AI Agent Open Source: A Comprehensive Guide for Developers

1# This is a simplified, conceptual example.
2# Actual AutoGPT code is much more complex.
3
4class AutoGPTAgent:
5    def __init__(self, llm_model):
6        self.llm = llm_model
7        self.memory = {}
8
9    def execute_task(self, task_description):
10        # 1. Break down the task into subtasks using LLM.
11        subtasks = self.llm.generate_subtasks(task_description)
12
13        # 2. Execute each subtask.
14        results = []
15        for subtask in subtasks:
16            action = self.llm.determine_action(subtask, self.memory)
17            result = self.perform_action(action)
18            results.append(result)
19            self.memory[subtask] = result #Store result to memory
20
21        # 3. Summarize the results using LLM.
22        final_result = self.llm.summarize_results(results)
23        return final_result
24
25    def perform_action(self, action):
26        # Placeholder for actual action execution (e.g., web search).
27        print(f"Performing action: {action}")
28        return f"Result of action: {action}"
29

1from langchain.llms import OpenAI
2from langchain.chains import LLMChain
3from langchain.prompts import PromptTemplate
4
5# Initialize LLM
6llm = OpenAI(temperature=0.9)
7
8# Define prompt template
9template = "What is a good name for a company that makes {product}?"
10prompt = PromptTemplate(template=template, input_variables=["product"])
11
12# Create LLMChain
13chain = LLMChain(llm=llm, prompt=prompt)
14
15# Run the chain
16product = "colorful socks"
17company_name = chain.run(product)
18
19print(f"The suggested name for a company making {product} is: {company_name}")
20

1# Conceptual representation of BabyAGI task execution
2
3import os
4import openai
5
6openai.api_key = os.getenv("OPENAI_API_KEY")
7
8objective = "Write a short summary about the history of AI"
9
10task_list = ["Research the history of AI"]  # Initial task
11
12while task_list:
13    # Get the first task
14    task = task_list.pop(0)
15    print(f"Executing task: {task}")
16
17    # Use LLM to execute the task
18    response = openai.Completion.create(
19        engine="text-davinci-003",
20        prompt=f"{objective}
21{task}
22Result:",
23        temperature=0.7,
24        max_tokens=150
25    )
26
27    result = response.choices[0].text.strip()
28    print(f"Task Result: {result}")
29
30    # Use LLM to create new tasks based on the result
31    new_tasks = openai.Completion.create(
32        engine="text-davinci-003",
33        prompt=f"{objective}
34Based on the result: {result}
35What are the next tasks to do?",
36        temperature=0.7,
37        max_tokens=100
38    ).choices[0].text.strip().split("
39")
40
41    # Add new tasks to the list
42    task_list.extend(new_tasks)
43
44    print(f"New Tasks: {new_tasks}")
45

1class SimpleAgent:
2    def __init__(self):
3        self.environment = None
4
5    def perceive(self):
6        # Get information from the environment
7        pass
8
9    def think(self):
10        # Decide what to do
11        pass
12
13    def act(self):
14        # Perform an action
15        pass
16
17    def run(self):
18        while True:
19            self.perceive()
20            self.think()
21            self.act()
22