Auto Generated Agent Chat: Solving Tasks Requiring Web Info
AutoGen offers conversable agents powered by LLM, tool, or human, which can be used to perform tasks collectively via automated chat. This framework allows tool use and human participation through multi-agent conversation. Please find documentation about this feature here.
In this notebook, we demonstrate how to use AssistantAgent
and
UserProxyAgent
to perform tasks which require acquiring info from the
web: * discuss a paper based on its URL. * discuss about the stock
market.
Here AssistantAgent
is an LLM-based agent that can write Python code
(in a Python coding block) for a user to execute for a given task.
UserProxyAgent
is an agent which serves as a proxy for a user to
execute the code written by AssistantAgent
. By setting
human_input_mode
properly, the UserProxyAgent
can also prompt the
user for feedback to AssistantAgent
. For example, when
human_input_mode
is set to “TERMINATE”, the UserProxyAgent
will
execute the code written by AssistantAgent
directly and return the
execution results (success or failure and corresponding outputs) to
AssistantAgent
, and prompt the user for feedback when the task is
finished. When user feedback is provided, the UserProxyAgent
will
directly pass the feedback to AssistantAgent
.
Requirements
AutoGen requires Python>=3.8
. To run this notebook example, please
install autogen and docker:
pip install autogen docker
# %pip install "autogen" docker
Set your API Endpoint
The
config_list_from_json
function loads a list of configurations from an environment variable or
a json file.
import autogen
config_list = autogen.config_list_from_json(
"OAI_CONFIG_LIST",
filter_dict={
"model": ["gpt4", "gpt-4-32k", "gpt-4-32k-0314", "gpt-4-32k-v0314"],
},
)
llm_config = {
"timeout": 600,
"cache_seed": 42,
"config_list": config_list,
"temperature": 0,
}
It first looks for environment variable “OAI_CONFIG_LIST” which needs to be a valid json string. If that variable is not found, it then looks for a json file named “OAI_CONFIG_LIST”. It filters the configs by models (you can filter by other keys as well). Only the models with matching names are kept in the list based on the filter condition.
The config list looks like the following:
config_list = [
{
'model': 'gpt-4-32k',
'api_key': '<your OpenAI API key here>',
},
{
'model': 'gpt4',
'api_key': '<your Azure OpenAI API key here>',
'base_url': '<your Azure OpenAI API base here>',
'api_type': 'azure',
'api_version': '2024-02-01',
},
{
'model': 'gpt-4-32k-0314',
'api_key': '<your Azure OpenAI API key here>',
'base_url': '<your Azure OpenAI API base here>',
'api_type': 'azure',
'api_version': '2024-02-01',
},
]
You can set the value of config_list in any way you prefer. Please refer to this notebook for full code examples of the different methods.
Construct Agents
We construct the assistant agent and the user proxy agent. We specify
human_input_mode
as “TERMINATE” in the user proxy agent, which will
ask for human feedback when it receives a “TERMINATE” signal from the
assistant agent.
# create an AssistantAgent instance named "assistant"
assistant = autogen.AssistantAgent(
name="assistant",
llm_config=llm_config,
)
# create a UserProxyAgent instance named "user_proxy"
user_proxy = autogen.UserProxyAgent(
name="user_proxy",
human_input_mode="TERMINATE",
max_consecutive_auto_reply=10,
is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE"),
code_execution_config={
"work_dir": "web",
"use_docker": False,
}, # Please set use_docker=True if docker is available to run the generated code. Using docker is safer than running the generated code directly.
llm_config=llm_config,
system_message="""Reply TERMINATE if the task has been solved at full satisfaction.
Otherwise, reply CONTINUE, or the reason why the task is not solved yet.""",
)
Example Task: Paper Talk from URL
We invoke the initiate_chat()
method of the user proxy agent to start
the conversation. When you run the cell below, you will be prompted to
provide feedback after the assistant agent sends a “TERMINATE” signal at
the end of the message. If you don’t provide any feedback (by pressing
Enter directly), the conversation will finish. Before the “TERMINATE”
signal, the user proxy agent will try to execute the code suggested by
the assistant agent on behalf of the user.
# the assistant receives a message from the user, which contains the task description
user_proxy.initiate_chat(
assistant,
message="""
Who should read this paper: https://arxiv.org/abs/2308.08155
""",
)
Who should read this paper: https://arxiv.org/abs/2306.01337
--------------------------------------------------------------------------------
To determine who should read the paper, we need to first understand the content and context of the paper. We can do this by fetching the abstract of the paper from the provided URL and analyzing it.
