load redis memory

This is comfyui_LLM_party's memory node for when a single JSON file stops cutting it. Same job as the pack's file-based memory node - give your LLM a persistent conversation history instead of starting fresh every run - but backed by a real Redis server instead of a flat file, which is the right call the moment more than one thing might be reading or writing that history at once.

That's the actual reason to pick this over the JSON-file version: file-based memory works fine for one user testing locally, but it gets shaky the moment multiple runs might touch the same history concurrently - a file on disk has no real answer for "two processes writing at the same time." Redis does, and it's fast enough that using it for chat history doesn't add noticeable latency to a conversational agent. If you're building something more than a personal single-session assistant - several users, several concurrent conversations, anything approaching production - this is the node built for that, not the JSON one.

Inputs that matter. system_prompt is a required multiline STRING, same as the file-based memory node - the instructions framing the conversation. history_key (default "1") is what identifies a specific conversation thread inside Redis - the equivalent of the filename in the JSON version, except here it's a key in a shared store rather than a separate file, which is exactly what makes concurrent multi-user memory practical: different keys, same server. redis_host (default "localhost") and redis_port (default 6379, the standard Redis port) point at your actual Redis instance. clear_memo is a boolean, defaulting off - flip it on to wipe the history stored under that specific history_key, a clean way to reset one conversation thread without touching the node graph or any other thread's memory. Outputs mirror the file-based version: system_prompt, user_history, and history_key all pass through as STRINGs, feeding your LLM node its context.

Installing it. This ships with the full comfyui_LLM_party pack - no separate step, but it does need a Python Redis client, pulled in as part of requirements.txt. Get the pack 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 using your ComfyUI Python environment, and restart.

Where this actually breaks. The node itself is simple; the failure mode is almost always the Redis connection, not the node's logic. A connection error means Redis isn't reachable at redis_host:redis_port - either it isn't running, it's on a different host or port than the defaults, or you're on a hosted/remote executor where localhost doesn't point at a Redis server the way it would on your own machine. You need an actual, network-reachable Redis instance - local, self-hosted, or a managed service - before this node has anything to talk to. If memory seems to silently vanish between runs, double-check you're passing the same history_key each time; a different key is, by design, a different conversation with no shared history at all. And if the whole pack fails to import rather than just this node erroring, that's this project's large bundled requirements.txt - worth trying the README's requirements_fixed.txt before assuming Redis is the problem.