Langchain-chatchat

提示：Langchain服务已经正常启动，swagger-ui页面也正常访问的情况下出现接口调试报错问题

前言

报错接口如下：

2023-11-03 15:23:50 | INFO | stdout | INFO:     127.0.0.1:53907 - "POST /v1/chat/chat HTTP/1.1" 404 Not Found
2023-11-03 15:24:01 | INFO | stdout | INFO:     127.0.0.1:53920 - "POST /v1 HTTP/1.1" 404 Not Found
2023-11-03 15:24:23 | INFO | stdout | INFO:     127.0.0.1:53938 - "GET /v1 HTTP/1.1" 404 Not Found
INFO:     127.0.0.1:54053 - "GET / HTTP/1.1" 307 Temporary Redirect
INFO:     127.0.0.1:54053 - "GET /docs HTTP/1.1" 200 OK
openai.api_key='EMPTY'
openai.api_base='http://127.0.0.1:20000/v1'
model='chatglm2-6b' messages=[OpenAiMessage(role='user', content='hello')] temperature=0.7 n=1 max_tokens=0 stop=[] stream=False presence_penalty=0 frequency_penalty=0
INFO:     127.0.0.1:54064 - "POST /chat/fastchat HTTP/1.1" 200 OK
2023-11-03 15:26:58 | INFO | stdout | INFO:     127.0.0.1:54065 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:26:58,486 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:26:58,487 - openai_chat.py[line:52] - ERROR: APIError: 获取ChatCompletion时出错：Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400)
openai.api_key='EMPTY'
openai.api_base='http://127.0.0.1:20000/v1'
model='chatglm2-6b' messages=[OpenAiMessage(role='user', content='hello')] temperature=0.7 n=1 max_tokens=0 stop=[] stream=False presence_penalty=0 frequency_penalty=0
INFO:     127.0.0.1:54064 - "POST /chat/fastchat HTTP/1.1" 200 OK
2023-11-03 15:27:02 | INFO | stdout | INFO:     127.0.0.1:54071 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:27:02,322 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:27:02,323 - openai_chat.py[line:52] - ERROR: APIError: 获取ChatCompletion时出错：Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400)
openai.api_key='EMPTY'
openai.api_base='http://127.0.0.1:20000/v1'
model='chatglm2-6b' messages=[OpenAiMessage(role='user', content='hello')] temperature=0.7 n=1 max_tokens=0 stop=[] stream=False presence_penalty=0 frequency_penalty=0
INFO:     127.0.0.1:54064 - "POST /chat/fastchat HTTP/1.1" 200 OK
2023-11-03 15:27:02 | INFO | stdout | INFO:     127.0.0.1:54073 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:27:02,968 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:27:02,970 - openai_chat.py[line:52] - ERROR: APIError: 获取ChatCompletion时出错：Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400)
123
INFO:     127.0.0.1:54096 - "POST /chat/chat HTTP/1.1" 200 OK
{'cache': None, 'verbose': True, 'callbacks': [<langchain.callbacks.streaming_aiter.AsyncIteratorCallbackHandler object at 0x000001DE719FE610>], 'callback_manager': None, 'tags': None, 'metadata': None, 'client': <class 'openai.api_resources.chat_completion.ChatCompletion'>, 'model_name': 'chatglm2-6b', 'temperature': 0.7, 'model_kwargs': {}, 'openai_api_key': 'EMPTY', 'openai_api_base': 'http://127.0.0.1:20000/v1', 'openai_organization': '', 'openai_proxy': '', 'request_timeout': None, 'max_retries': 6, 'streaming': True, 'n': 1, 'max_tokens': 0, 'tiktoken_model_name': None}
2023-11-03 15:27:36 | INFO | stdout | INFO:     127.0.0.1:54097 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:27:36,492 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:27:36,493 - before_sleep.py[line:65] - WARNING: Retrying langchain.chat_models.openai.acompletion_with_retry.<locals>._completion_with_retry in 4.0 seconds as it raised APIError: Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400).
2023-11-03 15:27:40 | INFO | stdout | INFO:     127.0.0.1:54103 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:27:40,498 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:27:40,499 - before_sleep.py[line:65] - WARNING: Retrying langchain.chat_models.openai.acompletion_with_retry.<locals>._completion_with_retry in 4.0 seconds as it raised APIError: Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400).
2023-11-03 15:27:44 | INFO | stdout | INFO:     127.0.0.1:54109 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:27:44,521 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:27:44,524 - before_sleep.py[line:65] - WARNING: Retrying langchain.chat_models.openai.acompletion_with_retry.<locals>._completion_with_retry in 4.0 seconds as it raised APIError: Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400).
2023-11-03 15:27:48 | INFO | stdout | INFO:     127.0.0.1:54113 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:27:48,537 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:27:48,537 - before_sleep.py[line:65] - WARNING: Retrying langchain.chat_models.openai.acompletion_with_retry.<locals>._completion_with_retry in 8.0 seconds as it raised APIError: Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400).
2023-11-03 15:27:56 | INFO | stdout | INFO:     127.0.0.1:54124 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:27:56,572 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:27:56,574 - before_sleep.py[line:65] - WARNING: Retrying langchain.chat_models.openai.acompletion_with_retry.<locals>._completion_with_retry in 10.0 seconds as it raised APIError: Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400).
2023-11-03 15:28:06 | INFO | stdout | INFO:     127.0.0.1:54138 - "POST /v1/chat/completions HTTP/1.1" 400 Bad Request
2023-11-03 15:28:06,598 - util.py[line:67] - INFO: message='OpenAI API response' path=http://127.0.0.1:20000/v1/chat/completions processing_ms=None request_id=None response_code=400
2023-11-03 15:28:06,604 - utils.py[line:26] - ERROR: APIError: Caught exception: Invalid response object from API: '{"object":"error","message":"0 is less than the minimum of 1 - \'max_tokens\'","code":40302}' (HTTP response code was 400)

问题分析

请求体重自带 【max_tokens】，如果有max_tokens=0，不为null，就去请求方案openAI接口，我们只要掉本地api接口就可了。所以解决方案就是将这个参数从请求体中删掉就可以正常请求了，这个问题困扰了我两天，真的是太反人类了。
希望对大家有帮助。

解决问题方案

删除【max_tokens】参数。