Here is a Python script that uses the BeautifulSoup library to scrape the abstract of the paper from the webpage.
```python
# Python script to scrape the abstract of the paper
import requests
from bs4 import BeautifulSoup
def get_abstract(url):
response = requests.get(url)
soup = BeautifulSoup(response.text, 'html.parser')
abstract = soup.find('blockquote', attrs={'class': 'abstract mathjax'}).text.strip()
return abstract
url = "https://arxiv.org/abs/2306.01337"
abstract = get_abstract(url)
print(abstract)
```
Please run this script and provide the output. Based on the abstract, I can suggest who might be interested in reading this paper.
--------------------------------------------------------------------------------
>>>>>>>> USING AUTO REPLY...
>>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...
exitcode: 0 (execution succeeded)
Code output:
Abstract: Employing Large Language Models (LLMs) to address mathematical problems is an
intriguing research endeavor, considering the abundance of math problems
expressed in natural language across numerous science and engineering fields.
While several prior works have investigated solving elementary mathematics
using LLMs, this work explores the frontier of using GPT-4 for solving more
complex and challenging math problems. We evaluate various ways of using GPT-4.
Some of them are adapted from existing work, and one is MathChat, a
conversational problem-solving framework newly proposed in this work. We
perform the evaluation on difficult high school competition problems from the
MATH dataset, which shows the advantage of the proposed conversational
approach.
--------------------------------------------------------------------------------
Based on the abstract, the paper is about using Large Language Models (LLMs), specifically GPT-4, to solve complex mathematical problems. The paper introduces a new conversational problem-solving framework called MathChat and evaluates its performance on difficult high school competition problems from the MATH dataset.
Given this, the paper would be of interest to the following groups:
1. **Researchers in Artificial Intelligence and Natural Language Processing**: The paper discusses the use of a large language model (GPT-4) for problem-solving, which is a key research area in AI and NLP.
2. **Mathematicians and Math Educators**: The paper focuses on solving complex mathematical problems, so those with a background in mathematics might find the techniques and results interesting.
3. **Data Scientists and Machine Learning Engineers**: These professionals often use models like GPT-4 in their work and might be interested in new applications and techniques.
4. **Students studying AI, NLP, or Mathematics**: The paper could provide valuable insights for these students into how AI can be used in problem-solving.
5. **Developers working on AI-based chatbots or conversational agents**: The paper introduces a new conversational problem-solving framework, which could be of interest to these developers.
Please note that while the paper is likely to be of interest to these groups, the specific relevance will depend on the individual's specific interests and research needs.
TERMINATE
--------------------------------------------------------------------------------
>>>>>>>> NO HUMAN INPUT RECEIVED.
Example Task: Chat about Stock Market
user_proxy.initiate_chat(
assistant,
message="""Show me the YTD gain of 10 largest technology companies as of today.""",
)
Show me the YTD gain of 10 largest technology companies as of today.
--------------------------------------------------------------------------------
To get the YTD gain of the 10 largest technology companies, we need to do the following:
1. Identify the 10 largest technology companies. We can use the list of the largest technology companies by market capitalization. This list can change frequently, so we need to get the latest data. We can use web scraping to get this data from a reliable source.
2. Get the YTD gain for each of these companies. We can use a financial data API to get this data. Yahoo Finance is a popular source for this kind of data.
Here is a Python script that uses the BeautifulSoup library for web scraping and the yfinance library to get data from Yahoo Finance. This script will print the 10 largest technology companies and their YTD gains.
```python
# filename: ytd_gain.py
import requests
from bs4 import BeautifulSoup
import yfinance as yf
import pandas as pd
# Get the list of largest tech companies by market cap
url = "https://companiesmarketcap.com/technology/largest-technology-companies-by-market-cap/"
response = requests.get(url)
soup = BeautifulSoup(response.text, 'html.parser')
# Find the table with the data
table = soup.find('table')
# Get the ticker symbols of the 10 largest companies
tickers = []
for row in table.find_all('tr')[1:11]: # Skip the header row
ticker = row.find_all('td')[1].text
tickers.append(ticker)
# Get the YTD gain for each company
ytd_gains = {}
for ticker in tickers:
ticker_data = yf.Ticker(ticker)
hist = ticker_data.history(period="ytd")
start_price = hist.iloc[0]['Close']
end_price = hist.iloc[-1]['Close']
ytd_gain = (end_price - start_price) / start_price * 100
ytd_gains[ticker] = ytd_gain
# Print the results
for ticker, ytd_gain in ytd_gains.items():
print(f"{ticker}: {ytd_gain:.2f}%")
```
To run this script, save it to a file named `ytd_gain.py`, then run it with the command `python ytd_gain.py`.
