app.py

from fastapi import FastAPI, File, UploadFile
from fastapi.responses import JSONResponse
import uvicorn
import os
from predictor import SentenceExtractor  # 保证导入规范文件名 `predictor.py`

# 创建 FastAPI 应用
app = FastAPI()

# 初始化 SentenceExtractor
extractor = SentenceExtractor(
    eval_keywords_path="evaluation_keywords2.json",  # 相对路径将被转换为绝对路径
    model_path="distilled_model.onnx"
)

@app.get("/")
async def root():
    return {"message": "API is running. Use POST /evaluate to upload files."}

@app.get("/health")
async def health():
    try:
        # 暴露关键运行状态，便于部署环境自检
        return JSONResponse(content={
            "model_loaded": getattr(extractor, "model_loaded", False),
            "model_path_abs": getattr(extractor, "model_path_abs", None),
            "model_sha256": getattr(extractor, "model_sha256", None),
            "providers": getattr(extractor, "providers", None),
            "tokenizer_loaded": getattr(extractor, "tokenizer_loaded", None),
            "last_tokenizer_error": getattr(extractor, "last_tokenizer_error", None),
            "aggregation_mode": extractor.aggregation_mode,
            "min_sentence_char_len": extractor.min_sentence_char_len,
            "merge_leading_punct": extractor.merge_leading_punct,
        })
    except Exception as e:
        return JSONResponse(content={"error": str(e)}, status_code=500)

@app.post("/evaluate")
async def evaluate_file(file: UploadFile = File(...)):
    try:
        # 读取上传文件内容
        content = await file.read()
        text = content.decode("utf-8", errors="ignore")

        # 调用 extractor 进行文本分析
        result = extractor.extract(text)

        # 格式化输出结果
        formatted_result = {
            "综合评分": result["comprehensive_grade"],
            "积极词语评价数": result["positive_word_count"],
            "消极词语评价数": result["negative_word_count"],
            "中性词语评价数": result["neutral_word_count"],
            "句子评分": []
        }
        for i, item in enumerate(result["scored_sentences"], 1):
            source = item.get('source', '?')
            reason = item.get('reason') or item.get('last_tokenizer_error')
            suffix = f" ({source})"
            if source == 'rule' and reason:
                # 将回退原因直接拼接到可见文本，便于客户端看到具体错误
                if len(reason) > 120:
                    reason = reason[:120] + '...'
                suffix += f" - reason: {reason}"
            formatted_result["句子评分"].append({
                f"句子{i}": f"{item['sentence']} - {item['grade']}{suffix}"
            })

        # 附加调试信息便于客户端确认
        formatted_result["_debug"] = result.get("debug", {})

        return JSONResponse(content=formatted_result)

    except Exception as e:
        return JSONResponse(content={"error": str(e)}, status_code=500)

if __name__ == "__main__":
    port = int(os.getenv("PORT", 7860))
    uvicorn.run(app, host="0.0.0.0", port=port)