Python script to convert ChatGPT exports into markdown files, suitable for importing into Obsidian et al

gpt2mdconvert.py

# Convert ChatGPT conversations into markdown
# source: https://dev.to/gavi/convert-chatgpt-conversations-to-obsidian-markdown-format-p61

# Improvements (tried to match original coding conventions)
# 2023.08.10: Fix bug accessing 'parts' in node['message']['content']['parts']
# 2024.01.06: Updated get_conversation to handle non-string content parts, i.e. images (which are not currently included in the ChatGPT exports)
# 2024.01.20: Added link to conversation in ChatGTP using conversation_id, and also set time stamps correctly, esp. on macOS

import json
import os
import sys
import time
import subprocess
import platform

def get_conversation(node_id, mapping, list):
    node = mapping[node_id]
    if 'message' in node and node['message'] is not None:
        if 'content' in node['message'] and node['message']['content'] is not None:
            if 'parts' in node['message']['content'] and node['message']['content']['parts'] is not None:
                content_parts = node['message']['content']['parts']
                if len(content_parts) > 0:
                    text_parts = []
                    for part in content_parts:
                        if isinstance(part, str):
                            text_parts.append(part)
                        elif isinstance(part, dict) and 'asset_pointer' in part:
                            # Replace with a Markdown image placeholder
                            text_parts.append(f"![Image]({part['asset_pointer']})")
                        # Add more conditions if there are other types of content
                        else:
                            print("Non-string content_parts found:", content_parts)
                            sys.exit(1)  # Exit the script
                    author_role = node['message']['author']['role']
                    list.append(f"## {author_role}\n {''.join(text_parts)}")
    for child_id in node.get('children', []):
        get_conversation(child_id, mapping, list)

def set_timestamps(file_path, create_time, update_time):
    # Setting the creation time (macOS specific)
    if platform.system() == 'Darwin':
        subprocess.run(["SetFile", "-d", time.strftime("%m/%d/%Y %H:%M:%S", time.localtime(create_time)), file_path])
    # Updating the modification time
    os.utime(file_path, (update_time, update_time))

if __name__ == '__main__':
    folder_path = "output"
    # Check if the directory exists
    if not os.path.isdir(folder_path):
        # If not, create the directory
        os.makedirs(folder_path)
    with open('conversations.json') as f:
        data = json.loads(f.read())
        for item in data:
            if 'title' in item and item["title"] is not None:
                title = item["title"]
            else:
                title = "New chat"
            title = title.replace("/","_").replace('"','')
            if title == "New chat":
                title = "New chat " + str(int(item["create_time"]))
            if 'conversation_id' in item and item["conversation_id"] is not None:
                url = "https://chat.openai.com/c/" + item["conversation_id"]
            else:
                url = "None"
            root_node_id = [node_id for node_id, node in item['mapping'].items() if node['parent'] is None][0]
            list = []
            get_conversation(root_node_id, item['mapping'],list)
            file_path = f'{folder_path}/{title}.md'
            with open(file_path,'w') as outfile:
                outfile.write(f'Source: {url}\n')
                outfile.write('\n'.join(list))
            set_timestamps(file_path, item["create_time"], item["update_time"])