Remove retry_strategy in LM and handle no-docstring functions in ReAct (#1725)

* Remove retry_strategy in LM and handle no-docstring functions in ReAct

* exponential_backoff_retry tests

Author: okhat

dspy/clients/lm.py

@@ -39,7 +39,7 @@ def __init__(
         cache: bool = True,
         launch_kwargs: Optional[Dict[str, Any]] = None,
         callbacks: Optional[List[BaseCallback]] = None,
-        num_retries: int = 8,
+        num_retries: int = 3,
         **kwargs,
     ):
         """
@@ -186,7 +186,6 @@ def litellm_completion(request, num_retries: int, cache={"no-cache": True, "no-s
     kwargs = ujson.loads(request)
     return litellm.completion(
         num_retries=num_retries,
-        retry_strategy="exponential_backoff_retry",
         cache=cache,
         **kwargs,
     )
@@ -223,7 +222,6 @@ def litellm_text_completion(request, num_retries: int, cache={"no-cache": True,
         api_base=api_base,
         prompt=prompt,
         num_retries=num_retries,
-        retry_strategy="exponential_backoff_retry",
         **kwargs,
     )
 

dspy/predict/react.py

@@ -12,7 +12,7 @@ def __init__(self, func: Callable, name: str = None, desc: str = None, args: dic
         annotations_func = func if inspect.isfunction(func) else func.__call__
         self.func = func
         self.name = name or getattr(func, '__name__', type(func).__name__)
-        self.desc = desc or getattr(func, '__doc__', None) or getattr(annotations_func, '__doc__', "No description")
+        self.desc = desc or getattr(func, '__doc__', None) or getattr(annotations_func, '__doc__', "")
         self.args = {
             k: v.schema() if isinstance((origin := get_origin(v) or v), type) and issubclass(origin, BaseModel)
             else get_annotation_name(v)
@@ -50,10 +50,10 @@ def __init__(self, signature, tools: list[Callable], max_iters=5):
         tools["finish"] = Tool(func=lambda **kwargs: "Completed.", name="finish", desc=finish_desc, args=finish_args)
 
         for idx, tool in enumerate(tools.values()):
-            desc = tool.desc.replace("\n", "  ")
             args = tool.args if hasattr(tool, 'args') else str({tool.input_variable: str})
-            desc = f"whose description is <desc>{desc}</desc>. It takes arguments {args} in JSON format."
-            instr.append(f"({idx+1}) {tool.name}, {desc}")
+            desc = (f", whose description is <desc>{tool.desc}</desc>." if tool.desc else ".").replace('\n', "  ")
+            desc += f" It takes arguments {args} in JSON format."
+            instr.append(f"({idx+1}) {tool.name}{desc}")
 
         signature_ = (
             dspy.Signature({**signature.input_fields}, "\n".join(instr))

tests/clients/test_lm.py

@@ -11,7 +11,7 @@ def test_lm_chat_respects_max_retries():
 
     assert litellm_completion_api.call_count == 1
     assert litellm_completion_api.call_args[1]["max_retries"] == 17
-    assert litellm_completion_api.call_args[1]["retry_strategy"] == "exponential_backoff_retry"
+    # assert litellm_completion_api.call_args[1]["retry_strategy"] == "exponential_backoff_retry"
 
 
 def test_lm_completions_respects_max_retries():
@@ -22,4 +22,4 @@ def test_lm_completions_respects_max_retries():
 
     assert litellm_completion_api.call_count == 1
     assert litellm_completion_api.call_args[1]["max_retries"] == 17
-    assert litellm_completion_api.call_args[1]["retry_strategy"] == "exponential_backoff_retry"
+    # assert litellm_completion_api.call_args[1]["retry_strategy"] == "exponential_backoff_retry"