Please note that this script requires the BeautifulSoup, requests, yfinance, and pandas libraries. If you don't have these libraries installed, you can install them with the command `pip install beautifulsoup4 requests yfinance pandas`.
--------------------------------------------------------------------------------
>>>>>>>> USING AUTO REPLY...
>>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...
exitcode: 1 (execution failed)
Code output:
Traceback (most recent call last):
File "ytd_gain.py", line 18, in <module>
for row in table.find_all('tr')[1:11]: # Skip the header row
AttributeError: 'NoneType' object has no attribute 'find_all'
--------------------------------------------------------------------------------
I apologize for the error. It seems that the script failed to find the table in the webpage. This could be due to changes in the webpage structure or because the page uses JavaScript to load the table, which BeautifulSoup can't handle.
Let's try a different approach. We'll use the yfinance library to get the data for the 10 largest technology companies. These are currently: Apple (AAPL), Microsoft (MSFT), Alphabet (GOOGL), Amazon (AMZN), Facebook (FB), Tesla (TSLA), Nvidia (NVDA), Paypal (PYPL), Adobe (ADBE), and Netflix (NFLX).
Here's the updated Python script:
```python
# filename: ytd_gain.py
import yfinance as yf
# List of the 10 largest tech companies by market cap
tickers = ['AAPL', 'MSFT', 'GOOGL', 'AMZN', 'FB', 'TSLA', 'NVDA', 'PYPL', 'ADBE', 'NFLX']
# Get the YTD gain for each company
ytd_gains = {}
for ticker in tickers:
ticker_data = yf.Ticker(ticker)
hist = ticker_data.history(period="ytd")
start_price = hist.iloc[0]['Close']
end_price = hist.iloc[-1]['Close']
ytd_gain = (end_price - start_price) / start_price * 100
ytd_gains[ticker] = ytd_gain
# Print the results
for ticker, ytd_gain in ytd_gains.items():
print(f"{ticker}: {ytd_gain:.2f}%")
```
To run this script, save it to a file named `ytd_gain.py`, then run it with the command `python ytd_gain.py`.
Please note that this script requires the yfinance library. If you don't have this library installed, you can install it with the command `pip install yfinance`.
--------------------------------------------------------------------------------
>>>>>>>> USING AUTO REPLY...
>>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...
exitcode: 1 (execution failed)
Code output:
FB: No data found, symbol may be delisted
Traceback (most recent call last):
File "ytd_gain.py", line 13, in <module>
start_price = hist.iloc[0]['Close']
File "/home/vscode/.local/lib/python3.9/site-packages/pandas/core/indexing.py", line 931, in __getitem__
return self._getitem_axis(maybe_callable, axis=axis)
File "/home/vscode/.local/lib/python3.9/site-packages/pandas/core/indexing.py", line 1566, in _getitem_axis
self._validate_integer(key, axis)
File "/home/vscode/.local/lib/python3.9/site-packages/pandas/core/indexing.py", line 1500, in _validate_integer
raise IndexError("single positional indexer is out-of-bounds")
IndexError: single positional indexer is out-of-bounds
--------------------------------------------------------------------------------
I apologize for the error. It seems that the script failed to get the data for Facebook (FB). This could be due to a temporary issue with the Yahoo Finance API.
Let's modify the script to handle this situation. We'll add a try-except block to skip any companies for which we can't get data.
Here's the updated Python script:
```python
# filename: ytd_gain.py
import yfinance as yf
# List of the 10 largest tech companies by market cap
tickers = ['AAPL', 'MSFT', 'GOOGL', 'AMZN', 'FB', 'TSLA', 'NVDA', 'PYPL', 'ADBE', 'NFLX']
# Get the YTD gain for each company
ytd_gains = {}
for ticker in tickers:
try:
ticker_data = yf.Ticker(ticker)
hist = ticker_data.history(period="ytd")
start_price = hist.iloc[0]['Close']
end_price = hist.iloc[-1]['Close']
ytd_gain = (end_price - start_price) / start_price * 100
ytd_gains[ticker] = ytd_gain
except IndexError:
print(f"Could not get data for {ticker}")
# Print the results
for ticker, ytd_gain in ytd_gains.items():
print(f"{ticker}: {ytd_gain:.2f}%")
```
To run this script, save it to a file named `ytd_gain.py`, then run it with the command `python ytd_gain.py`.
