palimpzest 0.6.1__py3-none-any.whl → 0.6.3__py3-none-any.whl

palimpzest/prompts/prompt_factory.py

@@ -1,4 +1,5 @@
 """This file contains factory methods which return template prompts and return messages for chat payloads."""
+
 import base64
 import json
 from string import Formatter
@@ -82,6 +83,7 @@ from palimpzest.prompts.util_phrases import (
 
 class PromptFactory:
     """Factory class for generating prompts for the Generator given the input(s)."""
+
     BASE_SYSTEM_PROMPT_MAP = {
         PromptStrategy.COT_BOOL: COT_BOOL_BASE_SYSTEM_PROMPT,
         PromptStrategy.COT_BOOL_IMAGE: COT_BOOL_BASE_SYSTEM_PROMPT,
@@ -92,6 +94,7 @@ class PromptFactory:
         PromptStrategy.COT_QA_IMAGE_CRITIC: None,
         PromptStrategy.COT_QA_IMAGE_REFINE: None,
         PromptStrategy.COT_MOA_PROPOSER: COT_MOA_PROPOSER_BASE_SYSTEM_PROMPT,
+        PromptStrategy.COT_MOA_PROPOSER_IMAGE: COT_MOA_PROPOSER_BASE_SYSTEM_PROMPT,
         PromptStrategy.COT_MOA_AGG: COT_MOA_AGG_BASE_SYSTEM_PROMPT,
     }
     BASE_USER_PROMPT_MAP = {
@@ -104,6 +107,7 @@ class PromptFactory:
         PromptStrategy.COT_QA_IMAGE_CRITIC: BASE_CRITIQUE_PROMPT,
         PromptStrategy.COT_QA_IMAGE_REFINE: BASE_REFINEMENT_PROMPT,
         PromptStrategy.COT_MOA_PROPOSER: COT_MOA_PROPOSER_BASE_USER_PROMPT,
+        PromptStrategy.COT_MOA_PROPOSER_IMAGE: COT_MOA_PROPOSER_BASE_USER_PROMPT,
         PromptStrategy.COT_MOA_AGG: COT_MOA_AGG_BASE_USER_PROMPT,
     }
 
@@ -148,13 +152,13 @@ class PromptFactory:
                 longest_field_name, longest_field_length = sorted_fields[0]
 
                 # trim the field
-                context_factor =  MIXTRAL_LLAMA_CONTEXT_TOKENS_LIMIT / (total_context_len * TOKENS_PER_CHARACTER)
+                context_factor = MIXTRAL_LLAMA_CONTEXT_TOKENS_LIMIT / (total_context_len * TOKENS_PER_CHARACTER)
                 keep_frac_idx = int(longest_field_length * context_factor)
                 context[longest_field_name] = context[longest_field_name][:keep_frac_idx]
 
                 # update total context length
                 total_context_len = len(json.dumps(context, indent=2))
-        
+
         return json.dumps(context, indent=2)
 
     def _get_input_fields(self, candidate: DataRecord, **kwargs) -> list[str]:
@@ -201,7 +205,11 @@ class PromptFactory:
         """
         output_fields_desc = ""
         output_schema: Schema = kwargs.get("output_schema")
-        if self.prompt_strategy.is_cot_qa_prompt():
+        if (
+            self.prompt_strategy.is_cot_qa_prompt()
+            or self.prompt_strategy.is_moa_proposer_prompt()
+            or self.prompt_strategy.is_moa_aggregator_prompt()
+        ):
             assert output_schema is not None, "Output schema must be provided for convert prompts."
 
             field_desc_map = output_schema.field_desc_map()
@@ -230,14 +238,16 @@ class PromptFactory:
 
         Args:
             kwargs: The keyword arguments provided by the user.
-        
+
         Returns:
             str | None: The original output.
         """
         original_output = kwargs.get("original_output")
         if self.prompt_strategy.is_critic_prompt() or self.prompt_strategy.is_refine_prompt():
-            assert original_output is not None, "Original output must be provided for critique and refinement operations."
-        
+            assert original_output is not None, (
+                "Original output must be provided for critique and refinement operations."
+            )
+
         return original_output
 
     def _get_critique_output(self, **kwargs) -> str | None:
@@ -246,7 +256,7 @@ class PromptFactory:
 
         Args:
             kwargs: The keyword arguments provided by the user.
-        
+
         Returns:
             str | None: The critique output.
         """
@@ -259,10 +269,10 @@ class PromptFactory:
     def _get_model_responses(self, **kwargs) -> str | None:
         """
         Returns the model responses for the mixture-of-agents aggregation operation.
-        
+
         Args:
             kwargs: The keyword arguments provided by the user.
-            
+
         Returns:
             str | None: The model responses.
         """
@@ -314,9 +324,7 @@ class PromptFactory:
         critique_criteria = None
         if self.prompt_strategy.is_critic_prompt():
             critique_criteria = (
-                COT_QA_IMAGE_CRITIQUE_CRITERIA
-                if self.prompt_strategy.is_image_prompt()
-                else COT_QA_CRITIQUE_CRITERIA
+                COT_QA_IMAGE_CRITIQUE_CRITERIA if self.prompt_strategy.is_image_prompt() else COT_QA_CRITIQUE_CRITERIA
             )
 
         return critique_criteria
@@ -467,16 +475,18 @@ class PromptFactory:
 
         return prompt_strategy_to_example_answer.get(self.prompt_strategy)
 
-    def _get_all_format_kwargs(self, candidate: DataRecord, input_fields: list[str], output_fields: list[str], **kwargs) -> dict:
+    def _get_all_format_kwargs(
+        self, candidate: DataRecord, input_fields: list[str], output_fields: list[str], **kwargs
+    ) -> dict:
         """
         Returns a dictionary containing all the format kwargs for templating the prompts.
-        
+
         Args:
             candidate (DataRecord): The input record.
             input_fields (list[str]): The input fields.
             output_fields (list[str]): The output fields.
             kwargs: The keyword arguments provided by the user.
-        
+
         Returns:
             dict: The dictionary containing all the format kwargs.
         """
@@ -517,7 +527,7 @@ class PromptFactory:
         Args:
             candidate (DataRecord): The input record.
             input_fields (list[str]): The list of input fields.
-        
+
         Returns:
             list[dict]: The image messages for the chat payload.
         """
@@ -529,15 +539,19 @@ class PromptFactory:
 
             # image filepath (or list of image filepaths)
             if isinstance(field_type, ImageFilepathField):
-                with open(field_value, 'rb') as f:
-                    base64_image_str = base64.b64encode(f.read()).decode('utf-8')
-                image_messages.append({"role": "user", "type": "image", "content": f"data:image/jpeg;base64,{base64_image_str}"})
+                with open(field_value, "rb") as f:
+                    base64_image_str = base64.b64encode(f.read()).decode("utf-8")
+                image_messages.append(
+                    {"role": "user", "type": "image", "content": f"data:image/jpeg;base64,{base64_image_str}"}
+                )
 
             elif hasattr(field_type, "element_type") and issubclass(field_type.element_type, ImageFilepathField):
                 for image_filepath in field_value:
-                    with open(image_filepath, 'rb') as f:
-                        base64_image_str = base64.b64encode(f.read()).decode('utf-8')
-                    image_messages.append({"role": "user", "type": "image", "content": f"data:image/jpeg;base64,{base64_image_str}"})
+                    with open(image_filepath, "rb") as f:
+                        base64_image_str = base64.b64encode(f.read()).decode("utf-8")
+                    image_messages.append(
+                        {"role": "user", "type": "image", "content": f"data:image/jpeg;base64,{base64_image_str}"}
+                    )
 
             # image url (or list of image urls)
             elif isinstance(field_type, ImageURLField):
@@ -550,12 +564,16 @@ class PromptFactory:
             # pre-encoded images (or list of pre-encoded images)
             elif isinstance(field_type, ImageBase64Field):
                 base64_image_str = field_value.decode("utf-8")
-                image_messages.append({"role": "user", "type": "image", "content": f"data:image/jpeg;base64,{base64_image_str}"})
+                image_messages.append(
+                    {"role": "user", "type": "image", "content": f"data:image/jpeg;base64,{base64_image_str}"}
+                )
 
             elif hasattr(field_type, "element_type") and issubclass(field_type.element_type, ImageBase64Field):
                 for base64_image in field_value:
                     base64_image_str = base64_image.decode("utf-8")
-                    image_messages.append({"role": "user", "type": "image", "content": f"data:image/jpeg;base64,{base64_image_str}"})
+                    image_messages.append(
+                        {"role": "user", "type": "image", "content": f"data:image/jpeg;base64,{base64_image_str}"}
+                    )
 
         return image_messages
 
@@ -595,15 +613,15 @@ class PromptFactory:
 
         # get any image messages for the chat payload (will be an empty list if this is not an image prompt)
         image_messages = (
-            self._create_image_messages(candidate, input_fields)
-            if self.prompt_strategy.is_image_prompt()
-            else []
+            self._create_image_messages(candidate, input_fields) if self.prompt_strategy.is_image_prompt() else []
         )
 
