fix: always call sample when a strategy is provided

mellea/stdlib/funcs.py

@@ -144,9 +144,9 @@ async def _act(
             "Must provide a SamplingStrategy when return_sampling_results==True"
         )
 
-    # if there is no reason to sample, just generate from the context.
-    if strategy is None or requirements is None or len(requirements) == 0:
-        if strategy is None and requirements is not None:
+    if strategy is None:
+        # Only use the strategy if one is provided. Add a warning if requirements were passed in though.
+        if requirements is not None and len(requirements) >= 0:
             FancyLogger.get_logger().warning(
                 "Calling the function with NO strategy BUT requirements. No requirement is being checked!"
             )
@@ -394,6 +394,13 @@ async def _validate(
     # Turn a solitary requirement in to a list of requirements, and then reqify if needed.
     reqs = [reqs] if not isinstance(reqs, list) else reqs
     reqs = [Requirement(req) if type(req) is str else req for req in reqs]
+
+    if len(reqs) == 0:
+        FancyLogger().get_logger().warning(
+            "Calling validate with no requirements. This is a no-op."
+        )
+        return []
+
     if output is None:
         validation_target_ctx = context
     else:

mellea/stdlib/sampling/base.py

@@ -86,7 +86,7 @@ async def sample(
         action: Component,
         context: Context,
         backend: Backend,
-        requirements: list[Requirement],
+        requirements: list[Requirement] | None,
         *,
         validation_ctx: Context | None = None,
         format: type[BaseModelSubclass] | None = None,
@@ -123,7 +123,7 @@ async def sample(
         show_progress = show_progress and flog.getEffectiveLevel() <= FancyLogger.INFO
 
         reqs = []
-        # global requirements supersede local requirements (global requiremenst can be defined by user)
+        # global requirements supersede local requirements (global requirements can be defined by user)
         # Todo: re-evaluate if this makes sense
         if self.requirements is not None:
             reqs += self.requirements

mellea/stdlib/sampling/best_of_n.py

@@ -22,7 +22,7 @@ async def sample(
         action: Component,
         context: Context,
         backend: Backend,
-        requirements: list[Requirement],
+        requirements: list[Requirement] | None,
         *,
         validation_ctx: Context | None = None,
         format: type[BaseModelSubclass] | None = None,

mellea/stdlib/sampling/types.py

@@ -81,7 +81,7 @@ async def sample(
         action: Component,
         context: Context,
         backend: Backend,
-        requirements: list[Requirement],
+        requirements: list[Requirement] | None,
         *,
         validation_ctx: Context | None = None,
         format: type[BaseModelSubclass] | None = None,

test/stdlib_basics/test_vision_ollama.py

@@ -61,7 +61,7 @@ def test_image_block_construction_from_pil(pil_image: Image.Image):
 
 def test_image_block_in_instruction(m_session: MelleaSession, pil_image: Image.Image, gh_run: int):
     image_block = ImageBlock.from_pil_image(pil_image)
-    instr = m_session.instruct("Is this image mainly blue? Answer yes or no.", images=[image_block])
+    instr = m_session.instruct("Is this image mainly blue? Answer yes or no.", images=[image_block], strategy=None)
     assert isinstance(instr, ModelOutputThunk)
 
     # if not on GH

test/stdlib_basics/test_vision_openai.py

@@ -65,7 +65,7 @@ def test_image_block_construction_from_pil(pil_image: Image.Image):
 
 def test_image_block_in_instruction(m_session: MelleaSession, pil_image: Image.Image, gh_run: int):
     image_block = ImageBlock.from_pil_image(pil_image)
-    instr = m_session.instruct("Is this image mainly blue? Answer yes or no.", images=[image_block])
+    instr = m_session.instruct("Is this image mainly blue? Answer yes or no.", images=[image_block], strategy=None)
     assert isinstance(instr, ModelOutputThunk)
 
     # if not on GH