reflexion cot/react math strat skeleton

agential/cog/strategies/reflexion/math.py

@@ -0,0 +1,72 @@
+"""Reflexion Agent strategies for Math."""
+
+from typing import Any, Dict, List, Tuple
+from langchain_core.language_models.chat_models import BaseChatModel
+from agential.cog.strategies.reflexion.base import (
+    ReflexionCoTBaseStrategy,
+    ReflexionReActBaseStrategy,
+)
+
+
+class ReflexionCoTMathStrategy(ReflexionCoTBaseStrategy):
+    def __init__(self, llm: BaseChatModel) -> None:
+        super().__init__(llm)
+
+    def generate(self, *args: Any, **kwargs: Any) -> str:
+        return super().generate(*args, **kwargs)
+
+    def generate_action(self, question: str, examples: str, reflections: str, prompt: str, additional_keys: Dict[str, str]) -> Tuple[str, str]:
+        return super().generate_action(question, examples, reflections, prompt, additional_keys)
+
+    def generate_observation(self, action_type: str, query: str, key: str) -> Tuple[bool | str]:
+        return super().generate_observation(action_type, query, key)
+
+    def create_output_dict(self, thought: str, action_type: str, obs: str, is_correct: bool, reflections: List[str]) -> Dict[str, Any]:
+        return super().create_output_dict(thought, action_type, obs, is_correct, reflections)
+
+    def halting_condition(self, idx: int, key: str, **kwargs: Any) -> bool:
+        return super().halting_condition(idx, key, **kwargs)
+
+    def reset(self, *args: Any, **kwargs: Any) -> None:
+        return super().reset(*args, **kwargs)
+
+    def reflect(self, reflect_strategy: str, question: str, examples: str, prompt: str, additional_keys: Dict[str, str]) -> Tuple[List[str] | str]:
+        return super().reflect(reflect_strategy, question, examples, prompt, additional_keys)
+
+    def reflect_condition(self, idx: int, reflect_strategy: str | None, key: str) -> bool:
+        return super().reflect_condition(idx, reflect_strategy, key)
+
+
+class ReflexionReActMathStrategy(ReflexionReActBaseStrategy):
+    def __init__(self, llm: BaseChatModel) -> None:
+        super().__init__(llm)
+
+    def generate(self, *args: Any, **kwargs: Any) -> str:
+        return super().generate(*args, **kwargs)
+
+    def generate_action(self, question: str, examples: str, reflections: str, prompt: str, additional_keys: Dict[str, str], **kwargs: Any) -> Tuple[str]:
+        return super().generate_action(question, examples, reflections, prompt, additional_keys, **kwargs)
+
+    def generate_observation(self, step_idx: int, action_type: str, query: str, key: str) -> Tuple[bool | str]:
+        return super().generate_observation(step_idx, action_type, query, key)
+
+    def create_output_dict(self, react_out: List[Dict[str, Any]], reflections: List[str]) -> Dict[str, Any]:
+        return super().create_output_dict(react_out, reflections)
+
+    def react_create_output_dict(self, thought: str, action_type: str, query: str, obs: str, is_correct: bool) -> Dict[str, str]:
+        return super().react_create_output_dict(thought, action_type, query, obs, is_correct)
+
+    def halting_condition(self, idx: int, key: str, **kwargs: Any) -> bool:
+        return super().halting_condition(idx, key, **kwargs)
+
+    def react_halting_condition(self, step_idx: int, question: str, examples: str, reflections: str, prompt: str, additional_keys: Dict[str, str], **kwargs: Any) -> bool:
+        return super().react_halting_condition(step_idx, question, examples, reflections, prompt, additional_keys, **kwargs)
+
+    def reset(self, *args: Any, **kwargs: Any) -> None:
+        return super().reset(*args, **kwargs)
+
+    def reflect(self, reflect_strategy: str, question: str, examples: str, prompt: str, additional_keys: Dict[str, str]) -> Tuple[List[str] | str]:
+        return super().reflect(reflect_strategy, question, examples, prompt, additional_keys)
+
+    def reflect_condition(self, step_idx: int, reflect_strategy: str | None, question: str, examples: str, key: str, prompt: str, additional_keys: Dict[str, str], **kwargs: Dict[str, str]) -> bool:
+        return super().reflect_condition(step_idx, reflect_strategy, question, examples, key, prompt, additional_keys, **kwargs)