AI/ragflow
1
0
Fork 0
mirror of https://git.mirrors.martin98.com/https://github.com/infiniflow/ragflow.git synced 2025-08-14 10:45:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

Change default error message to English (#3838)

Browse Source 
### What problem does this PR solve?

As title

### Type of change

- [x] Refactoring

---------

Signed-off-by: Jin Hai <haijin.chn@gmail.com>
This commit is contained in:
Jin Hai 2024-12-04 09:34:49 +08:00 committed by GitHub
parent 87455d79e4
commit 6657ca7cde
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 63 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

82
rag/llm/chat_model.py
View File

@ -22,7 +22,7 @@ from abc import ABC
from openai import OpenAI from openai import OpenAI
import openai import openai
from ollama import Client from ollama import Client
from rag.nlp import is_english from rag.nlp import is_chinese
from rag.utils import num_tokens_from_string from rag.utils import num_tokens_from_string
from groq import Groq from groq import Groq
import os import os
@ -30,6 +30,8 @@ import json
import requests import requests
import asyncio import asyncio
LENGTH_NOTIFICATION_CN = "······\n由于长度的原因，回答被截断了，要继续吗？"
LENGTH_NOTIFICATION_EN = "...\nFor the content length reason, it stopped, continue?"
class Base(ABC): class Base(ABC):
def __init__(self, key, model_name, base_url): def __init__(self, key, model_name, base_url):
@ -47,8 +49,10 @@ class Base(ABC):
**gen_conf) **gen_conf)
ans = response.choices[0].message.content.strip() ans = response.choices[0].message.content.strip()
if response.choices[0].finish_reason == "length": if response.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
return ans, response.usage.total_tokens return ans, response.usage.total_tokens
except openai.APIError as e: except openai.APIError as e:
return "**ERROR**: " + str(e), 0 return "**ERROR**: " + str(e), 0
@ -80,8 +84,10 @@ class Base(ABC):
else: total_tokens = resp.usage.total_tokens else: total_tokens = resp.usage.total_tokens
if resp.choices[0].finish_reason == "length": if resp.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
yield ans yield ans
except openai.APIError as e: except openai.APIError as e:
@ -167,8 +173,10 @@ class BaiChuanChat(Base):
**self._format_params(gen_conf)) **self._format_params(gen_conf))
ans = response.choices[0].message.content.strip() ans = response.choices[0].message.content.strip()
if response.choices[0].finish_reason == "length": if response.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese([ans]):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
return ans, response.usage.total_tokens return ans, response.usage.total_tokens
except openai.APIError as e: except openai.APIError as e:
return "**ERROR**: " + str(e), 0 return "**ERROR**: " + str(e), 0
@ -207,8 +215,10 @@ class BaiChuanChat(Base):
else resp.usage["total_tokens"] else resp.usage["total_tokens"]
) )
if resp.choices[0].finish_reason == "length": if resp.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese([ans]):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
yield ans yield ans
except Exception as e: except Exception as e:
@ -242,8 +252,10 @@ class QWenChat(Base):
ans += response.output.choices[0]['message']['content'] ans += response.output.choices[0]['message']['content']
tk_count += response.usage.total_tokens tk_count += response.usage.total_tokens
if response.output.choices[0].get("finish_reason", "") == "length": if response.output.choices[0].get("finish_reason", "") == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese([ans]):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
return ans, tk_count return ans, tk_count
return "**ERROR**: " + response.message, tk_count return "**ERROR**: " + response.message, tk_count
@ -276,8 +288,10 @@ class QWenChat(Base):
ans = resp.output.choices[0]['message']['content'] ans = resp.output.choices[0]['message']['content']
tk_count = resp.usage.total_tokens tk_count = resp.usage.total_tokens
if resp.output.choices[0].get("finish_reason", "") == "length": if resp.output.choices[0].get("finish_reason", "") == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
