palimpzest 0.7.9__tar.gz → 0.7.11__tar.gz

{palimpzest-0.7.9/src/palimpzest.egg-info → palimpzest-0.7.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: palimpzest
-Version: 0.7.9
+Version: 0.7.11
 Summary: Palimpzest is a system which enables anyone to process AI-powered analytical queries simply by defining them in a declarative language
 Author-email: MIT DSG Semantic Management Lab <michjc@csail.mit.edu>
 Project-URL: homepage, https://palimpzest.org

{palimpzest-0.7.9 → palimpzest-0.7.11}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "palimpzest"
-version = "0.7.9"
+version = "0.7.11"
 description = "Palimpzest is a system which enables anyone to process AI-powered analytical queries simply by defining them in a declarative language"
 readme = "README.md"
 requires-python = ">=3.8"

{palimpzest-0.7.9 → palimpzest-0.7.11}/src/palimpzest/query/generators/generators.py

@@ -215,14 +215,15 @@ class BaseGenerator(Generic[ContextType, InputType], ABC):
 
             # TODO: wrap non-list outputs in a list if expected output is a list
 
-            # common error: if the output is a singleton list which contains a list, but the expected field type
+            # common error for one-to-one: if the output is a singleton list which contains a list, but the expected field type
             # is a list of strings, or a list of floats, i.e. not a list of lists; then extract the inner list
-            for field, field_type in fields.items():
-                answer = field_answers[field]
-                field_type_is_not_list_of_lists = isinstance(field_type, ListField) and not issubclass(field_type.element_type, ListField)
-                answer_is_list_of_lists = isinstance(answer, list) and len(answer) == 1 and isinstance(answer[0], list)
-                if field_type_is_not_list_of_lists and answer_is_list_of_lists:
-                    field_answers[field] = answer[0]
+            if self.cardinality == Cardinality.ONE_TO_ONE:
+                for field, field_type in fields.items():
+                    answer = field_answers[field]
+                    field_type_is_not_list_of_lists = isinstance(field_type, ListField) and not issubclass(field_type.element_type, ListField)
+                    answer_is_list_of_lists = isinstance(answer, list) and len(answer) == 1 and isinstance(answer[0], list)
+                    if field_type_is_not_list_of_lists and answer_is_list_of_lists:
+                        field_answers[field] = answer[0]
 
             # prepare the field answers to match the expected output and return
             return self._prepare_field_answers(field_answers, fields)
@@ -416,6 +417,8 @@ class BaseGenerator(Generic[ContextType, InputType], ABC):
                 prompt += message["content"] + "\n" if message["type"] == "text" else "<image>\n"
         logger.debug(f"PROMPT:\n{prompt}")
         logger.debug(Fore.GREEN + f"{completion_text}\n" + Style.RESET_ALL)
+        print(f"PROMPT:\n{prompt}")
+        print(Fore.GREEN + f"{completion_text}\n" + Style.RESET_ALL)
 
         # parse reasoning
         reasoning = None

{palimpzest-0.7.9 → palimpzest-0.7.11}/src/palimpzest/query/optimizer/rules.py

@@ -323,7 +323,8 @@ class LLMConvertBondedRule(ImplementationRule):
             first_criteria = model in pure_vision_models and not is_image_conversion
             second_criteria = model in pure_text_models and is_image_conversion
             third_criteria = model.is_llama_model() and model.is_vision_model() and (num_image_fields > 1 or list_image_field)
-            if first_criteria or second_criteria or third_criteria:
+            fourth_criteria = model.is_embedding_model()
+            if first_criteria or second_criteria or third_criteria or fourth_criteria:
                 continue
 
             # construct multi-expression
@@ -470,7 +471,7 @@ class RAGConvertRule(ImplementationRule):
         physical_expressions = []
         for model in physical_op_params["available_models"]:
             # skip this model if this is a pure image model
-            if model in pure_vision_models:
+            if model in pure_vision_models or model.is_embedding_model():
                 continue
 
             for num_chunks_per_field in cls.num_chunks_per_fields:
@@ -666,7 +667,8 @@ class CriticAndRefineConvertRule(ImplementationRule):
             first_criteria = model in pure_vision_models and not is_image_conversion
             second_criteria = model in pure_text_models and is_image_conversion
             third_criteria = model.is_llama_model() and model.is_vision_model() and (num_image_fields > 1 or list_image_field)
-            if first_criteria or second_criteria or third_criteria:
+            fourth_criteria = model.is_embedding_model()
+            if first_criteria or second_criteria or third_criteria or fourth_criteria:
                 continue
 
             models.append(model)
@@ -746,7 +748,7 @@ class SplitConvertRule(ImplementationRule):
         physical_expressions = []
         for model in physical_op_params["available_models"]:
             # skip this model if this is a pure image model
-            if model in pure_vision_models:
+            if model in pure_vision_models or model.is_embedding_model():
                 continue
 
             for min_size_to_chunk in cls.min_size_to_chunk:
@@ -934,7 +936,8 @@ class LLMFilterRule(ImplementationRule):
             first_criteria = model in pure_vision_models and not is_image_filter
             second_criteria = model in pure_text_models and is_image_filter
             third_criteria = model.is_llama_model() and model.is_vision_model() and (num_image_fields > 1 or list_image_field)
-            if first_criteria or second_criteria or third_criteria:
+            fourth_criteria = model.is_embedding_model()
+            if first_criteria or second_criteria or third_criteria or fourth_criteria:
                 continue
 
             # construct multi-expression

{palimpzest-0.7.9 → palimpzest-0.7.11}/src/palimpzest/utils/model_helpers.py

@@ -25,13 +25,17 @@ def get_vision_models() -> list[Model]:
     return models
 
 
-def get_models(include_vision: bool = False) -> list[Model]:
+def get_models(include_vision: bool = False, include_embedding: bool = False) -> list[Model]:
     """
     Return the set of models which the system has access to based on the set environment variables.
     """
     models = []
     if os.getenv("OPENAI_API_KEY") is not None:
         openai_models = [model for model in Model if model.is_openai_model()]
+        if not include_embedding:
+            openai_models = [
+                model for model in openai_models if not model.is_embedding_model()
+            ]
         models.extend(openai_models)
 
     if os.getenv("TOGETHER_API_KEY") is not None:
@@ -40,6 +44,10 @@ def get_models(include_vision: bool = False) -> list[Model]:
             together_models = [
                 model for model in together_models if not model.is_vision_model()
             ]
+        if not include_embedding:
+            together_models = [
+                model for model in together_models if not model.is_embedding_model()
+            ]
         models.extend(together_models)
 
     if include_vision:

{palimpzest-0.7.9 → palimpzest-0.7.11/src/palimpzest.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: palimpzest
-Version: 0.7.9
+Version: 0.7.11
 Summary: Palimpzest is a system which enables anyone to process AI-powered analytical queries simply by defining them in a declarative language
 Author-email: MIT DSG Semantic Management Lab <michjc@csail.mit.edu>
 Project-URL: homepage, https://palimpzest.org