This commit is contained in:
2025-08-20 14:02:58 +08:00
2 changed files with 74 additions and 58 deletions

View File

@@ -47,19 +47,24 @@ async def train_document_task():
logging.info(f"开始处理文档:{document_name}, 还有{len(no_train_document_result) - 1}个文档需要处理!")
# 训练代码开始
# content = get_docx_content_by_pandoc(document_path)
train_result = True
try:
# 注意默认设置使用NetworkX
rag = await initialize_rag(working_dir)
# 获取docx文件的内容
content = get_docx_content_by_pandoc(document_path)
await rag.ainsert(content, ids=[document_name], file_paths=[document_name])
logger.info(f"Inserted content from {document_name}")
if content is not None:
await rag.ainsert(content, ids=[document_name], file_paths=[document_name])
logger.info(f"Inserted content from {document_name}")
else:
train_result = False
except Exception as e:
logger.error(f"An error occurred: {e}")
finally:
await rag.finalize_storages()
# 训练结束,更新训练状态
update_document_sql: str = " UPDATE t_ai_teaching_model_document SET train_flag = 2 WHERE id = " + str(document["id"])
train_flag = "2" if train_result else "7"
update_document_sql: str = " UPDATE t_ai_teaching_model_document SET train_flag = " + train_flag + " WHERE id = " + str(document["id"])
await execute_sql(update_document_sql, ())
elif document["train_flag"] == 3:
update_sql: str = " UPDATE t_ai_teaching_model_document SET train_flag = 4 WHERE id = " + str(document["id"])
@@ -80,22 +85,30 @@ async def train_document_task():
await execute_sql(update_document_sql, ())
# 整体更新主题状态
select_document_sql: str = f"select train_flag, count(1) as train_count from t_ai_teaching_model_document where theme_id = {theme['id']} and is_deleted = 0 and train_flag in (0,1,2) group by train_flag"
select_document_sql: str = f"select train_flag, count(1) as train_count from t_ai_teaching_model_document where theme_id = {theme['id']} and is_deleted = 0 and train_flag in (0,1,2,7) group by train_flag"
select_document_result = await find_by_sql(select_document_sql, ())
train_document_count_map = {}
for item in select_document_result:
train_document_count_map[str(item["train_flag"])] = int(item["train_count"])
train_document_count_0 = train_document_count_map.get("0", 0)
train_document_count_1 = train_document_count_map.get("1", 0)
train_document_count_2 = train_document_count_map.get("2", 0)
train_document_count_7 = train_document_count_map.get("7", 0)
search_flag = 0
if train_document_count_2 > 0:
update_theme_sql: str = f"UPDATE t_ai_teaching_model_theme SET search_flag = 1, train_flag = 2 WHERE id = {theme['id']}"
search_flag = 1
# 训练未开始:初始化| 或者没有is_deleted=0的文档
if train_document_count_0 == 0 and train_document_count_1 == 0 and train_document_count_2 == 0 and train_document_count_7 == 0:
update_theme_sql: str = f"UPDATE t_ai_teaching_model_theme SET search_flag = {search_flag}, train_flag = 0 WHERE id = {theme['id']}"
await execute_sql(update_theme_sql, ())
else:
if train_document_count_1 > 0:
update_theme_sql: str = f"UPDATE t_ai_teaching_model_theme SET search_flag = 0, train_flag = 1 WHERE id = {theme['id']}"
await execute_sql(update_theme_sql, ())
else:
update_theme_sql: str = f"UPDATE t_ai_teaching_model_theme SET search_flag = 0, train_flag = 0 WHERE id = {theme['id']}"
await execute_sql(update_theme_sql, ())
# 训练进行中:单个文档训练中|同时存在训练完成+未训练的文档
if train_document_count_2 > 0 and (train_document_count_1 > 0 or train_document_count_7 > 0):
update_theme_sql: str = f"UPDATE t_ai_teaching_model_theme SET search_flag = {search_flag}, train_flag = 1 WHERE id = {theme['id']}"
await execute_sql(update_theme_sql, ())
# 训练已完成所有is_deleted=0的文档都训练完成
if train_document_count_2 > 0 and train_document_count_0 == 0 and train_document_count_1 == 0 and train_document_count_7 == 0:
update_theme_sql: str = f"UPDATE t_ai_teaching_model_theme SET search_flag = {search_flag}, train_flag = 2 WHERE id = {theme['id']}"
await execute_sql(update_theme_sql, ())
# 添加适当的等待时间,避免频繁查询
await asyncio.sleep(120) # 每二分钟查询一次

