local file read tool

Instead of manually copy-pasting file contents into a prompt every time you want an LLM to look at something, this node hands the model the ability to read files from a folder itself, as one of its callable tools. The agent decides what to open and when, based on what the conversation actually needs - closer to how a coding assistant reads your repo than a fixed "load this one file" step.

What it does

Two fields: folder_path, which points at the directory you're giving the LLM read access to, and is_enable, the pack's standard toggle for pulling this tool in or out of an agent's available set without deleting the node. The default value on folder_path is worth reading literally - "folder path/to/you want to llm control" - awkward phrasing, but the instruction is clear: replace it with the actual folder you want the model to be able to read from.

One output, tool, a string that plugs into an LLM/agent node's tool-list input, the same pattern this pack uses across all its _tool nodes (DuckDuckGo Tool, Feishu Tool, and so on). Once it's in an agent's toolset, the model can choose to call it mid-conversation - "let me check what's in that folder" as agent behavior, rather than something your workflow does unconditionally on every run.

Where it fits

Give an agent this alongside its other tools and it goes from "answers based only on what's in the prompt" to "can go look at your files when relevant" - useful for building an assistant over a folder of notes, logs, or documents where you don't know in advance which file is going to be relevant to a given question. Compare with this pack's Word Vector Search node, which is the right choice when you specifically want RAG-style semantic retrieval over a fixed corpus; this tool is closer to giving the model a directory listing and a read command it can use as it sees fit.

Installing it

Standard pack install, nothing node-specific:

• ComfyUI Manager: search "comfyui_LLM_party", install, restart.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/heshengtao/comfyui_LLM_party, then pip install -r requirements.txt from inside the pack folder using ComfyUI's Python, then restart.

Common issues & troubleshooting

Scope this narrowly - it's a genuine security consideration, not just a config detail. This node hands the LLM read access to whatever folder_path points at, and once it's in an agent's toolset the model decides what to read within that scope on its own. Point it at the specific folder you actually want exposed - a project's docs directory, a notes folder - not your home directory or anything that might contain credentials, keys, or unrelated sensitive files. There's nothing in the node itself limiting what within that folder gets read once it's granted.

Agent never uses the tool. Check is_enable is on and the tool output is actually wired into the agent node's tool-list input - an unconnected tool node isn't part of the agent's available toolset no matter how well the folder_path is set.

folder_path still has the placeholder text, or points somewhere that doesn't exist. Both will mean the tool has nothing usable to read from - replace the default placeholder with a real, existing folder path before expecting the agent to get anything useful out of calling it.