = None
+    processing_time_seconds: Optional[float] = None
+    token_usage: Optional[Dict[str, int]] = None
+
+
+class TaskFailedEvent(WorkforceEventBase):
+    event_type: Literal["task_failed"] = "task_failed"
+    task_id: str
+    error_message: str
+    worker_id: Optional[str] = None
+
+
+class AllTasksCompletedEvent(WorkforceEventBase):
+    event_type: Literal["all_tasks_completed"] = "all_tasks_completed"
+
+
+class QueueStatusEvent(WorkforceEventBase):
+    event_type: Literal["queue_status"] = "queue_status"
+    queue_name: str
+    length: int
+    pending_task_ids: Optional[List[str]] = None
+    metadata: Optional[Dict[str, Any]] = None
+
+
+WorkforceEvent = Union[
+    TaskDecomposedEvent,
+    TaskCreatedEvent,
+    TaskAssignedEvent,
+    TaskStartedEvent,
+    TaskCompletedEvent,
+    TaskFailedEvent,
+    WorkerCreatedEvent,
+    WorkerDeletedEvent,
+    AllTasksCompletedEvent,
+    QueueStatusEvent,
+]