06 Document File Browser

Every document pipeline starts the same way: you have files on disk and you need to get their paths into the graph. "06 Document File Browser" is the node that does that - it scans a folder on the machine running ComfyUI, filters by extension and name, and hands you a path (plus a whole batch of them as a list) ready for Ingest, Upload, or anything else that wants a file_path_link. The display name says "File Browser" and it's exactly that, but the useful part is that it's batch-aware: it's built to feed streaming pipelines, not just one file at a time.

How it works

Pick a root_mode from the dropdown: Documents, Comfy Input (the standard input folder), Home, Current Working Directory, Filesystem Root, or Custom Root (where you fill in custom_root). Then:

• recursive (default true) - descend into subfolders
• extensions (default .pdf,.docx,.txt,.md,.csv,.json,.log,.rtf) - comma/space-separated; * means all files
• search_text - optional case-insensitive filename/path filter
• max_results (default 500, up to 10k) and max_depth (default 6, up to 25) - cap the scan so you don't walk your whole home directory
• include_hidden (default false) - the scanner also skips .git, node_modules, venv-style dirs, and system folders out of the box, which keeps scans sane

The single-file outputs - file_path, filename, content_type, size_bytes, exists, browser_root - describe the currently selected file (use selected_file dropdown or manual_path to pick it). But for pipelines, the list outputs are the real deal: file_paths and filenames as lists, discovered_count, indexed_file_paths (numbered), plus the standard selectable tail. And there are typed links - file_path_link and filename_link - which are what you wire into "06 Ingest Text PDF / DOCX" or "06 Upload Document File" so the file travels around the graph with its identity intact.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/CentralFloridaAttorney/ComfyUI-ZMongo

Restart ComfyUI, or install "ComfyUI-ZMongo" via ComfyUI Manager. The README's minimal deps (pymupdf, python-docx, requests, python-dotenv) cover this node; the heavier requirements.txt extras are only needed if you use the embedding/agent features.

Where people get burned

The single biggest confusion: paths are resolved on the ComfyUI machine, not your browser. If ComfyUI runs on a remote box and you browse your laptop's files, the node won't find them. The error payloads hammer this point for a reason - "must exist on the machine running ComfyUI."

Second, a directory is not a file. This node's file_path outputs point at files (or the browser root when nothing's selected). If you wire browser_root into an Upload or Ingest node expecting it to "just handle the folder," you'll get a clean "that's a directory" rejection. Select a real file.

And third: with recursive on and a big tree, you can hit max_results (500) quickly and think the scan found nothing weird - you just hit the cap. Raise it or narrow extensions/search_text. Watch discovered_count and you'll see it coming.