Load Wikipedia

Not every grounding problem needs a full RAG pipeline. Sometimes you just want one solid reference article dropped into an LLM's context so it stops hallucinating a date or a definition. Load Wikipedia is the lightweight option in comfyui_LLM_party's knowledge-base tooling - no embedding model, no vector database, no chunking step. Give it a query, get back article text.

Compare it to the pack's other knowbase nodes: Load File reads something already on disk, Save Embeddings Database builds a proper searchable index for repeated queries - this one is the quick, single-shot version of the same idea, good for a one-off fact check rather than a knowledge base you'll query repeatedly.

How it works

You give it a search query, it looks up matching Wikipedia content and returns the article text as a string. There's no chunking, embedding, or ranking involved - this is direct lookup, not retrieval over an index.

The inputs and outputs that matter

• query - the term or question to look up (default placeholder query).
• is_enable - the pack's standard bypass toggle.

Output: file_content (STRING) - the retrieved article text, ready to drop into an LLM node's context or combine with other text using one of the pack's String Combine nodes.

How to install it

Search comfyui_LLM_party in ComfyUI Manager and install, then restart. Manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/heshengtao/comfyui_LLM_party.git

Run pip install -r requirements.txt from the pack's own folder using ComfyUI's Python, then restart. This node needs an internet connection to reach Wikipedia but no API key and no local model - one of the lighter-weight nodes to get running in a pack whose combined install otherwise covers LLM clients, embeddings, OCR, and TTS libraries.

Common issues & troubleshooting

Gets the wrong article, or a disambiguation page instead of the one you meant. Wikipedia lookup by a raw query string is inherently approximate - an ambiguous term (a name that's also a place, a word with multiple meanings) can match a different article than the one you had in mind. Be as specific as you reasonably can in query (add a disambiguating word) rather than relying on a single generic term.

No result comes back. If the query genuinely has no matching Wikipedia article - a very obscure or made-up term - this node has nothing to return; that's expected behavior, not a failure, since it can only surface what actually exists on Wikipedia.

You need something more current than what's on Wikipedia. This node is a reference lookup, not a live web search - for anything time-sensitive (recent news, current prices, this week's events), reach for one of the pack's actual search tools (Bing, SearXNG) instead, which query the live web rather than an encyclopedia snapshot.