Calling openai from Leo

[Karthik Suresh]

Hi,

I've been thinking about Leo as a more general information management
tool for a while - I keep being surprised by how useful it is, after 21
years of using the tool.

For anyone that's interested, I thought  about using ChatGPT to help
with scripting custom extensions to do jobs that might be useful, one of
which is calling the OpenAI api itself to work on nodes. The following
code is what it came up with and I've been finding it quite useful.

Cheers,

Karthik

import leo.core.leoGlobals as g

from openai import OpenAI
# client = OpenAI()

# Set your OpenAI API key
client = OpenAI(api_key = "<YOUR KEY HERE>")

def summarize_and_reformat(content):
    """
    Sends the given content to OpenAI to summarize and reformat it
using the updated API.
    """
    try:
        # OpenAI API request
        response = client.chat.completions.create(
            model="gpt-4o-mini",  # Use "gpt-3.5-turbo" or "gpt-4"
based on your access
            messages=[
                {"role": "system", "content": "You are a helpful
assistant. Summarize and reformat the text you are given."},
                {"role": "user", "content": content},
            ],
            temperature=0.7,
            max_tokens=2000,  # Adjust as needed
        )
        # Extract and return the response
        message = response.choices[0].message.content
        g.es(f"Response from openai: {message}")
        return message
    except Exception as e:
        g.es(f"Error while interacting with OpenAI: {e}", color="red")
        return None

def process_node_with_openai(c, p):
    """
    Summarizes and reformats the content of the current node and
creates a new sub-node with the result.
    """
    # Get the current node's content
    content = p.b.strip()
    if not content:
        g.es("The node is empty. Nothing to summarize.", color="red")
        return
    # g.es(f"(content in the pane: {content}")
    g.es("Sending content to OpenAI for summarization and
reformatting...", color="blue")
    # Call the OpenAI function
    summarized_content = summarize_and_reformat(content)

    if summarized_content:
        # Create a new sub-node with the summarized content
        new_node = p.insertAsLastChild()
        new_node.h = "Summarized and Reformatted Content"
        new_node.b = summarized_content
        g.es("New sub-node created with summarized and reformatted
text.", color="green")
    else:
        g.es("Failed to create sub-node. See error logs for details.",
color="red")

# Add the command to the Leo editor
def add_openai_command(c):
    command_name = 'summarize-and-reformat'
    c.k.registerCommand(command_name, lambda event:
process_node_with_openai(c, c.p))
    g.es(f"Command '{command_name}' added to Leo.", color="green")

g.plugin_signon(__name__)
add_openai_command(c)

[Edward K. Ream]

On Thu, Nov 28, 2024 at 3:58 PM Karthik Suresh <k.su...@jayamony.co.uk> wrote:

For anyone that's interested, I thought  about using ChatGPT to help
with scripting custom extensions to do jobs that might be useful, one of
which is calling the OpenAI api itself to work on nodes. The following
code is what it came up with and I've been finding it quite useful.

Thanks for this script, Karthik. I'll take a closer look at it soon.

Edward