feat: feedback enrichment preparation

cognee/infrastructure/llm/prompts/feedback_reaction_prompt.txt

@@ -0,0 +1,14 @@
+A question was previously answered, but the answer received negative feedback.
+Please reconsider and improve the response.
+
+Question: {question}
+Context originally used: {context}
+Previous answer: {wrong_answer}
+Feedback on that answer: {negative_feedback}
+
+Task: Provide a better response. The new answer should be short and direct.
+Then explain briefly why this answer is better.
+
+Format your reply as:
+Answer: <improved answer>
+Explanation: <short explanation>

cognee/infrastructure/llm/prompts/feedback_report_prompt.txt

@@ -0,0 +1,13 @@
+Write a concise, stand-alone paragraph that explains the correct answer to the question below.
+The paragraph should read naturally on its own, providing all necessary context and reasoning
+so the answer is clear and well-supported.
+
+Question: {question}
+Correct answer: {improved_answer}
+Supporting context: {new_context}
+
+Your paragraph should:
+- First sentence clearly states the correct answer as a full sentence
+- Remainder flows from first sentence and provides explanation based on context
+- Use simple, direct language that is easy to follow
+- Use shorter sentences, no long-winded explanations

cognee/infrastructure/llm/prompts/feedback_user_context_prompt.txt

@@ -0,0 +1,5 @@
+Question: {question}
+Context: {context}
+
+Provide a one paragraph human readable summary of this interaction context,
+listing all the relevant facts and information in a simple and direct way.

cognee/tasks/feedback/models.py

@@ -0,0 +1,20 @@
+from typing import List, Optional, Union
+from uuid import UUID
+
+from cognee.infrastructure.engine import DataPoint
+from cognee.modules.engine.models import Entity
+from cognee.tasks.temporal_graph.models import Event
+
+
+class FeedbackEnrichment(DataPoint):
+    """Minimal DataPoint for feedback enrichment that works with extract_graph_from_data."""
+
+    text: str
+    contains: Optional[List[Union[Entity, Event]]] = None
+    metadata: dict = {"index_fields": ["text"]}
+
+    question: str
+    original_answer: str
+    improved_answer: str
+    feedback_id: UUID
+    interaction_id: UUID

cognee/tasks/feedback/utils.py

@@ -0,0 +1,57 @@
+from cognee.modules.retrieval.graph_completion_retriever import GraphCompletionRetriever
+from cognee.modules.retrieval.graph_completion_cot_retriever import GraphCompletionCotRetriever
+from cognee.modules.retrieval.graph_completion_context_extension_retriever import (
+    GraphCompletionContextExtensionRetriever,
+)
+from cognee.shared.logging_utils import get_logger
+
+
+logger = get_logger("feedback_utils")
+
+
+def create_retriever(
+    retriever_name: str = "graph_completion_cot",
+    top_k: int = 20,
+    user_prompt_path: str = "graph_context_for_question.txt",
+    system_prompt_path: str = "answer_simple_question.txt",
+):
+    """Factory for retriever instances with configurable top_k and prompt paths."""
+    if retriever_name == "graph_completion":
+        return GraphCompletionRetriever(
+            top_k=top_k,
+            save_interaction=False,
+            user_prompt_path=user_prompt_path,
+            system_prompt_path=system_prompt_path,
+        )
+    if retriever_name == "graph_completion_cot":
+        return GraphCompletionCotRetriever(
+            top_k=top_k,
+            save_interaction=False,
+            user_prompt_path=user_prompt_path,
+            system_prompt_path=system_prompt_path,
+        )
+    if retriever_name == "graph_completion_context_extension":
+        return GraphCompletionContextExtensionRetriever(
+            top_k=top_k,
+            save_interaction=False,
+            user_prompt_path=user_prompt_path,
+            system_prompt_path=system_prompt_path,
+        )
+    logger.warning(
+        "Unknown retriever, defaulting to graph_completion_cot", retriever=retriever_name
+    )
+    return GraphCompletionCotRetriever(
+        top_k=top_k,
+        save_interaction=False,
+        user_prompt_path=user_prompt_path,
+        system_prompt_path=system_prompt_path,
+    )
+
+
+def filter_negative_feedback(feedback_nodes):
+    """Filter for negative sentiment feedback using precise sentiment classification."""
+    return [
+        (node_id, props)
+        for node_id, props in feedback_nodes
+        if (props.get("sentiment", "").casefold() == "negative" or props.get("score", 0) < 0)
+    ]