Nodes/comfyui_LLM_party/searxng Search Tool
ComfyUI Node

searxng Search Tool

Give your LLM agent a private, self-hosted search engine

By heshengtao·Created 2 years ago·Updated 7 days ago· 2,321
searxng Search Tool
    • tool
    searxng_api_urlhttp://localhost:8080
    categoriesgeneral
    is_enabletrue

    Bing and Google search tools in comfyui_LLM_party need an API key and, past the free tier, a bill. SearXNG Search Tool takes a different route: SearXNG is a real, well-known open-source metasearch engine that you self-host, aggregating results from multiple search backends without tracking you or charging per query. This node wires an LLM agent up to your own SearXNG instance rather than a paid third-party API.

    Worth saying plainly since it's easy to assume otherwise: this isn't a hosted service ComfyUI can just reach out to. You need a SearXNG instance actually running somewhere this node can reach before it does anything.

    How it works

    The node sends a query to your SearXNG instance's API and returns results as a tool an LLM agent can call. Because it's just talking to a standard SearXNG deployment, any instance you control - a Docker container on the same machine, a box on your home network, a VPS - works as long as the URL is reachable from wherever ComfyUI is running.

    The inputs and outputs that matter

    • searxng_api_url (default http://localhost:8080) - where your SearXNG instance lives. This default only works if SearXNG is running on the same machine as ComfyUI - if you're on a remote or serverless GPU, localhost inside that environment won't reach a SearXNG instance running on your home network, and you'll need a real, publicly reachable URL instead.
    • categories - which SearXNG search vertical to query: general, images, videos, news, map, science, files, social_media, it, q&a, shopping, and more.
    • is_enable - the standard bypass toggle.

    Output: tool (STRING) - a tool descriptor, not search results directly. Wire it into the tool/tools input on the pack's LLM or agent node so the model can call it during a conversation.

    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 with ComfyUI's Python, then restart. This node itself needs no extra Python dependencies beyond the pack's shared install - what it actually needs is a running SearXNG server, which is a separate piece of infrastructure you stand up yourself (SearXNG ships its own Docker Compose setup) before this node has anything to talk to.

    Common issues & troubleshooting

    Every call fails or times out. Almost always searxng_api_url pointing at nothing reachable. Confirm you actually have a SearXNG instance running, that it's the API endpoint (not just the web UI URL) it expects, and - if ComfyUI is running remotely - that the URL is one that environment can actually reach, not localhost.

    Results come back empty. Try the query directly against your SearXNG instance's own web interface first - if it returns nothing there too, the issue is upstream (SearXNG's configured search engines, not this node).

    You'd rather not run your own server. If self-hosting SearXNG is more infrastructure than you want, the pack's Bing and Google-backed search tools trade that setup for an API key instead - pick based on whether you'd rather manage a server or a subscription.

    Category大模型派对(llm_party)/工具(tools)/联网(Networking)

    Inputs (3)

    NameTypeDefaultDescription
    searxng_api_urlSTRINGhttp://localhost:8080
    categoriesCOMBOgeneral12 options: general, images, videos, news, music, map, +6
    is_enableBOOLEANtrue

    Outputs (1)

    NameTypeDescription
    toolSTRING