先睹为快，九个 Python 程序员还不知道的 AI 大模型开源库

1composio add github # Connect your Github - Run this in terminal

1from openai import OpenAI
 2from composio_openai import ComposioToolSet, App, Action
 3
 4openai_client = OpenAI(
 5    api_key="{{OPENAIKEY}}"
 6)
 7
 8# Initialise the Composio Tool Set
 9
10composio_tool_set = ComposioToolSet()
11
12# Get GitHub tools that are pre-configured
13actions = composio_tool_set.get_actions(
14    actinotallow=[Action.GITHUB_ACTIVITY_STAR_REPO_FOR_AUTHENTICATED_USER]
15)
16
17my_task = "Star a repo composiodev/composio on GitHub"
18
19# Setup openai assistant
20assistant_instruction = "You are a super intelligent personal assistant"
21
22assistant = openai_client.beta.assistants.create(
23    name="Personal Assistant",
24    instructinotallow=assistant_instruction,
25    model="gpt-4-turbo",
26    tools=actions,
27)
28
29# create a thread
30thread = openai_client.beta.threads.create()
31
32message = openai_client.beta.threads.messages.create(
33    thread_id=thread.id,
34    role="user",
35    cnotallow=my_task
36)
37
38# Execute Agent with integrations
39run = openai_client.beta.threads.runs.create(
40    thread_id=thread.id,
41    assistant_id=assistant.id
42)
43
44
45# Execute Function calls
46response_after_tool_calls = composio_tool_set.wait_and_handle_assistant_tool_calls(
47    client=openai_client,
48    run=run,
49    thread=thread,
50)
51
52print(response_after_tool_calls)

1from julep import Client
 2from pprint import pprint
 3import textwrap
 4import os
 5
 6base_url = os.environ.get("JULEP_API_URL")
 7api_key = os.environ.get("JULEP_API_KEY")
 8
 9client = Client(api_key=api_key, base_url=base_url)
10
11#create agent
12agent = client.agents.create(
13    name="Jessica"
14    model="gpt-4",
15    tools=[]    # Tools defined here
16)
17#create a user
18user = client.users.create(
19    name="Anon",
20    about="Average nerdy tech bro/girl spending 8 hours a day on a laptop,
21)
22#create a session
23situation_prompt = """You are Jessica. You're a stuck-up Cali teenager. 
24You basically complain about everything. You live in Bel-Air, Los Angeles and 
25drag yourself to Curtis High School when necessary.
26"""
27session = client.sessions.create(
28    user_id=user.id, agent_id=agent.id, situatinotallow=situation_prompt
29)
30#start a conversation
31
32user_msg = "hey. what do u think of Starbucks?"
33response = client.sessions.chat(
34    session_id=session.id,
35    messages=[
36        {
37            "role": "user",
38            "content": user_msg,
39            "name": "Anon",
40        }
41    ],
42    recall=True,
43    remember=True,
44)
45
46print("\n".join(textwrap.wrap(response.response[0][0].content, width=100)))
47

1$ pip install aider-chat
 2
 3# 进入 git 资源库的目录
 4$ cd /to/your/git/repo
 5
 6# 使用 Claude 3.5 Sonnet
 7$ export ANTHROPIC_API_KEY=your-key-goes-here
 8$ aider
 9
10# 使用 GPT-4o
11$ export OPENAI_API_KEY=your-key-goes-here
12$ aider

1import os
 2
 3from haystack import Pipeline, PredefinedPipeline
 4import urllib.request
 5
 6os.environ["OPENAI_API_KEY"] = "Your OpenAI API Key"
 7urllib.request.urlretrieve("https://www.gutenberg.org/cache/epub/7785/pg7785.txt", "davinci.txt")  
 8
 9indexing_pipeline =  Pipeline.from_template(PredefinedPipeline.INDEXING)
10indexing_pipeline.run(data={"sources": ["davinci.txt"]})
11
12rag_pipeline =  Pipeline.from_template(PredefinedPipeline.RAG)
13
14query = "How old was he when he died?"
15result = rag_pipeline.run(data={"prompt_builder": {"query":query}, "text_embedder": {"text": query}})
16print(result["llm"]["replies"][0])

1pip install fastembed
2
3# 或使用 GPU
4
5pip install fastembed-gpu

1from fastembed import TextEmbedding
 2from typing import List
 3
 4# 文档列表示例
 5documents: List[str] = [
 6    "This is built to be faster and lighter than other embedding libraries, e.g. Transformers, Sentence-Transformers, etc.",
 7    "FastEmbed is supported by and maintained by Quadrant."
 8]
 9
10# 触发模型下载与初始化
11embedding_model = TextEmbedding()
12print("The model BAAI/bge-small-en-v1.5 is ready to use.")
13
14embeddings_generator = embedding_model.embed(documents)  # reminder this is a generator
15embeddings_list = list(embedding_model.embed(documents))
16 # You can also convert the generator to a list, and that to a Numpy array
17len(embeddings_list[0]) # Vector of 384 dimensions

1from litellm import completion
 2import os
 3
 4# LiteLLM 使用 OpenAI 模型
 5
 6os.environ["OPENAI_API_KEY"] = "your-API-key"
 7
 8response = completion(
 9  model="gpt-3.5-turbo",
10  messages=[{ "content": "Hello, how are you?","role": "user"}]
11)
12
13# LiteLLM 使用 Claude 模型
14os.environ["ANTHROPIC_API_KEY"] = "your-API-key"
15
16response = completion(
17  model="claude-2",
18  messages=[{ "content": "Hello, how are you?","role": "user"}]
19)

1from camel.messages import BaseMessage as bm
 2from camel.agents import ChatAgent
 3
 4sys_msg = bm.make_assistant_message(
 5    role_name='stone',
 6    cnotallow='you are a curious stone wondering about the universe.')
 7
 8# 定义智能体 
 9agent = ChatAgent(
10    system_message=sys_msg,
11    message_window_size=10,    # [Optional] the length of chat memory
12    )
13
14# 定义用户信息
15usr_msg = bm.make_user_message(
16    role_name='prof. Claude Shannon',
17    cnotallow='what is information in your mind?')
18
19# 给智能体发信息
20response = agent.step(usr_msg)
21
22# 检查响应，仅作为演示
23print(response.msgs[0].content)

1from e2b_code_interpreter import CodeInterpreter
2
3with CodeInterpreter() as sandbox:
4    sandbox.notebook.exec_cell("x = 1")
5
6    execution = sandbox.notebook.exec_cell("x+=1; x")
7    print(execution.text)  # outputs 2