Ollama Connection

Ollama Connection is the address book half of the ComfyBros LLM pair. It takes an endpoint URL and an API key and packages them into an OLLAMA_CONNECTION object - endpoint, key, and a pre-built Authorization: Bearer <key> header - that the pack's chat node consumes. Like its configuration sibling, it produces no visible output on its own; it's a struct you wire somewhere.

The default endpoint is revealing: https://api.runpod.ai/v2/YOUR_ENDPOINT_ID/runsync. So despite the "Ollama" name, this pack's LLM story is less "talk to your local Ollama server" and more "call an OpenAI-compatible chat endpoint - often one hosted on a RunPod serverless instance running an LLM." The word Ollama is doing a lot of heavy lifting; the actual contract is an OpenAI-style /v1/chat/completions endpoint. If you expected to point this at http://localhost:11434 and chat with your local llama, check what the consuming node actually does with the connection before you assume.

Inputs

• endpoint_url - the chat endpoint. The default is a RunPod runsync URL with YOUR_ENDPOINT_ID as a placeholder you must replace.
• api_key - the bearer token for that endpoint.

That's the entire input surface: two strings. The node builds the headers dict (with the Authorization header always included, even if the key is empty - unlike the pack's Endpoint Configuration node, which omits it when blank).

Output

One output: OLLAMA_CONNECTION. It's designed to pair with the pack's Ollama Converse node (which reads the endpoint and headers from it and constructs the chat request). The OLLAMA_CONFIG from Ollama Configuration slots in alongside it.

Installing it

ComfyBros installs like any custom node pack. ComfyUI Manager - search "ComfyBros" - or:

cd ComfyUI/custom_nodes
git clone https://github.com/turnbros/ComfyBros

Then restart ComfyUI. The README's requirements.txt doesn't exist; the declared deps are in pyproject.toml. The node lives under ComfyBros/LLM.

Gotchas

Two things to keep straight. First, this is a config builder, not a chat node - nothing visible happens when you run it. Second, the "Ollama" branding is loose; what you're really configuring is an OpenAI-compatible endpoint, and the default URL is a RunPod one you must customize. If you're pointing at a self-hosted server, make sure it actually speaks the OpenAI chat-completions dialect the consuming node expects. Get those two straight and this node becomes the boring, reliable "plug the endpoint in here" step it's meant to be.