matrix-show/backend/app/services/script_parser.py

import json
import logging
from typing import Optional

from app.services.hermes_client import hermes_client

logger = logging.getLogger(__name__)


class ScriptParser:
    async def parse(self, content: str, file_type: str = "natural_language") -> dict:
        if file_type == "json":
            return self._parse_json(content)
        return await self._parse_natural_language(content)

    def _parse_json(self, content: str) -> dict:
        try:
            data = json.loads(content)
            return {
                "title": data.get("title", "未命名剧本"),
                "background": data.get("background", ""),
                "characters": data.get("characters", []),
                "clues": data.get("clues", []),
                "phases": data.get("phases", [
                    {"name": "intro", "order": 0, "duration": 120},
                    {"name": "round1_speak", "order": 1, "duration": 300},
                    {"name": "round1_search", "order": 2, "duration": 180},
                    {"name": "round2_speak", "order": 3, "duration": 300},
                    {"name": "round2_search", "order": 4, "duration": 180},
                    {"name": "final_discuss", "order": 5, "duration": 300},
                    {"name": "voting", "order": 6, "duration": 120},
                    {"name": "reveal", "order": 7, "duration": 120},
                ]),
                "character_count": len(data.get("characters", [])),
            }
        except json.JSONDecodeError as e:
            logger.error(f"JSON parse error: {e}")
            return {"title": "未命名剧本", "background": "", "characters": [], "clues": [], "phases": [], "character_count": 0}

    async def _parse_natural_language(self, content: str) -> dict:
        prompt = f"""你是一个剧本杀解析器。请分析以下自然语言描述的剧本，提取结构化信息。

输出格式必须是合法的JSON：
{{
    "title": "剧本名称（从内容推断）",
    "background": "案件背景",
    "characters": [
        {{
            "name": "角色名",
            "personality": "性格特征",
            "speaking_style": "说话风格",
            "secret": "该角色隐藏的秘密",
            "motive": "该角色的动机"
        }}
    ],
    "clues": [
        {{
            "id": "线索ID",
            "content": "线索内容",
            "owner": "线索属于哪个角色",
            "phase": "线索在哪个阶段可用（round1_search/round2_search）"
        }}
    ],
    "phases": [
        {{"name": "intro", "order": 0, "duration": 120}},
        {{"name": "round1_speak", "order": 1, "duration": 300}},
        {{"name": "round1_search", "order": 2, "duration": 180}},
        {{"name": "round2_speak", "order": 3, "duration": 300}},
        {{"name": "round2_search", "order": 4, "duration": 180}},
        {{"name": "final_discuss", "order": 5, "duration": 300}},
        {{"name": "voting", "order": 6, "duration": 120}},
        {{"name": "reveal", "order": 7, "duration": 120}}
    ]
}}

剧本内容：
{content[:8000]}

请只返回JSON，不要包含其他解释文字。"""

        try:
            response = await hermes_client.chat(profile_name="default", message=prompt)
            response = response.strip()
            if response.startswith("```"):
                lines = response.split("\n")
                response = "\n".join(lines[1:-1]) if len(lines) >= 3 else response
            data = json.loads(response)
            data.setdefault("title", "未命名剧本")
            data.setdefault("background", "")
            data.setdefault("characters", [])
            data.setdefault("clues", [])
            data["character_count"] = len(data.get("characters", []))
            if not data.get("phases"):
                data["phases"] = [
                    {"name": "intro", "order": 0, "duration": 120},
                    {"name": "round1_speak", "order": 1, "duration": 300},
                    {"name": "round1_search", "order": 2, "duration": 180},
                    {"name": "round2_speak", "order": 3, "duration": 300},
                    {"name": "round2_search", "order": 4, "duration": 180},
                    {"name": "final_discuss", "order": 5, "duration": 300},
                    {"name": "voting", "order": 6, "duration": 120},
                    {"name": "reveal", "order": 7, "duration": 120},
                ]
            return data
        except (json.JSONDecodeError, Exception) as e:
            logger.warning(f"LLM parse failed: {e}, using regex fallback")
            return self._regex_fallback(content)

    def _regex_fallback(self, content: str) -> dict:
        import re

        title = "未命名剧本"
        first_line = content.strip().split("\n")[0].strip()
        if len(first_line) <= 50:
            title = first_line.lstrip("#").strip()

        characters = []
        char_pattern = re.compile(r"^[#\-\*]*\s*(.{1,10})(?:[：:]\s*(.+))?$", re.MULTILINE)
        name_keywords = re.compile(r"(角色|人物|嫌疑人|侦探|凶手|死者|被害人)", re.IGNORECASE)

        char_section = False
        for line in content.split("\n"):
            line = line.strip()
            if name_keywords.search(line):
                char_section = True
                continue
            if char_section and line.startswith("#"):
                break
            if char_section and len(line) <= 30 and line:
                name = re.sub(r"[-:：\s].*$", "", line).strip()
                if name and len(name) <= 10:
                    characters.append({"name": name, "personality": "", "speaking_style": "", "secret": "", "motive": ""})

        return {
            "title": title,
            "background": content[:200],
            "characters": characters,
            "clues": [],
            "phases": [
                {"name": "intro", "order": 0, "duration": 120},
                {"name": "round1_speak", "order": 1, "duration": 300},
                {"name": "round1_search", "order": 2, "duration": 180},
                {"name": "round2_speak", "order": 3, "duration": 300},
                {"name": "round2_search", "order": 4, "duration": 180},
                {"name": "final_discuss", "order": 5, "duration": 300},
                {"name": "voting", "order": 6, "duration": 120},
                {"name": "reveal", "order": 7, "duration": 120},
            ],
            "character_count": len(characters),
        }


script_parser = ScriptParser()