Load File Folder

This is the node that turns "a folder full of documents" into "text an LLM can actually read." It's the entry point for most of comfyui_LLM_party's knowledge-base features - before you can do keyword search or embedding-based retrieval over your own files, something has to actually read those files off disk and hand back their contents as a single string. That's this node's entire job.

The mechanism is straightforward: point it at a directory, and it walks the folder, pulls the readable content out of the files it finds, and concatenates it into one text blob. It doesn't chunk, embed, or search anything itself - that's deliberately left to the nodes downstream of it. This pack splits knowledge-base work into a loader step and a retrieval step, and this node is purely the loader half. The retrieval half is usually one of two nodes: Load Keyword Searcher for free, no-API keyword matching, or the OpenAI Word Vector Search node if you want real semantic embedding search and are fine paying for (or self-hosting) the embedding calls.

Inputs that matter. folder_path is required and defaults to C://Users/ - obviously a placeholder from the author's own Windows machine, not a real suggestion, so set it to wherever your actual source documents live before running anything. is_enable is the pack's standard on/off toggle, letting you disable this loader without unplugging it from the graph. There's also an optional related_characters field with no default and no tooltip in the node's own schema - treat it as a way to scope or tag what gets loaded rather than something you need for basic use; leaving it blank is the safe default, and it's worth experimenting with directly in the node rather than guessing at exact behavior nobody's documented plainly. The single output is file_content, a STRING, which plugs straight into the file_content input on either of the retrieval nodes mentioned above.

Installing it. Load File Folder ships with the full comfyui_LLM_party pack - no separate install. Get it via ComfyUI Manager (search "comfyui_LLM_party") or cd ComfyUI/custom_nodes && git clone https://github.com/heshengtao/comfyui_LLM_party, then pip install -r requirements.txt from inside the project folder in your ComfyUI Python environment, and restart. This particular node needs no model download and no API key - it's just reading local files - so on its own there's little to misconfigure beyond pointing it at the right path.

Where it actually trips people up. The most common practical issue isn't the node itself, it's the path. folder_path needs to be a location the ComfyUI process can actually see - on your own machine that's trivial, but on a hosted or containerized execution environment, a path that only exists on your laptop obviously won't resolve, and you need to make sure your source files are somewhere the running executor's filesystem can reach before pointing this node at them. Beyond that, if the pack fails to load entirely rather than just this node misbehaving, that's the broader requirements.txt for this project - it's large, covering dozens of optional integrations at once, and a single version conflict there can take the whole node set down; the README's requirements_fixed.txt is the built-in fallback worth trying first.