-
Notifications
You must be signed in to change notification settings - Fork 0
/
kimi_tools_call.py
130 lines (117 loc) · 5.86 KB
/
kimi_tools_call.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
from typing import *
import json
from openai import OpenAI
import os
import json
from diophila import OpenAlex
openalex = OpenAlex()
client = OpenAI(
api_key="sk-106i1MhEY7PtQtNMYPwOly1pf2LuvTgxseX3XvF9pKrhoB0D", # 在这里将 MOONSHOT_API_KEY 替换为你从 Kimi 开放平台申请的 API Key
base_url="https://api.moonshot.cn/v1",
)
tools = [
{
"type": "function", # 约定的字段 type，目前支持 function 作为值
"function": {
"name": "get_weather",
"description": """
获取当前天气
当你的知识无法回答用户提出的问题，或用户请求你进行联网搜索时，调用此工具。
""",
"parameters": { # 使用 parameters 字段来定义函数接收的参数
"type": "object",
"required": ["city"], # 使用 required 字段告诉 Kimi 大模型哪
"properties": {
"city": {
"type": "string",
"description": "The name of the city to get weather information for.",
},
},
}
}
},
{
"type": "function",
"function": {
"name": "get_paper",
"description": """
通过openAlex进行检索文献,当用户请求你进行文献查找时，调用此工具。
""",
"parameters": { # 使用 parameters 字段来定义函数接收的参数
"type": "object",
"required": ["search","filters"], # 使用 required 字段告诉 Kimi 大模型哪
"properties": {
"search": {
"type": "string",
"description": "结合上下文，用户要查找的论文的主题",
},
"filters": {
"type": "dict",
"description": """结合上下文，用户要查找的文献的筛选条件，只有以下这些过滤条件：abstract.search,alternate_host_venues.id,alternate_host_venues.license,alternate_host_venues.version,author.id,author.orcid,authorships.author.id,authorships.author.orcid,authorships.institutions.country_code,authorships.institutions.id,authorships.institutions.ror,authorships.institutions.type,cited_by,cited_by_count,cites,concept.id,concepts.id,concepts.wikidata,display_name,display_name.search,doi,from_created_date,from_publication_date,from_updated_date,has_doi,has_oa_accepted_or_published_version,has_oa_submitted_version,host_venue.id,host_venue.issn,host_venue.publisher,ids.mag,ids.openalex,ids.pmcid,ids.pmid,institution.id,institutions.country_code,institutions.id,institutions.ror,institutions.type,is_oa,is_paratext,is_retracted,journal.id,mag,oa_status,open_access.is_oa,open_access.oa_status,openalex,openalex_id,pmcid,pmid,publication_date,publication_year,raw_affiliation_string.search,referenced_works,related_to,title.search,to_publication_date,type.例如：{"is_oa": "true"，"type":"review"}"""
}
},
}
}
},
]
openalex = OpenAlex()
def get_paper(search):
# print(search)
grouped_concepts_list = openalex.get_list_of_works(search=search["search"],pages=[i for i in range(1, 2)],filters=search["filters"],sort={"relevance_score":"desc",})
concepts_list = list(grouped_concepts_list)
title, doi = [], []
for id in concepts_list[0]["results"]:
title.append(id["title"])
doi.append(id["doi"])
save = {"title":title, "doi":doi}
return save
MAX_CALLS = 30 # 设置最大函数调用次数
num_calls = 0 # 初始化调用计数器
messages = [
{"role": "system",
"content": "你是一个文献搜索助手，返回用户查找的文献的名称与doi号"},
{"role": "user", "content": "使用openAlex查找有关carbon material的2024年的文章"} # 在提问中要求 Kimi 大模型联网搜索
]
finish_reason = None
tool_map = {
"get_paper": get_paper,
}
while finish_reason is None or finish_reason == "tool_calls":
if num_calls >= MAX_CALLS: # 超过最大调用次数时候终止
break
completion = client.chat.completions.create(
model="moonshot-v1-8k",
messages=messages,
tool_choice="auto",
temperature=0.3,
tools=tools,
)
print(completion.choices[0].model_dump_json(indent=4))
choice = completion.choices[0]
finish_reason = choice.finish_reason
if finish_reason == "tool_calls": # <-- 判断当前返回内容是否包含 tool_calls
messages.append(choice.message) # <-- 我们将 Kimi 大模型返回给我们的 assistant 消息也添加到上下文中，以便于下次请求时 Kimi 大模型能理解
for tool_call in choice.message.tool_calls: # <-- tool_calls 可能是多个，因此我们使用循环逐个执行
tool_call_name = tool_call.function.name
tool_call_arguments = json.loads(tool_call.function.arguments) # <-- arguments 是序列化后的 JSON Object，我们需要使用 json.loads 反序列化一下
tool_function = tool_map[tool_call_name] # <-- 通过 tool_map 快速找到需要执行哪个函数
tool_result = tool_function(tool_call_arguments)
messages.append({
"role": "tool",
"tool_call_id": tool_call.id,
"name": tool_call_name,
"content": json.dumps(tool_result), # <-- 我们约定使用字符串格式向 Kimi 大模型提交工具调用结果，因此在这里使用 json.dumps 将执行结果序列化成字符串
})
messages.append({"role": "user", "content": "请按相关性进行排序"} )
num_calls += 1
print(choice.message.content) # <-- 在这里，我们才将模型生成的回复返回给用户
print(messages)
completion = client.chat.completions.create(
model="moonshot-v1-8k",
messages=messages,
temperature=0.3,
tools=tools,
)
print(completion.choices[0].model_dump_json(indent=4))
choice = completion.choices[0]
print(choice.message.content)