from typing import Any

from sqlalchemy.orm import Session

from app.modules.ai_agent.adapters import get_adapter
from app.modules.audit.schemas import AuditLogCreate
from app.modules.audit.service import AuditService


class AIService:
    """Coordinate AI provider calls and audit logging."""

    def __init__(self, db: Session):
        self.db = db
        self.audit = AuditService(db)

    def ask(
        self,
        prompt: str,
        context: dict[str, Any] | None = None,
        actor: str = "api",
        source: str = "api",
    ) -> dict[str, Any]:
        adapter = get_adapter()
        result = adapter.ask(prompt, context or {})
        response = {
            "provider": adapter.provider_name,
            "answer": result["answer"],
            "raw": result.get("raw", {}),
        }
        self.audit.log(
            AuditLogCreate(
                actor=actor,
                source=source,
                action="ai.ask",
                target_type="ai",
                risk_level="medium",
                request_payload={"prompt": prompt, "context": context or {}},
                response_payload=response,
            )
        )
        return response

    def draft_policy(
        self,
        title: str,
        policy_type: str,
        requirements: list[str],
        actor: str,
    ) -> dict[str, Any]:
        prompt = (
            f"Draft a company policy in Chinese. Title: {title}. Type: {policy_type}. "
            "Include purpose, scope, roles, process, approval rules, audit rules, and KPI linkage. "
            f"Requirements: {requirements}"
        )
        return self.ask(prompt, actor=actor, source="policy")

    def draft_investment_research(
        self,
        symbol_or_topic: str,
        risk_preference: str,
        actor: str,
    ) -> dict[str, Any]:
        prompt = (
            "Create an investment research memo in Chinese. Do not give direct trading "
            "instructions. Include thesis, risks, data needed, position sizing constraints, "
            "and human approval checklist. "
            f"Topic: {symbol_or_topic}. Risk preference: {risk_preference}."
        )
        return self.ask(prompt, actor=actor, source="investment")