View File

@@ -93,50 +93,53 @@ def get_docx_content_by_pandoc(docx_file):
resize_images_in_directory('./static/Images/' + md5_value + '/media')
# 读取然后修改内容,输出到新的文件
img_idx = 0 # 图片索引
with open(temp_markdown, 'r', encoding='utf-8') as f:
for line in f:
line = line.strip()
if not line:
continue
# 跳过图片高度描述行
if line.startswith('height=') and (line.endswith('in"}') or line.endswith('in"')):
continue
# height="1.91044072615923in"
# 使用find()方法安全地检查图片模式
is_img = line.find("![](") >= 0 and (
line.find(".png") > 0 or
line.find(".jpg") > 0 or
line.find(".jpeg") > 0
)
if is_img:
# ![](media/image3.png){width="3.1251607611548557in"
# height="3.694634733158355in"}
# ![](../static/Images/01b20e04085e406ea5375791da58a60f/media/image3.png){width="3.1251607611548557in"
pos = line.find(")")
q = line[:pos + 1]
q = q.replace("./static", ".")
# Modify by Kalman.CHENG ☆: 增加逻辑对图片路径处理,在(和static之间加上/
left_idx = line.find("(")
static_idx = line.find("static")
if left_idx == -1 or static_idx == -1 or left_idx > static_idx:
print("路径中不包含(+~+static的已知格式")
else:
between_content = q[left_idx+1:static_idx].strip()
if between_content:
q = q[:left_idx+1] + '\\' + q[static_idx:]
if os.path.exists(temp_markdown):
with open(temp_markdown, 'r', encoding='utf-8') as f:
for line in f:
line = line.strip()
if not line:
continue
# 跳过图片高度描述行
if line.startswith('height=') and (line.endswith('in"}') or line.endswith('in"')):
continue
# height="1.91044072615923in"
# 使用find()方法安全地检查图片模式
is_img = line.find("![](") >= 0 and (
line.find(".png") > 0 or
line.find(".jpg") > 0 or
line.find(".jpeg") > 0
)
if is_img:
# ![](media/image3.png){width="3.1251607611548557in"
# height="3.694634733158355in"}
# ![](../static/Images/01b20e04085e406ea5375791da58a60f/media/image3.png){width="3.1251607611548557in"
pos = line.find(")")
q = line[:pos + 1]
q = q.replace("./static", ".")
# Modify by Kalman.CHENG ☆: 增加逻辑对图片路径处理,在(和static之间加上/
left_idx = line.find("(")
static_idx = line.find("static")
if left_idx == -1 or static_idx == -1 or left_idx > static_idx:
print("路径中不包含(+~+static的已知格式")
else:
q = q[:static_idx] + '\\' + q[static_idx:]
print(f"q3{q}")
#q = q[4:-1]
#q='<img src="'+q+'" alt="我是图片">'
img_idx += 1
content += q + "\n"
else:
content += line.strip().replace("**", "") + "\n"
content = content.replace("\phantom", "")
# 将content回写到markdown文件
with open(temp_markdown, 'w', encoding='utf-8') as f:
f.write(content)
# 删除临时文件 output_file
# os.remove(temp_markdown)
return content.replace("\n\n", "\n").replace("\\", "/")
between_content = q[left_idx+1:static_idx].strip()
if between_content:
q = q[:left_idx+1] + '\\' + q[static_idx:]
else:
q = q[:static_idx] + '\\' + q[static_idx:]
print(f"q3{q}")
#q = q[4:-1]
#q='<img src="'+q+'" alt="我是图片">'
img_idx += 1
content += q + "\n"
else:
content += line.strip().replace("**", "") + "\n"
content = content.replace("\phantom", "")
# 将content回写到markdown文件
with open(temp_markdown, 'w', encoding='utf-8') as f:
f.write(content)
# 删除临时文件 output_file
# os.remove(temp_markdown)
return content.replace("\n\n", "\n").replace("\\", "/")
else:
return None