Split llms-full.txt into llms-chat-apps.txt and llms-agents.txt

[jhaaaa]

Split LLM ingestion bundles into llms-chat-apps.txt and llms-agents.txt and update generator in generate_llms_full.py to build three targeted files from docs/pages

Introduce a configurable generate_llms_txt in generate_llms_full.py that assembles content from selected docs/pages subdirectories and generates llms-chat-apps.txt, llms-agents.txt, and llms-full.txt; update the workflow to ignore llms-*.txt and add an npm script to run the generator; revise chat-apps intro docs to reference the new files.

📍Where to Start

Start with the __main__ script block and the generate_llms_txt function in generate_llms_full.py.

---

📊 Macroscope summarized 2abc30b. 1 file reviewed, 3 issues evaluated, 3 issues filtered, 0 comments posted

🗂️ Filtered Issues

llms/generate_llms_full.py — 0 comments posted, 3 evaluated, 3 filtered

• line 24: os.makedirs(output_dir) is called outside the surrounding try block and without exist_ok=True. This can raise a FileExistsError or PermissionError due to a race between the existence check and directory creation, or if a file already exists at output_dir. Because it is outside the try, the exception will terminate the script without the intended error handling/logging. Move the creation inside the try or use os.makedirs(output_dir, exist_ok=True) and catch exceptions. [ Already posted ]
• line 78: Error detection for content uses isinstance(content, str) and content.startswith("Error:"). This will falsely classify a legitimate markdown file whose content begins with "Error:" as unreadable and skip it. Use a structured return (e.g., (ok, content)), exceptions, or a sentinel type/value to distinguish read errors from valid file content. [ Already posted ]
• line 91: The function logs print(f"Successfully processed {len(processed_files)} Markdown files into {output_file}") even when processed_files is empty. The calling script treats an empty result as a failure and exits with error. This creates contradictory success/failure messaging for the same operation, confusing operators and automated log parsers. Adjust logging to report zero-file cases as a warning or failure consistently with the caller. [ Low confidence ]

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
docs-xmtp-org	Ready	Preview	Comment	Nov 13, 2025 1:47am