ChatGPT Conversation Parser

This is a Java 17 CLI tool that walks a directory tree of saved ChatGPT conversations (HTML exports) and produces nicely formatted text files that summarize what was said plus the attachment list per conversation.

Features

• Scans the provided root directory and every child folder for .html/.htm exports generated via Save Page As from ChatGPT.

• Extracts the conversation title plus each user/assistant message using Jsoup, and renders them as:

  user said : <text>
  
  chatgpt said : <text>
  

  Each question/answer pair is separated by an empty line, and the first user message mentions image attachments when present.

• Outputs one .txt file per HTML export (same base name) and logs the generated file paths.

• Every invocation writes under an output folder (default chatgpt-conversations-parser-OUTPUT next to the input) inside a timestamped subfolder (yyyyMMdd-HHmmss-SSS), or a custom location when a second argument is provided.

Requirements

• Java 17 (JDK 17+)
• Maven
• Network access only required for the first build to download Jsoup from Maven Central.

Build

mvn -DskipTests package

This produces:

• target/chatgpt-conversation-parser-0.1.0.jar
• target/chatgpt-conversation-parser-0.1.0-jar-with-dependencies.jar

The assembly jar includes Jsoup and can be executed directly.

Usage

java -jar target/chatgpt-conversation-parser-0.1.0-jar-with-dependencies.jar <root-directory> [output-directory]

• <root-directory> — required. The directory holding conversation subfolders or HTML exports. The tool searches the root itself plus every immediate subdirectory.
• [output-directory] — optional. When omitted, the tool creates chatgpt-conversations-parser-OUTPUT next to the root directory; otherwise it uses the provided path. Each execution adds a timestamped folder inside that base location and writes individual .txt files there.

Example

cd target
java -jar chatgpt-conversation-parser-0.1.0-jar-with-dependencies.jar "C:/path/to/ChatGPT/exports"

Outputs will look like:

• .../chatgpt-conversations-parser-OUTPUT/20260306-170012-123/CV - SCA1.txt
• .../chatgpt-conversations-parser-OUTPUT/20260306-170012-123/CV - SCA2.txt
• etc.

Each .txt mirrors the conversation and notes attachments so it can be archived, searched, or shared easily.

Troubleshooting

• If you receive Provided path is not a directory, double-check the root argument is the folder with the exports, not a single HTML file.
• If “No ChatGPT conversations were detected” appears, ensure there are .html files in the root or its subdirectories, and that they were saved via the browser’s “Save Page As” so the markup is parsable.
• For parsing tweaks, inspect src/main/java/com/chatgpt/conversation/parser/ConversationScanner.java and adjust the selectors (currently div[data-author], div[data-role=message], article, and div[class*=message]).

Extending

• Add unit tests under src/test/java when you formalize selectors.
• Improve attachment-to-message correlation by expanding ConversationScanner’s TODOs.
• If you need JSON/CSV output, replace the string formatting logic in ConversationParserApp.formatConversation.

Support

File issues or ideas on the repository issue tracker or reach out via the appropriate project channel.