Please note that this script requires the yfinance library. If you don't have this library installed, you can install it with the command `pip install yfinance`.
--------------------------------------------------------------------------------
>>>>>>>> USING AUTO REPLY...
>>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...
exitcode: 0 (execution succeeded)
Code output:
Could not get data for FB
AAPL: 46.41%
MSFT: 37.63%
GOOGL: 51.13%
AMZN: 56.22%
TSLA: 131.46%
NVDA: 237.68%
PYPL: -16.49%
ADBE: 59.79%
NFLX: 43.41%
--------------------------------------------------------------------------------
The script has successfully fetched the Year-to-Date (YTD) gain for the largest technology companies, except for Facebook (FB). The YTD gains are as follows:
- Apple (AAPL): 46.41%
- Microsoft (MSFT): 37.63%
- Alphabet (GOOGL): 51.13%
- Amazon (AMZN): 56.22%
- Tesla (TSLA): 131.46%
- Nvidia (NVDA): 237.68%
- Paypal (PYPL): -16.49%
- Adobe (ADBE): 59.79%
- Netflix (NFLX): 43.41%
Unfortunately, the script could not fetch the data for Facebook (FB). This could be due to a temporary issue with the Yahoo Finance API or the ticker symbol might have changed.
Please note that these values are subject to change as the stock market fluctuates.
TERMINATE
--------------------------------------------------------------------------------
FB -> META
--------------------------------------------------------------------------------
Thank you for the update. Facebook Inc. has indeed rebranded to Meta Platforms Inc. and its ticker symbol has changed from FB to META on the NASDAQ stock exchange.
Let's update the script to reflect this change:
```python
# filename: ytd_gain.py
import yfinance as yf
# List of the 10 largest tech companies by market cap
# Updated Facebook (FB) to Meta Platforms (META)
tickers = ['AAPL', 'MSFT', 'GOOGL', 'AMZN', 'META', 'TSLA', 'NVDA', 'PYPL', 'ADBE', 'NFLX']
# Get the YTD gain for each company
ytd_gains = {}
for ticker in tickers:
try:
ticker_data = yf.Ticker(ticker)
hist = ticker_data.history(period="ytd")
start_price = hist.iloc[0]['Close']
end_price = hist.iloc[-1]['Close']
ytd_gain = (end_price - start_price) / start_price * 100
ytd_gains[ticker] = ytd_gain
except IndexError:
print(f"Could not get data for {ticker}")
# Print the results
for ticker, ytd_gain in ytd_gains.items():
print(f"{ticker}: {ytd_gain:.2f}%")
```
To run this script, save it to a file named `ytd_gain.py`, then run it with the command `python ytd_gain.py`.
Please note that this script requires the yfinance library. If you don't have this library installed, you can install it with the command `pip install yfinance`.
--------------------------------------------------------------------------------
>>>>>>>> USING AUTO REPLY...
>>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...
exitcode: 0 (execution succeeded)
Code output:
AAPL: 46.34%
MSFT: 37.55%
GOOGL: 51.08%
AMZN: 56.17%
META: 133.67%
TSLA: 131.02%
NVDA: 237.38%
PYPL: -16.55%
ADBE: 59.71%
NFLX: 43.44%
--------------------------------------------------------------------------------
The script has successfully fetched the Year-to-Date (YTD) gain for the largest technology companies, including Meta Platforms (formerly Facebook). The YTD gains are as follows:
- Apple (AAPL): 46.34%
- Microsoft (MSFT): 37.55%
- Alphabet (GOOGL): 51.08%
- Amazon (AMZN): 56.17%
- Meta Platforms (META): 133.67%
- Tesla (TSLA): 131.02%
- Nvidia (NVDA): 237.38%
- Paypal (PYPL): -16.55%
- Adobe (ADBE): 59.71%
- Netflix (NFLX): 43.44%
Please note that these values are subject to change as the stock market fluctuates.
TERMINATE
--------------------------------------------------------------------------------
>>>>>>>> NO HUMAN INPUT RECEIVED.