"""
Deduplication Router - 使用 Embedding 或 LLM 去重描述

提供 API 端點將相似的創新描述分組，幫助識別重複的想法。
支援兩種方法：
- Embedding: 快速向量相似度比較
- LLM: 精準語意判斷（較慢但更準確）
"""

import logging
from fastapi import APIRouter, HTTPException

from ..models.schemas import DeduplicationRequest, DeduplicationResult, DeduplicationMethod
from ..services.embedding_service import embedding_service
from ..services.llm_deduplication_service import llm_deduplication_service

logger = logging.getLogger(__name__)

router = APIRouter(prefix="/api/deduplication", tags=["deduplication"])


@router.post("/deduplicate", response_model=DeduplicationResult)
async def deduplicate_descriptions(request: DeduplicationRequest) -> DeduplicationResult:
    """
    去重描述

    支援兩種方法：
    - embedding: 使用向量相似度（快速）
    - llm: 使用 LLM 成對比較（精準但較慢）

    Args:
        request: 去重請求，包含描述列表、方法選擇和相關參數

    Returns:
        DeduplicationResult: 去重結果，包含分組資訊

    Raises:
        HTTPException: 如果去重處理失敗
    """
    method = request.method
    logger.info(f"Deduplication request: {len(request.descriptions)} descriptions, method={method.value}, threshold={request.similarity_threshold}")

    if not request.descriptions:
        return DeduplicationResult(
            total_input=0,
            total_groups=0,
            total_duplicates=0,
            groups=[],
            threshold_used=request.similarity_threshold,
            method_used=method,
            model_used=request.model or ("nomic-embed-text" if method == DeduplicationMethod.EMBEDDING else "qwen3:4b")
        )

    try:
        if method == DeduplicationMethod.EMBEDDING:
            # 使用 Embedding 相似度去重
            result = await embedding_service.deduplicate(
                descriptions=request.descriptions,
                threshold=request.similarity_threshold,
                model=request.model
            )
        else:
            # 使用 LLM 成對比較去重
            result = await llm_deduplication_service.deduplicate(
                descriptions=request.descriptions,
                model=request.model
            )
        return result
    except ValueError as e:
        logger.error(f"Deduplication failed: {e}")
        raise HTTPException(status_code=500, detail=str(e))
    except Exception as e:
        logger.error(f"Unexpected error during deduplication: {e}")
        raise HTTPException(status_code=500, detail=f"Deduplication failed: {str(e)}")


@router.get("/models")
async def list_embedding_models():
    """
    列出可用的 Embedding 模型

    Returns:
        dict: 可用模型列表和建議的預設模型
    """
    return {
        "default": "nomic-embed-text",
        "available": [
            {"name": "nomic-embed-text", "description": "Fast and efficient embedding model"},
            {"name": "mxbai-embed-large", "description": "High quality embeddings"},
            {"name": "all-minilm", "description": "Lightweight embedding model"},
        ],
        "note": "Run 'ollama pull <model>' to install a model"
    }