yield ans yield ans
else: else:
yield ans + "\n**ERROR**: " + resp.message if not re.search(r" (key|quota)", str(resp.message).lower()) else "Out of credit. Please set the API key in **settings > Model providers.**" yield ans + "\n**ERROR**: " + resp.message if not re.search(r" (key|quota)", str(resp.message).lower()) else "Out of credit. Please set the API key in **settings > Model providers.**"
@ -308,8 +322,10 @@ class ZhipuChat(Base):
) )
ans = response.choices[0].message.content.strip() ans = response.choices[0].message.content.strip()
if response.choices[0].finish_reason == "length": if response.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
return ans, response.usage.total_tokens return ans, response.usage.total_tokens
except Exception as e: except Exception as e:
return "**ERROR**: " + str(e), 0 return "**ERROR**: " + str(e), 0
@ -333,8 +349,10 @@ class ZhipuChat(Base):
delta = resp.choices[0].delta.content delta = resp.choices[0].delta.content
ans += delta ans += delta
if resp.choices[0].finish_reason == "length": if resp.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
tk_count = resp.usage.total_tokens tk_count = resp.usage.total_tokens
if resp.choices[0].finish_reason == "stop": tk_count = resp.usage.total_tokens if resp.choices[0].finish_reason == "stop": tk_count = resp.usage.total_tokens
yield ans yield ans
@ -525,8 +543,10 @@ class MiniMaxChat(Base):
response = response.json() response = response.json()
ans = response["choices"][0]["message"]["content"].strip() ans = response["choices"][0]["message"]["content"].strip()
if response["choices"][0]["finish_reason"] == "length": if response["choices"][0]["finish_reason"] == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
return ans, response["usage"]["total_tokens"] return ans, response["usage"]["total_tokens"]
except Exception as e: except Exception as e:
return "**ERROR**: " + str(e), 0 return "**ERROR**: " + str(e), 0
@ -594,8 +614,10 @@ class MistralChat(Base):
**gen_conf) **gen_conf)
ans = response.choices[0].message.content ans = response.choices[0].message.content
if response.choices[0].finish_reason == "length": if response.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
return ans, response.usage.total_tokens return ans, response.usage.total_tokens
except openai.APIError as e: except openai.APIError as e:
return "**ERROR**: " + str(e), 0 return "**ERROR**: " + str(e), 0
@ -618,8 +640,10 @@ class MistralChat(Base):
ans += resp.choices[0].delta.content ans += resp.choices[0].delta.content
total_tokens += 1 total_tokens += 1
if resp.choices[0].finish_reason == "length": if resp.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
yield ans yield ans
except openai.APIError as e: except openai.APIError as e:
@ -811,8 +835,10 @@ class GroqChat:
) )
ans = response.choices[0].message.content ans = response.choices[0].message.content
if response.choices[0].finish_reason == "length": if response.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
return ans, response.usage.total_tokens return ans, response.usage.total_tokens
except Exception as e: except Exception as e:
return ans + "\n**ERROR**: " + str(e), 0 return ans + "\n**ERROR**: " + str(e), 0
@ -838,8 +864,10 @@ class GroqChat:
ans += resp.choices[0].delta.content ans += resp.choices[0].delta.content
total_tokens += 1 total_tokens += 1
if resp.choices[0].finish_reason == "length": if resp.choices[0].finish_reason == "length":
ans += "...\nFor the content length reason, it stopped, continue?" if is_english( if is_chinese(ans):
[ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？" ans += LENGTH_NOTIFICATION_CN
else:
ans += LENGTH_NOTIFICATION_EN
yield ans yield ans
except Exception as e: except Exception as e:

8
rag/nlp/__init__.py
View File

@ -230,6 +230,14 @@ def is_english(texts):
return True return True
return False return False
def is_chinese(text):
chinese = 0
for ch in text:
if '\u4e00' <= ch <= '\u9fff':
chinese += 1
if chinese / len(text) > 0.2:
return True
return False
def tokenize(d, t, eng): def tokenize(d, t, eng):
d["content_with_weight"] = t d["content_with_weight"] = t