增加调试语句
This commit is contained in:
parent
e60bdecd20
commit
a04cfcabce
@ -218,8 +218,21 @@ async def converse_with_chat_assistant(
|
||||
|
||||
# 流式响应
|
||||
if converse_params.stream:
|
||||
# 创建缓存存储回调函数
|
||||
async def make_cache_store(chat_id: str, question: str):
|
||||
async def store_answer(answer: str):
|
||||
if redis and answer and len(answer.strip()) >= 10:
|
||||
try:
|
||||
await _async_store_qa(chat_id, question, answer, redis)
|
||||
logger.info(f'[RAG_CACHE] 语义缓存存储成功 | chat_id={chat_id} | answer_length={len(answer)}')
|
||||
except Exception as e:
|
||||
logger.warning(f'语义缓存存储失败: {e}')
|
||||
return store_answer
|
||||
|
||||
cache_store = await make_cache_store(converse_params.chat_id, converse_params.question)
|
||||
|
||||
return StreamingResponse(
|
||||
stream_ragflow_response(result, converse_params.chat_id, start_time),
|
||||
stream_ragflow_response(result, converse_params.chat_id, start_time, cache_store_func=cache_store),
|
||||
media_type='text/event-stream',
|
||||
headers={
|
||||
'Cache-Control': 'no-cache',
|
||||
@ -229,47 +242,12 @@ async def converse_with_chat_assistant(
|
||||
}
|
||||
)
|
||||
|
||||
# 非流式响应
|
||||
response = parse_result(result)
|
||||
|
||||
# 记录原生RAG响应日志
|
||||
if isinstance(result, dict) and result.get('code') == 0:
|
||||
answer_data = result.get('data', {})
|
||||
answer_text = answer_data.get('answer') if isinstance(answer_data, dict) else str(answer_data)
|
||||
logger.info(f'[RAG_SOURCE] 原生RAG非流式响应完成 | chat_id={converse_params.chat_id} | answer_length={len(answer_text) if answer_text else 0}')
|
||||
|
||||
# 设置语义缓存（存储问答对）
|
||||
if not converse_params.stream and redis and isinstance(result, dict) and result.get('code') == 0:
|
||||
try:
|
||||
answer_data = result.get('data', {})
|
||||
answer_text = answer_data.get('answer') if isinstance(answer_data, dict) else str(answer_data)
|
||||
if answer_text and len(answer_text.strip()) >= 10:
|
||||
# 异步存储，不阻塞响应返回
|
||||
asyncio.create_task(_async_store_qa(
|
||||
converse_params.chat_id,
|
||||
converse_params.question,
|
||||
answer_text,
|
||||
redis
|
||||
))
|
||||
except Exception as e:
|
||||
logger.warning(f"语义缓存存储失败: {e}")
|
||||
|
||||
# 设置原有精确缓存
|
||||
if redis and cache_key and isinstance(result, dict) and result.get('code') == 0:
|
||||
try:
|
||||
await redis.set(cache_key, json.dumps(result.get('data', {}), ensure_ascii=False), ex=60)
|
||||
except Exception as e:
|
||||
logger.warning(f"缓存设置失败: {e}")
|
||||
|
||||
logger.info(f'[RAG_PROFILE] Total Sync Duration ({converse_params.chat_id}): {time.perf_counter() - start_time:.3f}s')
|
||||
return response
|
||||
|
||||
except Exception as e:
|
||||
logger.exception(f'[RAG_PROFILE] ragflow对话异常: {e}')
|
||||
return ResponseUtil.error(msg=f'对话服务异常: {str(e)}')
|
||||
|
||||
|
||||
def stream_ragflow_response(result: Generator, chat_id: str, start_time: float) -> Generator[str, None, None]:
|
||||
def stream_ragflow_response(result: Generator, chat_id: str, start_time: float, cache_store_func=None) -> Generator[str, None, None]:
|
||||
"""
|
||||
流式处理RAGFlow响应 - 同步版本，修复首token延迟
|
||||
"""
|
||||
@ -361,6 +339,14 @@ def stream_ragflow_response(result: Generator, chat_id: str, start_time: float)
|
||||
finally:
|
||||
total_time = time.perf_counter() - start_time
|
||||
logger.info(f'[RAG_SERVER {time.time():.3f}] ⏱️ Total Stream Duration ({chat_id}): {total_time:.3f}s')
|
||||
|
||||
# 流结束后存储缓存
|
||||
if cache_store_func and last_answer and len(last_answer.strip()) >= 10:
|
||||
try:
|
||||
cache_store_func(last_answer)
|
||||
logger.info(f'[RAG_CACHE] 流式响应缓存已存储 | chat_id={chat_id} | answer_length={len(last_answer)}')
|
||||
except Exception as cache_err:
|
||||
logger.warning(f'[RAG_CACHE] 流式响应缓存存储失败: {cache_err}')
|
||||
|
||||
|
||||
# 聊天助手列表
|
||||
|
||||
@ -485,14 +485,15 @@ if __name__ == "__main__":
|
||||
)
|
||||
else:
|
||||
# 性能测试模式
|
||||
# 定义三个测试问题
|
||||
# 问题1和2完全相同，用于测试精确缓存
|
||||
# 问题3与问题1意思相同但表述略不同，用于测试语义缓存
|
||||
questions = [
|
||||
"公司有什么产品？", # 问题1
|
||||
"公司有什么产品？", # 问题2（完全相同，测试精确缓存）
|
||||
"公司有哪些产品？", # 问题3（表述略不同，测试语义缓存）
|
||||
]
|
||||
if question and question != DEFAULT_QUESTION:
|
||||
questions = [question, question, question]
|
||||
print(f"📝 使用命令行参数的问题进行性能测试: {question}")
|
||||
else:
|
||||
questions = [
|
||||
"公司有什么产品？", # 问题1
|
||||
"公司有什么产品？", # 问题2（完全相同，测试精确缓存）
|
||||
"公司有哪些产品？", # 问题3（表述略不同，测试语义缓存）
|
||||
]
|
||||
|
||||
asyncio.run(
|
||||
run_performance_test(
|
||||
|
||||
@ -169,6 +169,9 @@ class MatchService:
|
||||
kw1 = set(self.extract_keywords(q1))
|
||||
kw2 = set(self.extract_keywords(q2))
|
||||
|
||||
logger.info(f'[MatchDebug] q1={q1} | n1={n1} | kw1={kw1}')
|
||||
logger.info(f'[MatchDebug] q2={q2} | n2={n2} | kw2={kw2}')
|
||||
|
||||
if not kw1 or not kw2:
|
||||
return 0.0
|
||||
|
||||
@ -192,11 +195,14 @@ class MatchService:
|
||||
|
||||
for candidate_text, candidate_data in candidates:
|
||||
similarity = self.calculate_similarity(query, candidate_text)
|
||||
logger.info(f'[MatchDebug] compare | query="{query}" | candidate="{candidate_text}" | sim={similarity:.3f}')
|
||||
|
||||
if similarity > best_similarity:
|
||||
best_similarity = similarity
|
||||
best_match = candidate_data
|
||||
|
||||
logger.info(f'[MatchDebug] best | query="{query}" | best_sim={best_similarity:.3f} | threshold={threshold}')
|
||||
|
||||
if best_similarity >= threshold:
|
||||
return best_match, best_similarity
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user