Actions

​

Overview

• Python Code: Execute custom Python scripts in isolated containers for data processing, API integrations, and custom logic.
• API Call: Make HTTP requests to external APIs and services.

• fetch real-time data
• process data with custom scripts
• retrieve current information
• integrate with third-party tools

​

How Actions Work

1. When to Trigger This Action: The Guru checks if the user’s question matches your defined conditions
2. Parameter Extraction: Parameters are extracted from the user’s question based on your descriptions
3. Execution Decision: The action runs only if all required parameters can be extracted (or have default values)
4. Execution: Python code runs in an isolated container, or an API call is made to the configured endpoint
5. Response Handling: The Usage Instructions guide the AI on how to interpret and present the results

​

Creating an Action

1. Basic Information and Parameters
2. Action Type Configuration (Python Code or API Call)
3. Usage Instructions and Testing

​

Basic Information and Parameters

• Action Name: Provide a descriptive name that appears in your actions list
• When to Trigger This Action: Be specific about keywords, patterns, or user intents to ensure it only runs when appropriate (e.g., “When the user wants to guess the secret number”)

• API Call: Use them in your API endpoint, headers, body using {parameter_name} syntax
• Python Code: Use them as environment variables via os.environ.get()

• Name: Use only letters and underscores (e.g., username, user_id). Cannot conflict with secret names.
• Type: Choose from String (text), Number (numeric), or Boolean (true/false)
• What to Extract: Clear description for parameter extraction from user questions
• Required: Check if essential for the action to work
• Default Value: Optional fallback when parameter can’t be extracted

​

Usage Instructions and Testing

• Usage Instructions: Provide instructions for how the AI should interpret and use the response data (API response or Python stdout)
  • Example: “Use stdout to decide if the guess was correct”
  • Example: “Extract the user’s name and bio from the JSON response and present this information in a clear summary”
• Testing: Use the “Test Action” button to verify your configuration
  • If required parameters exist, you’ll be prompted to provide test values
  • Test results show success status, response data, and execution details

​

Action Types

• Python Code: Execute custom Python scripts in isolated containers
• API Call: Make HTTP requests to external APIs and services

​

Python Code Actions

• Isolated execution: Each execution runs in a fresh container that is destroyed after use
• Pre-installed libraries: Popular data science and HTTP libraries ready to use
• Results via stdout: Your code’s print() output is captured and returned to the Guru
• Secrets protection: All secrets are masked in output (supports plain text, base64, hex, and URL-encoded detection)

​

Sandbox Environment

​

Pre-installed Python Packages:

anthropic==0.76.0
beautifulsoup4==4.14.3
httpx==0.28.1
instructor==1.14.3
lxml==6.0.2
matplotlib==3.10.8
numpy==2.2.3
openai==2.15.0
opencv-python==4.12.0.88
pandas==2.3.3
pillow==12.1.0
pydantic==2.12.5
requests==2.32.5
scikit-learn==1.8.0
scipy==1.17.0
seaborn==0.13.2
urllib3==2.6.3

​

Execution Limits

​

Security Restrictions:

• Read-only filesystem (only /tmp is writable)
• /tmp mounted with noexec flag (cannot execute binaries)
• All Linux capabilities dropped
• Network access allowed (for API calls)

import os

# Access a parameter extracted from user's question
username = os.environ.get('USERNAME')

# Access a secret configured in Guru Settings
api_key = os.environ.get('MY_API_KEY')

print(f"Fetching data for {username}...")

​

Examples

import os

guess = int(os.environ.get('GUESS'))
secret_n = int(os.environ.get('SECRET_NUMBER'))

if guess == secret_n:
    print("Correct! You guessed the secret number!")
elif guess < secret_n:
    print("Too low! Try a higher number.")
else:
    print("Too high! Try a lower number.")

import os
import requests

api_key = os.environ.get('WEATHER_API_KEY')
city = os.environ.get('CITY', 'London')

response = requests.get(
    "https://api.weatherapi.com/v1/current.json",
    params={"key": api_key, "q": city}
)

data = response.json()
print(f"Weather in {city}: {data['current']['condition']['text']}")
print(f"Temperature: {data['current']['temp_c']}°C")

import pandas as pd
import requests

# Fetch users data from JSONPlaceholder API
response = requests.get("https://jsonplaceholder.typicode.com/users")
df = pd.DataFrame(response.json())

# Process and summarize
print(f"Total users: {len(df)}")
print(f"Columns: {', '.join(df.columns)}")
print("\nUser list:")
print(df[['id', 'name', 'email', 'phone']].to_string(index=False))

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ.get('OPENAI_API_KEY'))
user_text = "Hello, how are you?"

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": f"{user_text}"}]
)

print(response.choices[0].message.content)

​

API Call Actions

• HTTP Method: Select from GET, POST, PUT, PATCH, DELETE
• Endpoint URL: Enter the API endpoint. Use {parameter_name} or {SECRET_NAME} to insert parameter/secret values
  • Example: https://api.github.com/users/{username}
• Headers: Configure authentication or content type using {parameter_name} or {SECRET_NAME} syntax
  • Example: Authorization: Bearer {API_KEY}
• Request Body: For POST, PUT, and PATCH methods, include JSON payload supporting {parameter_name} or {SECRET_NAME} syntax:

{
  "test": "{TEST}",
  "new_address": "{new_address}"
}

​

Managing Actions

• View all configured actions with their name, type (“Python Code” or “API Call”), and status (Enabled/Disabled)
• Use the top action bar:
  • History: View execution history of all actions
  • Import: Import an action from a JSON configuration file (exported from another action)
  • Create an Action: Create a new action from scratch
• Click the menu icon (⋮) next to any action to:
  • Edit: Modify the action’s configuration
  • Export: Download the action configuration as a JSON file
  • Trigger Action: Manually run the action immediately
  • Disable: Temporarily disable without deleting
  • Delete: Permanently remove the action
• Enable or disable actions at any time
• Disabled actions won’t be triggered by user questions
• Note: If a secret used by an action is deleted, that action will be automatically disabled

​

Usage in the Answer

​

Secrets Management

• Secrets are encrypted at rest (AES-256) and masked in all output
• Access secrets in Python via os.environ.get('SECRET_NAME')
• Use secrets in API calls with {SECRET_NAME} syntax
• Secret names cannot conflict with parameter names
• Deleting a secret automatically disables all actions using it

• Python: api_key = os.environ.get('STRIPE_API_KEY')
• API Call: Authorization: Bearer {STRIPE_API_KEY}

​

Common Mistakes to Avoid

​

Next Steps