0.1.187
This commit is contained in:
217
PyPI/src/guan/AI_chat.py
Normal file
217
PyPI/src/guan/AI_chat.py
Normal file
@@ -0,0 +1,217 @@
|
||||
# Module: AI_chat
|
||||
|
||||
# AI 对话
|
||||
def chat(prompt='你好', model=1, stream=1, stream_label=0):
|
||||
import requests
|
||||
url = "http://api.guanjihuan.com/chat"
|
||||
data = {
|
||||
"prompt": prompt,
|
||||
"model": model,
|
||||
}
|
||||
if stream == 1:
|
||||
if stream_label == 1:
|
||||
print('\n--- Start Chat Stream Message ---\n')
|
||||
requests_response = requests.post(url, json=data, stream=True)
|
||||
response = ''
|
||||
if requests_response.status_code == 200:
|
||||
for line in requests_response.iter_lines():
|
||||
if line:
|
||||
if stream == 1:
|
||||
print(line.decode('utf-8'), end='', flush=True)
|
||||
response += line.decode('utf-8')
|
||||
print()
|
||||
else:
|
||||
pass
|
||||
if stream == 1:
|
||||
if stream_label == 1:
|
||||
print('\n--- End Chat Stream Message ---\n')
|
||||
return response
|
||||
|
||||
# 加上函数代码的 AI 对话
|
||||
def chat_with_function_code(function_name, prompt='', model=1, stream=1):
|
||||
import guan
|
||||
function_source = guan.get_source(function_name)
|
||||
if prompt == '':
|
||||
response = guan.chat(prompt=function_source, model=model, stream=stream)
|
||||
else:
|
||||
response = guan.chat(prompt=function_source+'\n\n'+prompt, model=model, stream=stream)
|
||||
return response
|
||||
|
||||
# 机器人自动对话
|
||||
def auto_chat(prompt='你好', round=2, model=1, stream=1):
|
||||
import guan
|
||||
response0 = prompt
|
||||
for i0 in range(round):
|
||||
print(f'\n【对话第 {i0+1} 轮】\n')
|
||||
print('机器人 1: ')
|
||||
response1 = guan.chat(prompt=response0, model=model, stream=stream)
|
||||
print('机器人 2: ')
|
||||
response0 = guan.chat(prompt=response1, model=model, stream=stream)
|
||||
|
||||
# 机器人自动对话（引导对话）
|
||||
def auto_chat_with_guide(prompt='你好', guide_message='（回答字数少于30个字，最后反问我一个问题）', round=5, model=1, stream=1):
|
||||
import guan
|
||||
response0 = prompt
|
||||
for i0 in range(round):
|
||||
print(f'\n【对话第 {i0+1} 轮】\n')
|
||||
print('机器人 1: ')
|
||||
response1 = guan.chat(prompt=response0+guide_message, model=model, stream=stream)
|
||||
print('机器人 2: ')
|
||||
response0 = guan.chat(prompt=response1+guide_message, model=model, stream=stream)
|
||||
|
||||
# 使用 LangChain 无记忆对话（需要 API Key)
|
||||
def langchain_chat_without_memory(prompt="你好", temperature=0.7, system_message=None, print_show=1):
|
||||
from langchain_openai import ChatOpenAI
|
||||
from langchain_core.prompts import ChatPromptTemplate
|
||||
import dotenv
|
||||
import os
|
||||
dotenv.load_dotenv()
|
||||
llm = ChatOpenAI(
|
||||
api_key=os.getenv("OPENAI_API_KEY"),
|
||||
base_url=os.getenv("DASHSCOPE_BASE_URL"),
|
||||
model="qwen-plus",
|
||||
temperature=temperature,
|
||||
streaming=True,
|
||||
)
|
||||
if system_message == None:
|
||||
langchain_prompt = ChatPromptTemplate.from_messages([
|
||||
("human", "{question}")
|
||||
])
|
||||
else:
|
||||
langchain_prompt = ChatPromptTemplate.from_messages([
|
||||
("system", system_message),
|
||||
("human", "{question}")
|
||||
])
|
||||
chain = langchain_prompt | llm
|
||||
response = ''
|
||||
for chunk in chain.stream({"question": prompt}):
|
||||
response += chunk.content
|
||||
if print_show:
|
||||
print(chunk.content, end="", flush=True)
|
||||
if print_show:
|
||||
print()
|
||||
return response
|
||||
|
||||
# 使用 LangChain 有记忆对话（记忆临时保存在函数的属性上，需要 API Key)
|
||||
def langchain_chat_with_memory(prompt="你好", temperature=0.7, system_message=None, session_id="default", print_show=1):
|
||||
from langchain_openai import ChatOpenAI
|
||||
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
|
||||
from langchain_core.runnables.history import RunnableWithMessageHistory
|
||||
from langchain_community.chat_message_histories import ChatMessageHistory
|
||||
from dotenv import load_dotenv
|
||||
import os
|
||||
load_dotenv()
|
||||
llm = ChatOpenAI(
|
||||
api_key=os.getenv("OPENAI_API_KEY"),
|
||||
base_url=os.getenv("DASHSCOPE_BASE_URL"),
|
||||
model="qwen-plus",
|
||||
temperature=temperature,
|
||||
streaming=True,
|
||||
)
|
||||
if system_message == None:
