more

Author: nqyy

mbodied/agents/backends/anthropic_backend.py

@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import os
 from typing import Any, List
 
 import anthropic
@@ -80,7 +81,7 @@ def __init__(self, api_key: str | None, client: anthropic.Anthropic | None = Non
             client: An optional client for the Anthropic service.
             kwargs: Additional keyword arguments.
         """
-        self.api_key = api_key
+        self.api_key = api_key or os.getenv("ANTHROPIC_API_KEY")
         self.client = client
 
         self.model = kwargs.pop("model", self.DEFAULT_MODEL)
@@ -117,13 +118,6 @@ def predict(
         )
         return completion.content[0].text
 
-    async def async_predict(
-        self, message: Message, context: List[Message] | None = None, model: Any | None = None
-    ) -> str:
-        """Asynchronously predict the next message in the conversation."""
-        # For now, we'll use the synchronous method since Anthropic doesn't provide an async API
-        return self.predict(message, context, model)
-
     def stream(
         self, message: Message, context: List[Message] = None, model: str = "claude-3-5-sonnet-20240620", **kwargs
     ):

mbodied/agents/backends/gemini_backend.py

@@ -105,7 +105,7 @@ def __init__(
             client: An optional client for the Gemini service.
             **kwargs: Additional keyword arguments.
         """
-        self.api_key = api_key or os.getenv("GOOGLE_API_KEY") or os.getenv("MBODI_API_KEY")
+        self.api_key = api_key or os.getenv("GOOGLE_API_KEY")
         self.client = client
 
         self.model = kwargs.pop("model", self.DEFAULT_MODEL)

mbodied/agents/backends/openai_backend.py

@@ -100,7 +100,7 @@ def __init__(
             aclient: Whether to use the asynchronous client.
             **kwargs: Additional keyword arguments.
         """
-        self.api_key = api_key or os.getenv("OPENAI_API_KEY") or os.getenv("MBODI_API_KEY")
+        self.api_key = api_key or os.getenv("OPENAI_API_KEY")
         self.client = client
         if self.client is None:
             from openai import AsyncOpenAI, OpenAI
@@ -137,7 +137,7 @@ def predict(
             **kwargs: Additional keyword arguments.
 
         Returns:
-            str | tuple[str, List[ToolCall]]: 
+            str | tuple[str, List[ToolCall]]:
                 When tools are not provided: Just the text response
                 When tools are provided: A tuple of (text_response, tool_calls)
         """
@@ -163,12 +163,7 @@ def predict(
         return completion.choices[0].message.content
 
     def stream(
-        self, 
-        message: Message, 
-        context: List[Message] = None, 
-        model: str = "gpt-4o", 
-        tools: List[Tool] = None,
-        **kwargs
+        self, message: Message, context: List[Message] = None, model: str = "gpt-4o", tools: List[Tool] = None, **kwargs
     ):
         """Streams a completion for the given messages using the OpenAI API standard.
 
@@ -178,7 +173,7 @@ def stream(
             model: The model to be used for the completion.
             tools: Optional list of tools (function calls) available to the model.
             **kwargs: Additional keyword arguments.
-            
+
         Yields:
             When tools is None:
                 str: Content delta chunks
@@ -196,7 +191,7 @@ def stream(
             tools=tools,
             **kwargs,
         )
-        
+
         if not tools:
             for chunk in stream:
                 yield chunk.choices[0].delta.content or ""
@@ -208,12 +203,7 @@ def stream(
                 yield content, tool_calls
 
     async def astream(
-        self, 
-        message: Message, 
-        context: List[Message] = None, 
-        model: str = "gpt-4o", 
-        tools: List[Tool] = None,
-        **kwargs
+        self, message: Message, context: List[Message] = None, model: str = "gpt-4o", tools: List[Tool] = None, **kwargs
     ):
         """Streams a completion asynchronously for the given messages using the OpenAI API standard.
 
@@ -223,7 +213,7 @@ async def astream(
             model: The model to be used for the completion.
             tools: Optional list of tools (function calls) available to the model.
             **kwargs: Additional keyword arguments.
-            
+
         Yields:
             When tools is None:
                 str: Content delta chunks
@@ -243,7 +233,7 @@ async def astream(
             tools=tools,
             **kwargs,
         )
-        
+
         if not tools:
             async for chunk in stream:
                 yield chunk.choices[0].delta.content or ""

mbodied/agents/language/language_agent.py

@@ -86,14 +86,19 @@ def __getitem__(self, key):
             raise IndexError("Invalid index")
 
 
-def make_context_list(context: list[str | Image | Message] | Image | str | Message | None) -> List[Message]:
+def make_context_list(
+    context: list[str | Image | Message] | Image | str | Message | None, model_src: str
+) -> List[Message]:
     """Convert the context to a list of messages."""
     if isinstance(context, list):
         return [Message(content=c) if not isinstance(c, Message) else c for c in context]
     if isinstance(context, Message):
         return [context]
     if isinstance(context, str | Image):
-        return [Message(role="user", content=[context]), Message(role="assistant", content="Understood.")]
+        if model_src == "openai":
+            return [Message(role="system", content=[context])]
+        else:
+            return [Message(role="user", content=[context]), Message(role="assistant", content="Understood.")]
     return []
 
 
@@ -150,7 +155,7 @@ def __init__(
         | DirectoryPath
         | NewPath = "openai",
         context: list | Image | str | Message = None,
-        api_key: str | None = os.getenv("OPENAI_API_KEY"),
+        api_key: str | None = None,
         model_kwargs: dict = None,
         recorder: Literal["default", "omit"] | str = "omit",
         recorder_kwargs: dict = None,
@@ -199,7 +204,7 @@ def __init__(
             api_key=api_key,
         )
 
-        self.context = make_context_list(context)
+        self.context = make_context_list(context, model_src)
 
     def forget_last(self) -> Message:
         """Forget the last message in the context."""

mbodied/agents/language/utils.py

@@ -104,8 +104,8 @@ def function_to_tool(func: Callable) -> Tool:
     return Tool.model_validate(tool_data)
 
 
-def main():
-    """Example usage of function_to_tool utility"""
+def main() -> None:
+    """Example usage of function_to_tool utility."""
     from mbodied.agents.language import LanguageAgent
 
     def move_by(x: float, y: float, z: float, speed: float = 1.0) -> bool: