Fix bugs, clean code, bring everything up-to-date and standardize a bit more.

CF_Code.yaml

@@ -65,14 +65,13 @@ query_message_prompt_template:
     code_placeholder: "{{python_code}}"
   template_format: jinja2
 
+input_data_transformations: []
 input_keys:
   - "problem_description"
   - "input_description"
   - "output_description"
   - "io_examples_and_explanation"
 
-output_keys:
-  - "code"
 output_data_transformations:
   - _target_: flows.data_transformations.RegexFirstOccurrenceExtractor
     regex: '(?<=```python)([\s\S]*?)(?=```)'
@@ -82,3 +81,5 @@ output_data_transformations:
     strip: True
     assert_unique: True
     verbose: True
+output_keys:
+  - "code"

CF_CodeCollab.yaml

@@ -2,20 +2,24 @@ name: "CodeCollab_Flow"
 verbose: True
 description: "ToDO: add description"
 
+reset_generator_every_round: False
+reset_critic_every_round: True
+max_rounds: 2 # ToDO: To increase to 4
+early_exit_key: "end_of_interaction"
+
+input_data_transformations: []
 input_keys:
   - "problem_description"
   - "input_description"
   - "output_description"
   - "io_examples_and_explanation"
 
+output_data_transformations:
+  - _target_: flows.data_transformations.KeyRename
+    old_key2new_key:
+      raw_response.code: "code"
 output_keys:
   - "code"
-  - "end_of_interaction"
-
-reset_generator_every_round: False
-reset_critic_every_round: True
-max_rounds: 2 # ToDO: To increase to 4
-early_exit_key: "end_of_interaction"
 
 subflows_config:
   - _target_: flows.flow_verse.instantiate_flow
@@ -23,6 +27,7 @@ subflows_config:
     class_name: CF_Code
     overrides:
       name: "CodeGenerator"
+      model_name: "gpt-4"
       human_message_prompt_template:
         _target_: langchain.PromptTemplate
         template: |2-

CF_CodeDebug.yaml

@@ -7,12 +7,18 @@ reset_critic_every_round: True
 max_rounds: 2 # ToDo: To increase to 4
 early_exit_key: "end_of_interaction"
 
+input_data_transformations: []
 input_keys:
   - "problem_description"
   - "input_description"
   - "output_description"
   - "io_examples_and_explanation"
   - "public_tests_individual_io"
+
+output_data_transformations:
+  - _target_: flows.data_transformations.KeyRename
+    old_key2new_key:
+      raw_response.code: "code"
 output_keys:
   - "code"
 
@@ -25,7 +31,7 @@ subflows_config:
       model_name: "gpt-4"
       human_message_prompt_template:
         template: |2-
-          {{query}}
+          {{testing_results_summary}}
 
 
           Consider the problem statement, the last proposed solution, and its issue. Provide a corrected version of the code that solves the original problem and resolves the issue, without any explanation, in the following format:
@@ -33,13 +39,10 @@ subflows_config:
           {{code_placeholder}}
           ```
         input_variables:
-          - query
+          - testing_results_summary
         partial_variables:
           code_placeholder: "{{python_code}}"
-      input_data_transformations:
-        - _target_: flows.data_transformations.KeyRename
-          old_key2new_key:
-            "test_results_summary": "query"
+      default_human_input_key: "testing_results_summary"
       output_data_transformations:
         - _target_: flows.data_transformations.RegexFirstOccurrenceExtractor
           regex: '(?<=```python)([\s\S]*?)(?=```)'
@@ -56,7 +59,6 @@ subflows_config:
       output_keys:
         - "code"
         - "end_of_interaction"
-
   - _target_: flows.flow_verse.instantiate_flow
     repository_id: ${oc.env:CC_FLOWS}
     class_name: CF_CodeTesting

CF_CodeReflect.yaml

@@ -7,11 +7,17 @@ reset_critic_every_round: True
 max_rounds: 2 # ToDo: To increase to 4
 early_exit_key: "end_of_interaction"
 
+input_data_transformations: []
 input_keys:
   - "problem_description"
   - "input_description"
   - "output_description"
   - "io_examples_and_explanation"
+
+output_data_transformations:
+  - _target_: flows.data_transformations.KeyRename
+    old_key2new_key:
+      raw_response.code: "code"
 output_keys:
   - "code"
 
@@ -25,12 +31,11 @@ subflows_config:
       input_data_transformations:
         - _target_: flows.data_transformations.KeyRename
           old_key2new_key:
-            "code_reflect_message": "query"
+            code_reflect_message: "query"
       output_data_transformations:
         - _target_: flows.data_transformations.RegexFirstOccurrenceExtractor
           regex: '(?<=```python)([\s\S]*?)(?=```)'
           regex_fallback: '(?<=```)([\s\S]*?)(?=```)'
-          input_key: "raw_response"
           output_key: "code"
           strip: True
           assert_unique: True

CF_CodeTesting.yaml

@@ -1,16 +1,19 @@
 description: "ToDo: add description"
+
+input_data_transformations: []
 input_keys:
   - "code"
   - "public_tests_individual_io"
+
 output_keys:
   - "all_tests_passed"
-  - "tests_results_summary"
+  - "testing_results_summary"
 output_data_transformations:
   - _target_: CC_flows.src.data_transformations.CorrectnessFlag  # ToDo: This import style would not work if the flow is synced in the current implementation (the outer directory is a hash and not the name; figure out how to do the import robustly; e.g., using relative imports)
-    input_key: "public_tests_results"  # ToDo: Add support for nested keys and update this to raw_response.public_tests_results?
+    input_key: "raw_response.public_tests_results"
     output_key: "all_tests_passed"
-  - _target_: CC_flows.src.data_transformations.TestResultsSummaryGeneration
-    output_key: "tests_results_summary"
+  - _target_: CC_flows.src.data_transformations.TestingResultsSummaryGeneration
+    output_key: "testing_results_summary"
 
     single_test_error_message: True
 

CF_CodeWithPlan.yaml

@@ -71,6 +71,7 @@ query_message_prompt_template:
     code_placeholder: "{{python_code}}"
   template_format: jinja2
 
+input_data_transformations: []
 input_keys:
   - "problem_description"
   - "input_description"
@@ -78,9 +79,6 @@ input_keys:
   - "io_examples_and_explanation"
   - "plan"
 
-output_keys:
-  - "code"
-
 output_data_transformations:
   - _target_: flows.data_transformations.RegexFirstOccurrenceExtractor
     regex: '(?<=```python)([\s\S]*?)(?=```)'
@@ -89,3 +87,5 @@ output_data_transformations:
     strip: True
     assert_unique: True
     verbose: True
+output_keys:
+  - "code"

CF_Plan.yaml

@@ -13,6 +13,25 @@ generation_parameters:
     frequency_penalty: 0
     presence_penalty: 0
 
+input_data_transformations: []
+input_keys:
+  - "problem_description"
+  - "input_description"
+  - "output_description"
+  - "io_examples_and_explanation"
+
+output_data_transformations:
+  - _target_: flows.data_transformations.RegexFirstOccurrenceExtractor
+    regex: '(?<=Conceptual solution)([\s\S]*?)(?=\n\n# [A-Z]|\Z)'
+    regex_fallback:
+      - '(?<=Conceptual solution:)([\s\S]*?)(?=\n\n# [A-Z]|\Z)'
+    output_key: "plan"
+    strip: True
+    assert_unique: True
+    verbose: True
+output_keys:
+  - "plan"
+
 system_message_prompt_template:
   _target_: langchain.PromptTemplate
   template: |2-
@@ -66,21 +85,3 @@ query_message_prompt_template:
     plan_placeholder: "{{conceptual_solution}}"
   template_format: jinja2
 
-input_keys:
-  - "problem_description"
-  - "input_description"
-  - "output_description"
-  - "io_examples_and_explanation"
-
-output_keys:
-  - "plan"
-
-output_data_transformations:
-  - _target_: flows.data_transformations.RegexFirstOccurrenceExtractor
-    regex: '(?<=Conceptual solution)([\s\S]*?)(?=\n\n# [A-Z]|\Z)'
-    regex_fallback:
-      - '(?<=Conceptual solution:)([\s\S]*?)(?=\n\n# [A-Z]|\Z)'
-    output_key: "plan"
-    strip: True
-    assert_unique: True
-    verbose: True

CF_PlanCollab.yaml

@@ -2,20 +2,24 @@ name: "PlanCollab_Flow"
 verbose: True
 description: "ToDO: add description"
 
+reset_generator_every_round: False
+reset_critic_every_round: True
+max_rounds: 2 # ToDo: To increase to 4
+early_exit_key: "end_of_interaction"
+
+input_data_transformations: []
 input_keys:
   - "problem_description"
   - "input_description"
   - "output_description"
   - "io_examples_and_explanation"
 
+output_data_transformations:
+  - _target_: flows.data_transformations.KeyRename
+    old_key2new_key:
+      raw_response.plan: "plan"
 output_keys:
   - "plan"
-  - "end_of_interaction"
-
-reset_generator_every_round: False
-reset_critic_every_round: True
-max_rounds: 2 # ToDo: To increase to 4
-early_exit_key: "end_of_interaction"
 
 subflows_config:
   - _target_: flows.flow_verse.instantiate_flow

CF_PlanCollab_Code.yaml

@@ -1,12 +1,19 @@
-name: "PlanCollab_Code"
+name: "PlanCollab_Code_Flow"
 description: "ToDO: add description"
 
+early_exit_key: null
+
+input_data_transformations: []
 input_keys:
   - "problem_description"
   - "input_description"
   - "output_description"
   - "io_examples_and_explanation"
 
+output_data_transformations:
+  - _target_: flows.data_transformations.KeyRename
+    old_key2new_key:
+      raw_response.code: "code"
 output_keys:
   - "code"
 

CF_PlanReflect.yaml

@@ -7,12 +7,17 @@ reset_critic_every_round: True
 max_rounds: 2 # ToDo: increase to 4
 early_exit_key: "end_of_interaction"
 
+input_data_transformations: []
 input_keys:
   - "problem_description"
   - "input_description"
   - "output_description"
   - "io_examples_and_explanation"
 
+output_data_transformations:
+  - _target_: flows.data_transformations.KeyRename
+    old_key2new_key:
+      raw_response.plan: "plan"
 output_keys:
   - "plan"
 
@@ -21,31 +26,30 @@ subflows_config:
     repository_id: ${oc.env:CC_FLOWS}
     class_name: CF_Plan
     overrides:
-      name: "PlanGenerator"
+      name: "PlanGenerator_Flow"
       model_name: "gpt-4"
-      response_annotators:
-        end_of_interaction_annotator:
-          _target_: flows.message_annotators.EndOfInteraction
+      input_data_transformations:
+        - _target_: flows.data_transformations.KeyRename
+          old_key2new_key:
+            plan_reflect_message: "query"
+      output_data_transformations:
+        - _target_: flows.data_transformations.RegexFirstOccurrenceExtractor
+          regex: '(?<=Conceptual solution)([\s\S]*?)(?=\n\n# [A-Z]|\Z)'
+          regex_fallback:
+            - '(?<=Conceptual solution:)([\s\S]*?)(?=\n\n# [A-Z]|\Z)'
+          output_key: "plan"
+          strip: True
+          assert_unique: True
+          verbose: True
+        - _target_: flows.data_transformations.EndOfInteraction
           end_of_interaction_string: "Final answer"
-          key: "end_of_interaction"
+          output_key: "end_of_interaction"
           verbose: True
       output_keys:
         - "plan"
         - "end_of_interaction"
+
   - _target_: flows.flow_verse.instantiate_flow
     repository_id: ${oc.env:CC_FLOWS}
-    class_name: CF_Reflect
-    overrides:
-      name: "PlanFixedReplyCritic"
-      description: "ToDo: Add description"
-      input_keys:
-        - "plan"
-      output_keys:
-        - "query"
-      fixed_reply: |2-
-        Consider the problem statement and the last proposed solution. Are you sure that the solution is provided in the requested format, and crucially, solves the problem?
-        If that is not the case, provide the corrected version of the conceptual solution in the following format:
-        # Conceptual solution
-        {{conceptual_solution}}
-        otherwise, reply:
-        "Final answer."
+    class_name: FixedReply_PlanReflect
+

CF_Plan_Code.yaml

@@ -3,7 +3,19 @@ description: "ToDO: add description"
 
 early_exit_key: null
 
+input_data_transformations: []
+input_keys:
+  - "problem_description"
+  - "input_description"
+  - "output_description"
+  - "io_examples_and_explanation"
 
+output_data_transformations:
+  - _target_: flows.data_transformations.KeyRename
+    old_key2new_key:
+      raw_response.code: "code"
+output_keys:
+  - "code"
 
 subflows_config:
   - _target_: flows.flow_verse.instantiate_flow
@@ -16,12 +28,3 @@ subflows_config:
     class_name: CF_CodeWithPlan
     overrides:
       model_name: "gpt-4"
-
-input_keys:
-  - "problem_description"
-  - "input_description"
-  - "output_description"
-  - "io_examples_and_explanation"
-
-output_keys:
-  - "code"

FixedReply_CodeReflect.yaml

@@ -2,24 +2,24 @@ name: "CodeReflectCritic"
 description: "A flow that prompts the user to reflect on their code and provide a corrected version if necessary."
 verbose: True
 
-input_keys:
-  - "code"
+input_data_transformations: []
+input_keys: []
 
-output_keys:
-  - "code_reflect_message"
 output_data_transformations:
   - _target_: flows.data_transformations.KeyRename
     old_key2new_key:
       raw_response: "code_reflect_message"
+output_keys:
+  - "code_reflect_message"
 
 fixed_reply: |2-
-    Consider the problem statement and the last proposed solution. Are you sure that the solution is provided in the requested format, and crucially, solves the problem?
-    If that is not the case, provide the corrected version of the code in the following format:
-    ```python
-    {{python_code}}
-    ```
-    otherwise, reply:
-    "Final answer."
+  Consider the problem statement and the last proposed solution. Are you sure that the solution is provided in the requested format, and crucially, solves the problem?
+  If that is not the case, provide the corrected version of the code in the following format:
+  ```python
+  {{python_code}}
+  ```
+  otherwise, reply:
+  "Final answer."
 
 
 

FixedReply_PlanReflect.py

@@ -0,0 +1,6 @@
+from flows.base_flows import FixedReplyAtomicFlow
+
+
+class FixedReply_PlanReflect(FixedReplyAtomicFlow):
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)

FixedReply_PlanReflect.yaml

@@ -0,0 +1,21 @@
+name: "PlanReflectCritic"
+description: "A flow that prompts the user to reflect on their conceptual solution and provide a corrected version if necessary."
+verbose: True
+
+input_data_transformations: []
+input_keys: []
+
+output_data_transformations:
+  - _target_: flows.data_transformations.KeyRename
+    old_key2new_key:
+      raw_response: "plan_reflect_message"
+output_keys:
+  - "plan_reflect_message"
+
+fixed_reply: |2-
+  Consider the problem statement and the last proposed solution. Are you sure that the solution is provided in the requested format, and crucially, solves the problem?
+  If that is not the case, provide the corrected version of the conceptual solution in the following format:
+  # Conceptual solution
+  {{conceptual_solution}}
+  otherwise, reply:
+  "Final answer."

__init__.py

@@ -15,6 +15,7 @@ from .CF_CodeWithPlan import CF_CodeWithPlan
 from .CF_Plan_Code import CF_Plan_Code
 
 # cf-plan_reflect-code
+from .FixedReply_PlanReflect import FixedReply_PlanReflect
 from .CF_PlanReflect import CF_PlanReflect
 from .CF_PlanReflect_Code import CF_PlanReflect_Code
 
@@ -26,6 +27,3 @@ from .CF_PlanCollab_Code import CF_PlanCollab_Code
 # cf-code_debug
 from .CF_CodeTesting import CF_CodeTesting
 from .CF_CodeDebug import CF_CodeDebug
-
-# from .CF_Debug import CF_Debug
-# from .CF_CodeTestDebug import CF_CodeTestDebug

src/data_transformations/__init__.py

@@ -1,2 +1,2 @@
 from .correctness_flag import CorrectnessFlag
-from .test_results_summary_generation import TestResultsSummaryGeneration
+from .testing_results_summary_generation import TestingResultsSummaryGeneration

src/data_transformations/correctness_flag.py

@@ -9,7 +9,6 @@ class CorrectnessFlag(DataTransformation):
         self.input_key = input_key
 
     def __call__(self, data_dict: Dict[str, Any], **kwargs) -> Dict[str, Any]:
-        test_data = data_dict["raw_response"]
-        all_tests_passed = all([test_result["status"] for test_result in test_data[self.input_key]])
+        all_tests_passed = all([test_result["status"] for test_result in data_dict[self.input_key]])
         data_dict[self.output_key] = all_tests_passed
         return data_dict

src/data_transformations/{test_results_summary_generation.py → testing_results_summary_generation.py}

@@ -3,9 +3,10 @@ from typing import Dict, Any
 import jinja2
 
 from flows.data_transformations.abstract import DataTransformation
+from flows.utils.general_helpers import unflatten_dict
 
 
-class TestResultsSummaryGeneration(DataTransformation):
+class TestingResultsSummaryGeneration(DataTransformation):
     def __init__(self, output_key, **kwargs):
         super().__init__(output_key)
         self.params = kwargs
@@ -13,10 +14,10 @@ class TestResultsSummaryGeneration(DataTransformation):
     def __call__(self, data_dict: Dict[str, Any], **kwargs) -> Dict[str, Any]:
         if data_dict["all_tests_passed"]:
             # the execution did not result in any errors
-            data_dict["test_results_summary"] = self.params["no_error_template"]
+            data_dict[self.output_key] = self.params["no_error_template"]
             return data_dict
 
-        test_data = data_dict["raw_response"]
+        test_data = unflatten_dict(data_dict)["raw_response"]
 
         if not test_data["compilation_status"]:
             # compilation error occurred
@@ -98,5 +99,5 @@ class TestResultsSummaryGeneration(DataTransformation):
                         )
 
                     message_content = self.params["tests_separator"].join(parts)
-        data_dict["test_results_summary"] = message_content
+        data_dict[self.output_key] = message_content
         return data_dict