|
||||
langchain_prompt = ChatPromptTemplate.from_messages([
|
||||
MessagesPlaceholder("history"),
|
||||
("human", "{question}")
|
||||
])
|
||||
else:
|
||||
langchain_prompt = ChatPromptTemplate.from_messages([
|
||||
("system", system_message),
|
||||
MessagesPlaceholder("history"),
|
||||
("human", "{question}")
|
||||
])
|
||||
chain = langchain_prompt | llm
|
||||
if not hasattr(langchain_chat_with_memory, "store"):
|
||||
langchain_chat_with_memory.store = {}
|
||||
|
||||
def get_session_history(sid: str):
|
||||
if sid not in langchain_chat_with_memory.store:
|
||||
langchain_chat_with_memory.store[sid] = ChatMessageHistory()
|
||||
return langchain_chat_with_memory.store[sid]
|
||||
|
||||
chatbot = RunnableWithMessageHistory(
|
||||
chain,
|
||||
lambda sid: get_session_history(sid),
|
||||
input_messages_key="question",
|
||||
history_messages_key="history",
|
||||
)
|
||||
response = ''
|
||||
for chunk in chatbot.stream({"question": prompt}, config={"configurable": {"session_id": session_id}}):
|
||||
response += chunk.content
|
||||
if print_show:
|
||||
print(chunk.content, end="", flush=True)
|
||||
if print_show:
|
||||
print()
|
||||
return response
|
||||
|
||||
# 使用 Ollama 本地模型对话（需要运行 Ollama 和下载对应的模型）
|
||||
def ollama_chat(prompt='你好/no_think', model="qwen3:0.6b", temperature=0.8, print_show=1):
|
||||
import ollama
|
||||
response_stream = ollama.chat(model=model, messages=[{"role": "user", "content": prompt}], stream=True, options={"temperature": temperature})
|
||||
response = ''
|
||||
start_thinking = 1
|
||||
for part in response_stream:
|
||||
response += part['message']['content']
|
||||
if print_show == 1:
|
||||
thinking = part['message'].get('thinking')
|
||||
if thinking is not None:
|
||||
if start_thinking == 1:
|
||||
print('<think>')
|
||||
start_thinking = 0
|
||||
print(f"{thinking}", end='', flush=True)
|
||||
else:
|
||||
if start_thinking == 0:
|
||||
print('</think>')
|
||||
start_thinking = 1
|
||||
print(part['message']['content'], end='', flush=True)
|
||||
if print_show == 1:
|
||||
print()
|
||||
return response
|
||||
|
||||
# ModelScope 加载本地模型和分词器（只加载一次）
|
||||
def load_modelscope_model(model_name="D:/models/Qwen/Qwen3-0.6B"):
|
||||
from modelscope import AutoModelForCausalLM, AutoTokenizer
|
||||
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
||||
model = AutoModelForCausalLM.from_pretrained(model_name)
|
||||
return model, tokenizer
|
||||
|
||||
# 使用 ModelScope 本地模型聊天
|
||||
def modelscope_chat(model, tokenizer, prompt='你好 /no_think', history=[], temperature=0.7, top_p=0.8):
|
||||
messages = history + [{"role": "user", "content": prompt}]
|
||||
text = tokenizer.apply_chat_template(
|
||||
messages,
|
||||
tokenize=False,
|
||||
add_generation_prompt=True
|
||||
)
|
||||
inputs = tokenizer(text, return_tensors="pt")
|
||||
response_ids = model.generate(**inputs, max_new_tokens=32768, temperature=temperature, top_p=top_p, do_sample=True)[0][len(inputs.input_ids[0]):].tolist()
|
||||
response = tokenizer.decode(response_ids, skip_special_tokens=True)
|
||||
new_history = history + [
|
||||
{"role": "user", "content": prompt},
|
||||
{"role": "assistant", "content": response}
|
||||
]
|
||||
return response, new_history
|
||||
|
||||
# LLaMA 加载本地模型（只加载一次）
|
||||
def load_llama_model(model_path="D:/models/Qwen/Qwen3-0.6B-GGUF/Qwen3-0.6B-Q8_0.gguf"):
|
||||
from llama_cpp import Llama
|
||||
llm = Llama(
|
||||
model_path=model_path,
|
||||
n_ctx=32768,
|
||||
verbose=False,
|
||||
chat_format="chatml",
|
||||
logits_all=False
|
||||
)
|
||||
return llm
|
||||
|
||||
# 使用 LLaMA 本地模型聊天
|
||||
def llama_chat(llm, prompt, history=[], temperature=0.7, top_p=0.8):
|
||||
new_history = history + [{"role": "user", "content": prompt}]
|
||||
llm_response = llm.create_chat_completion(
|
||||
messages=new_history,
|
||||
temperature=temperature,
|
||||
top_p=top_p,
|
||||
repeat_penalty=1.5,
|
||||
)
|
||||
response = llm_response["choices"][0]["message"]["content"].strip()
|
||||
new_history.append({"role": "assistant", "content": response})
|
||||
return response, new_history
|
||||
Reference in New Issue
Block a user