         # get any original messages for critique and refinement operations
         original_messages = kwargs.get("original_messages")
         if self.prompt_strategy.is_critic_prompt() or self.prompt_strategy.is_refine_prompt():
-            assert original_messages is not None, "Original messages must be provided for critique and refinement operations."
+            assert original_messages is not None, (
+                "Original messages must be provided for critique and refinement operations."
+            )
 
         # construct the user messages based on the prompt strategy
         user_messages = []
@@ -661,17 +679,19 @@ class PromptFactory:
                     f"Input fields: {input_fields}\n"
                 )
             assert fields_check, err_msg
-        
+
         # build set of format kwargs
         format_kwargs = {
-            field_name: "<bytes>" if isinstance(candidate.get_field_type(field_name), BytesField) else candidate[field_name]
+            field_name: "<bytes>"
+            if isinstance(candidate.get_field_type(field_name), BytesField)
+            else candidate[field_name]
             for field_name in input_fields
         }
 
         # split prompt on <<image-placeholder>> if it exists
         if "<<image-placeholder>>" in user_prompt:
             raise NotImplementedError("Image prompts are not yet supported.")
-        
+
         prompt_sections = user_prompt.split("<<image-placeholder>>")
         messages = [{"role": "user", "type": "text", "content": prompt_sections[0].format(**format_kwargs)}]
 
@@ -686,7 +706,7 @@ class PromptFactory:
     def create_messages(self, candidate: DataRecord, output_fields: list[str], **kwargs) -> list[dict]:
         """
         Creates the messages for the chat payload based on the prompt strategy.
-        
+
         Each message will be a dictionary with the following format:
         {
             "role": "user" | "system",

palimpzest/query/optimizer/optimizer.py

@@ -219,6 +219,9 @@ class Optimizer:
             allow_conventional_query=self.allow_conventional_query,
             allow_code_synth=self.allow_code_synth,
             allow_token_reduction=self.allow_token_reduction,
+            allow_rag_reduction=self.allow_rag_reduction,
+            allow_mixtures=self.allow_mixtures,
+            allow_critic=self.allow_critic,
             optimization_strategy_type=self.optimization_strategy_type,
             use_final_op_quality=self.use_final_op_quality,
         )

palimpzest/query/processor/query_processor_factory.py

@@ -123,6 +123,9 @@ class QueryProcessorFactory:
             allow_conventional_query=config.allow_conventional_query,
             allow_code_synth=config.allow_code_synth,
             allow_token_reduction=config.allow_token_reduction,
+            allow_rag_reduction=config.allow_rag_reduction,
+            allow_mixtures=config.allow_mixtures,
+            allow_critic=config.allow_critic,
             optimization_strategy_type=optimizer_strategy,
             use_final_op_quality=config.use_final_op_quality
         )

{palimpzest-0.6.1.dist-info → palimpzest-0.6.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: palimpzest
-Version: 0.6.1
+Version: 0.6.3
 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-0.6.1.dist-info → palimpzest-0.6.3.dist-info}/RECORD

