YepCode Executor: Serverless Code Execution for AG2 Agents#

import os

from autogen import ConversableAgent, LLMConfig
from autogen.coding import YepCodeCodeExecutor

# Ensure you have your API tokens set
assert os.getenv("YEPCODE_API_TOKEN"), "Please set YEPCODE_API_TOKEN environment variable"
assert os.getenv("OPENAI_API_KEY"), "Please set OPENAI_API_KEY environment variable (or configure another LLM provider)"

# Initialize YepCode executor
yepcode_executor = YepCodeCodeExecutor(
    timeout=120,  # Timeout in seconds
    remove_on_done=False,  # Keep execution logs for debugging
    sync_execution=True,  # Wait for execution to complete
)

# Create an agent with YepCode executor
code_executor_agent = ConversableAgent(
    "code_executor_agent",
    llm_config=False,  # Turn off LLM for this agent
    code_execution_config={"executor": yepcode_executor},
    human_input_mode="NEVER",
)

print("YepCode executor agent created successfully!")

# Test basic Python execution
task = """
Run this Python code:
```python
import datetime
import math

now = datetime.datetime.now()
print(f"Current time: {now}")
print(f"Square root of 144: {math.sqrt(144)}")

# Return a message
return "Hello from YepCode serverless execution!"
```
"""

reply = code_executor_agent.generate_reply(messages=[{"role": "user", "content": task}])
print(reply)

# Configure LLM (using OpenAI as example)
llm_config = LLMConfig(
    model="gpt-5-nano",
    api_key=os.environ["OPENAI_API_KEY"],
)

# Create code writer agent
code_writer_system_message = """
You are a helpful AI assistant that writes and executes code using YepCode's serverless platform.
When writing code:
1. Use Python or JavaScript as appropriate for the task
2. Include all necessary imports and dependencies (they will be installed automatically)
3. Write complete, executable code blocks
4. Return meaningful results using 'return' statements
5. Add informative print/console.log statements to show progress

For data analysis tasks, prefer Python with pandas, numpy, matplotlib, etc.
For web scraping or API calls, you can use either Python (requests, beautifulsoup) or JavaScript (axios, cheerio).

Always wrap your code in appropriate code blocks with language specification.
Reply 'TERMINATE' when the task is completed successfully.
"""

with llm_config:
    code_writer_agent = ConversableAgent(
        "code_writer_agent",
        system_message=code_writer_system_message,
        max_consecutive_auto_reply=5,
        code_execution_config=False,  # This agent writes code but doesn't execute it
    )

# Update code executor agent to work with the writer
code_executor_agent = ConversableAgent(
    "code_executor_agent",
    llm_config=False,
    code_execution_config={"executor": yepcode_executor},
    human_input_mode="NEVER",
    is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE"),
)

print("AI coding team with YepCode executor ready!")

data_analysis_task = """
Create a dataset of 100 random sales records with the following columns:
- date (random dates from the last 12 months)
- product_category (choose from: Electronics, Clothing, Books, Home, Sports)
- sales_amount (random amounts between $10 and $500)
- region (choose from: North, South, East, West)

Then perform the following analysis:
1. Calculate total sales by category
2. Find the average sales amount by region
3. Identify the top-selling month
4. Create a summary report

Present the results in a clear, formatted output.
"""

chat_result = code_executor_agent.initiate_chat(
    code_writer_agent,
    message=data_analysis_task,
    max_turns=10,
)

print("\n" + "=" * 50)
print("Data Analysis Task Completed!")
print("=" * 50)

api_task = """
Fetch cryptocurrency data from the CoinGecko API and analyze it:

1. Get the current price data for the top 5 cryptocurrencies by market cap
2. Calculate the total market cap of these top 5 coins
3. Find which coins have gained or lost the most in the last 24 hours
4. Create a formatted summary table showing:
   - Coin name and symbol
   - Current price (USD)
   - 24h price change (%)
   - Market cap

Use the CoinGecko API: https://api.coingecko.com/api/v3/coins/markets?vs_currency=usd&order=market_cap_desc&per_page=5&page=1
"""

chat_result = code_executor_agent.initiate_chat(
    code_writer_agent,
    message=api_task,
    max_turns=10,
)

print("\n" + "=" * 50)
print("Cryptocurrency Analysis Completed!")
print("=" * 50)

from autogen.coding import CodeExecutorFactory

# Create YepCode executor using factory
factory_executor = CodeExecutorFactory.create({
    "executor": "yepcode",
    "yepcode": {
        "timeout": 90,
        "remove_on_done": True,  # Clean up after execution
        "sync_execution": True,
    },
})

# Create agent with factory-created executor
factory_agent = ConversableAgent(
    "factory_agent",
    llm_config=False,
    code_execution_config={"executor": factory_executor},
    human_input_mode="NEVER",
)

# Test quick calculation
math_task = """
Calculate the following using Python:
```python
import math

# Calculate compound interest
principal = 10000  # Initial investment
rate = 0.07        # 7% annual interest
time = 10          # 10 years
n = 4              # Compounded quarterly

amount = principal * (1 + rate/n)**(n*time)
interest_earned = amount - principal

print(f"Principal: ${principal:,.2f}")
print(f"Final Amount: ${amount:,.2f}")
print(f"Interest Earned: ${interest_earned:,.2f}")

return {
    'principal': principal,
    'final_amount': round(amount, 2),
    'interest_earned': round(interest_earned, 2)
}
```
"""

result = factory_agent.generate_reply(messages=[{"role": "user", "content": math_task}])
print(result)