Merge pull request #406 from PrefectHQ/temp-none

Set temperature to None by default

Author: jlowin

src/controlflow/agents/agent.py

@@ -83,7 +83,7 @@ class Agent(ControlFlowModel, abc.ABC):
         default=False,
         description="If True, the agent is given tools for interacting with a human user.",
     )
-    memories: list[Memory] | list[AsyncMemory] = Field(
+    memories: list[Union[Memory, AsyncMemory]] = Field(
         default=[],
         description="A list of memory modules for the agent to use.",
     )
@@ -345,7 +345,7 @@ def _run_model(
 
         create_markdown_artifact(
             markdown=f"""
-{response.content or '(No content)'}
+{response.content or "(No content)"}
 
 #### Payload
 ```json
@@ -409,7 +409,7 @@ async def _run_model_async(
 
         create_markdown_artifact(
             markdown=f"""
-{response.content or '(No content)'}
+{response.content or "(No content)"}
 
 #### Payload
 ```json

src/controlflow/defaults.py

@@ -40,9 +40,7 @@ class Defaults(ControlFlowModel):
     model: Optional[Any]
     history: History
     agent: Agent
-    memory_provider: (
-        Optional[Union[MemoryProvider, str]] | Optional[Union[AsyncMemoryProvider, str]]
-    )
+    memory_provider: Optional[Union[MemoryProvider, AsyncMemoryProvider, str]]
 
     # add more defaults here
     def __repr__(self) -> str:

src/controlflow/llm/models.py

@@ -52,6 +52,8 @@ def get_model(
                 "To use Google as an LLM provider, please install the `langchain_google_genai` package."
             )
         cls = ChatGoogleGenerativeAI
+        if temperature is None:
+            temperature = 0.7
     elif provider == "groq":
         try:
             from langchain_groq import ChatGroq
@@ -60,6 +62,8 @@ def get_model(
                 "To use Groq as an LLM provider, please install the `langchain_groq` package."
             )
         cls = ChatGroq
+        if temperature is None:
+            temperature = 0.7
     elif provider == "ollama":
         try:
             from langchain_ollama import ChatOllama
@@ -73,7 +77,9 @@ def get_model(
             f"Could not load provider `{provider}` automatically. Please provide the LLM class manually."
         )
 
-    return cls(model=model, temperature=temperature, **kwargs)
+    if temperature is not None:
+        kwargs["temperature"] = temperature
+    return cls(model=model, **kwargs)
 
 
 def _get_initial_default_model() -> BaseChatModel:

src/controlflow/orchestration/orchestrator.py

@@ -188,7 +188,7 @@ def get_tools(self) -> list[Tool]:
         tools = as_tools(tools)
         return tools
 
-    def get_memories(self) -> list[Memory] | list[AsyncMemory]:
+    def get_memories(self) -> list[Union[Memory, AsyncMemory]]:
         memories = set()
 
         memories.update(self.agent.memories)
@@ -525,7 +525,7 @@ def compile_prompt(self) -> str:
         ]
 
         prompt = "\n\n".join([p for p in prompts if p])
-        logger.debug(f"{'='*10}\nCompiled prompt: {prompt}\n{'='*10}")
+        logger.debug(f"{'=' * 10}\nCompiled prompt: {prompt}\n{'=' * 10}")
         return prompt
 
     def compile_messages(self) -> list[BaseMessage]:

src/controlflow/orchestration/prompt_templates.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List, Optional
+from typing import Any, Dict, List, Optional, Union
 
 from pydantic import model_validator
 
@@ -98,7 +98,7 @@ def should_render(self) -> bool:
 
 class MemoryTemplate(Template):
     template_path: str = "memories.jinja"
-    memories: list[Memory] | list[AsyncMemory]
+    memories: list[Union[Memory, AsyncMemory]]
 
     def should_render(self) -> bool:
         return bool(self.memories)

src/controlflow/settings.py

@@ -109,7 +109,9 @@ def _validate_pretty_print_agent_events(cls, data: dict) -> dict:
         default="openai/gpt-4o",
         description="The default LLM model for agents.",
     )
-    llm_temperature: float = Field(0.7, description="The temperature for LLM sampling.")
+    llm_temperature: Union[float, None] = Field(
+        None, description="The temperature for LLM sampling."
+    )
     max_input_tokens: int = Field(
         100_000, description="The maximum number of tokens to send to an LLM."
     )

src/controlflow/tasks/task.py

@@ -75,7 +75,7 @@ def __getitem__(self, item):
         return self.root[item]
 
     def __repr__(self) -> str:
-        return f'Labels: {", ".join(self.root)}'
+        return f"Labels: {', '.join(self.root)}"
 
 
 class TaskStatus(Enum):
@@ -162,7 +162,7 @@ class Task(ControlFlowModel):
         description="Agents that are allowed to mark this task as complete. If None, all agents are allowed.",
     )
     interactive: bool = False
-    memories: list[Memory] | list[AsyncMemory] = Field(
+    memories: list[Union[Memory, AsyncMemory]] = Field(
         default=[],
         description="A list of memory modules for the task to use.",
     )