127 lines
4.5 KiB
Python
127 lines
4.5 KiB
Python
"""
|
||
回忆录整理 Agent：基于传记结构，将口语改写为书面语，归类到章节
|
||
支持异步调用
|
||
"""
|
||
|
||
import json
|
||
from typing import Dict, List, Optional
|
||
|
||
from app.agents.memoir.prompts import (
|
||
CHAPTER_CATEGORIES,
|
||
STAGE_TO_ORDER,
|
||
get_chapter_classification_prompt,
|
||
get_text_rewrite_prompt,
|
||
)
|
||
from app.core.dependencies import get_llm_provider
|
||
from app.core.langchain_llm import bind_json_object_mode
|
||
from app.core.logging import get_logger
|
||
from app.features.memoir.memoir_images.json_payload import extract_json_payload
|
||
|
||
logger = get_logger(__name__)
|
||
|
||
|
||
def _get_langchain_llm():
|
||
try:
|
||
provider = get_llm_provider()
|
||
return getattr(provider, "langchain_llm", None)
|
||
except Exception:
|
||
return None
|
||
|
||
|
||
class MemoryAgent:
|
||
"""回忆录整理 Agent（支持异步）"""
|
||
|
||
def __init__(self):
|
||
self.llm = _get_langchain_llm()
|
||
|
||
async def classify_chapter(self, segments_text: str) -> str:
|
||
if not self.llm:
|
||
return "childhood"
|
||
try:
|
||
prompt = get_chapter_classification_prompt(segments_text)
|
||
response = await self.llm.ainvoke(prompt)
|
||
content = (
|
||
response.content if hasattr(response, "content") else str(response)
|
||
)
|
||
category = content.strip().lower()
|
||
if category in CHAPTER_CATEGORIES:
|
||
return category
|
||
except Exception as e:
|
||
logger.error("分类章节失败: %s", e)
|
||
return "childhood"
|
||
|
||
async def rewrite_to_literary(
|
||
self,
|
||
segments_text: str,
|
||
chapter_category: str,
|
||
existing_content: Optional[str] = None,
|
||
) -> Dict:
|
||
if not self.llm:
|
||
return {
|
||
"title": CHAPTER_CATEGORIES.get(chapter_category, "章节"),
|
||
"content": segments_text,
|
||
"summary": "",
|
||
"image_suggestions": [],
|
||
}
|
||
try:
|
||
prompt = get_text_rewrite_prompt(
|
||
segments_text, chapter_category, existing_content or ""
|
||
)
|
||
json_llm = bind_json_object_mode(self.llm, max_tokens=4096)
|
||
response = await json_llm.ainvoke(prompt)
|
||
content = (
|
||
response.content if hasattr(response, "content") else str(response)
|
||
)
|
||
content = content.strip()
|
||
result = json.loads(extract_json_payload(content))
|
||
return result
|
||
except json.JSONDecodeError:
|
||
raw = response.content if hasattr(response, "content") else str(response)
|
||
return {
|
||
"title": CHAPTER_CATEGORIES.get(chapter_category, "章节"),
|
||
"content": raw,
|
||
"summary": "",
|
||
"image_suggestions": [],
|
||
}
|
||
except Exception as e:
|
||
logger.error("改写文本失败: %s", e)
|
||
return {
|
||
"title": CHAPTER_CATEGORIES.get(chapter_category, "章节"),
|
||
"content": segments_text,
|
||
"summary": "",
|
||
"image_suggestions": [],
|
||
}
|
||
|
||
async def process_segments(
|
||
self,
|
||
segments: List[Dict],
|
||
existing_chapters: Optional[Dict[str, Dict]] = None,
|
||
) -> Dict[str, Dict]:
|
||
if existing_chapters is None:
|
||
existing_chapters = {}
|
||
segments_by_category: Dict[str, List[str]] = {}
|
||
for segment in segments:
|
||
text = segment.get("transcript_text", "")
|
||
if not text:
|
||
continue
|
||
category = await self.classify_chapter(text)
|
||
if category not in segments_by_category:
|
||
segments_by_category[category] = []
|
||
segments_by_category[category].append(text)
|
||
updated_chapters = existing_chapters.copy()
|
||
for category, texts in segments_by_category.items():
|
||
combined_text = "\n\n".join(texts)
|
||
existing_content = existing_chapters.get(category, {}).get("content", "")
|
||
result = await self.rewrite_to_literary(
|
||
combined_text, category, existing_content
|
||
)
|
||
updated_chapters[category] = {
|
||
"title": result.get("title", CHAPTER_CATEGORIES.get(category, "章节")),
|
||
"content": result.get("content", ""),
|
||
"summary": result.get("summary", ""),
|
||
"image_suggestions": result.get("image_suggestions", []),
|
||
"category": category,
|
||
"order_index": STAGE_TO_ORDER.get(category, 999),
|
||
}
|
||
return updated_chapters
|