Explainability

# Chain-of-Thought for explainable reasoning
from openai import OpenAI

client = OpenAI()

def get_explainable_answer(question: str) -> dict:
    """Get an answer with step-by-step reasoning"""
    
    prompt = f"""Answer the following question. 
Before giving your final answer, explain your reasoning step-by-step.

Question: {question}

Think through this step by step:
1. First, identify the key information needed
2. Then, analyze each relevant factor
3. Finally, synthesize into a conclusion

Format your response as:
REASONING:
[Your step-by-step thinking]

ANSWER:
[Your final answer]

CONFIDENCE: [High/Medium/Low]
"""

    response = client.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "user", "content": prompt}]
    )
    
    output = response.choices[0].message.content
    
    # Parse response into components
    return {
        "reasoning": extract_section(output, "REASONING"),
        "answer": extract_section(output, "ANSWER"),
        "confidence": extract_section(output, "CONFIDENCE")
    }

# RAG with source citations
class ExplainableRAG:
    def __init__(self, retriever, llm):
        self.retriever = retriever
        self.llm = llm
    
    def query(self, question: str) -> dict:
        # Retrieve relevant documents
        docs = self.retriever.search(question, top_k=5)
        
        # Build context with source markers
        context = ""
        for i, doc in enumerate(docs):
            context += f"[Source {i+1}]: {doc.content}\n\n"
        
        prompt = f"""Based on the following sources, answer the question.
Cite your sources using [Source N] format.

Sources:
{context}

Question: {question}

Answer with citations:"""
        
        answer = self.llm.generate(prompt)
        
        return {
            "answer": answer,
            "sources": [
                {
                    "id": i + 1,
                    "title": doc.metadata.get("title"),
                    "url": doc.metadata.get("url"),
                    "relevance_score": doc.score
                }
                for i, doc in enumerate(docs)
            ]
        }