@@ -20,7 +20,7 @@ palimpzest/prompts/critique_and_refine_convert_prompts.py,sha256=WoXExBxQ7twswd9
 palimpzest/prompts/filter_prompts.py,sha256=iQjn-39h3L0E5wng_UPgAXRHrP1ok329TXpOgZ6Wn1w,2372
 palimpzest/prompts/moa_aggregator_convert_prompts.py,sha256=BQRrtGdr53PTqvXzmFh8kfQ_w9KoKw-zTtmdo-8RFjo,2887
 palimpzest/prompts/moa_proposer_convert_prompts.py,sha256=d_hOh0-0m6HWBDAxUu7W3WyQtSTlUvqio3nzpnX2bxM,3642
-palimpzest/prompts/prompt_factory.py,sha256=VzZNH9kblFXYn4YKVKudJ21Y5Q-3tL6ZgFmNhBNTGjQ,31921
+palimpzest/prompts/prompt_factory.py,sha256=w6h4eQnaGVxv49A8hlAzij7yP8GaKl6tgSQbt3HXFDU,32350
 palimpzest/prompts/util_phrases.py,sha256=NWrcHfjJyiOY16Jyt7R50moVnlJDyvSBZ9kBqyX2WQo,751
 palimpzest/query/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 palimpzest/query/execution/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -46,7 +46,7 @@ palimpzest/query/operators/scan.py,sha256=z6wUVxuhr5VqPIeUxb3hxhkaljKpDc_exzGMe4
 palimpzest/query/operators/token_reduction_convert.py,sha256=yy9GYMPt-LQxPdwIgVyhCb9hi_8FRorGU8XqK_3jq9g,8513
 palimpzest/query/optimizer/__init__.py,sha256=pl1co0dCwDZkAQ-0oiwT81GjvB0Oc59WiwmmYF8k73s,3109
 palimpzest/query/optimizer/cost_model.py,sha256=zSK2Nsya96pR5Zh67cr_O5q0qtPa08--Tchn0cYvE58,44837
-palimpzest/query/optimizer/optimizer.py,sha256=lBfNYgbyyE_0bdZCGnz9oicyG3gFUdkRnzcyJ31_36o,20644
+palimpzest/query/optimizer/optimizer.py,sha256=EQ5yKb1oa0EdqiK2R8Ey7PPhC-YjXZquA-XWIWq3fso,20794
 palimpzest/query/optimizer/optimizer_strategy.py,sha256=-1xx_cviSJw6PH8XiQQK9qe4YPnAmxZEAhNVKdxRgH4,12894
 palimpzest/query/optimizer/plan.py,sha256=xlWB3sY5qDac3o6IHoWcuGK5Azv-4C2_zKKx4PzxEh4,5768
 palimpzest/query/optimizer/primitives.py,sha256=ikaX8YcDM3IrxKt98OX-mYujRYQtdMlDgsFKyjchMMA,4061
@@ -57,7 +57,7 @@ palimpzest/query/processor/config.py,sha256=kOhBxAZ3OeDDlQ2qMII1i2EorFpSSQbEFVFr
 palimpzest/query/processor/mab_sentinel_processor.py,sha256=xShPVW8ejhNR_wQ8ofPF2hL7iGj8tlFhsc2wiiIPb9Y,45600
 palimpzest/query/processor/nosentinel_processor.py,sha256=alPq1tnZvqxCSO5LYRCjlF4CB4v7NbzH_BHB-DSuehI,26478
 palimpzest/query/processor/query_processor.py,sha256=aR0OBmaZZt4_KSBjHy_KCT6pIBI8WTfT8TTcIkgPBt4,11109
-palimpzest/query/processor/query_processor_factory.py,sha256=663_V-AJK0VsBZNwgnqYu84g0rmtKf-U-xJWnps3XWs,8239
+palimpzest/query/processor/query_processor_factory.py,sha256=OjmSDVAwWcpwLbi7b97pBkV2sJNo4Jv3hkxD94VyOlI,8395
 palimpzest/query/processor/random_sampling_sentinel_processor.py,sha256=VFC0HFY3OLxAjpdmBt41K0rw8C_1ylECFjbRYsui3rU,30440
 palimpzest/query/processor/streaming_processor.py,sha256=4-XvgAjUTnO3Dgdxm9VSw4udREjNWTW526Rggy5Do7s,6501
 palimpzest/schemabuilder/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -80,8 +80,8 @@ palimpzest/utils/progress.py,sha256=GYmPUBdG7xmqbqj1UiSNP-pWZKmRMLX797MBgrOPugM,
 palimpzest/utils/sandbox.py,sha256=Ge96gmzqeOGlNkMCG9A95_PB8wRQbvTFua136of8FcA,6465
 palimpzest/utils/token_reduction_helpers.py,sha256=Ob95PcqCsbGLiBdQ-4YQsWGWRppb2hvQyt0gi1fzL-Y,3855
 palimpzest/utils/udfs.py,sha256=LjHic54B1az-rKgNLur0wOpaz2ko_UodjLEJrazkxvY,1854
-palimpzest-0.6.1.dist-info/LICENSE,sha256=5GUlHy9lr-Py9kvV38FF1m3yy3NqM18fefuE9wkWumo,1079
-palimpzest-0.6.1.dist-info/METADATA,sha256=VxPI4-vfq3Fm3l3PjxTpdHGbDclIQNHo1Ag1enfAyMU,7837
-palimpzest-0.6.1.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-palimpzest-0.6.1.dist-info/top_level.txt,sha256=raV06dJUgohefUn3ZyJS2uqp_Y76EOLA9Y2e_fxt8Ew,11
-palimpzest-0.6.1.dist-info/RECORD,,
+palimpzest-0.6.3.dist-info/LICENSE,sha256=5GUlHy9lr-Py9kvV38FF1m3yy3NqM18fefuE9wkWumo,1079
+palimpzest-0.6.3.dist-info/METADATA,sha256=FxQ3szdZPEMH-y0zkOrTPEAN2jLVPOcVS7A9am-Hy6o,7837
+palimpzest-0.6.3.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
+palimpzest-0.6.3.dist-info/top_level.txt,sha256=raV06dJUgohefUn3ZyJS2uqp_Y76EOLA9Y2e_fxt8Ew,11
+palimpzest-0.6.3.dist-info/RECORD,,