👾 Gemini Chat API 🔑

If you want a cloud LLM inside ComfyUI but OpenAI's pricing irritates you, OmixGeminiPro ("👾 Gemini Chat API 🔑") is the Google-flavored entry in OmiXDev's chat trio. The default model is gemini-2.5-flash-lite, which is fast, cheap, and has a usable free tier - a fine engine for the "draft me a detailed prompt for this character and composition" job. Same basic shape as the OpenAI node: key in, message in, reply out.

How it works

The node uses Google's google.generativeai SDK: it configures with your key, spins up a GenerativeModel, and - the nice bit - actually respects send_history. When it's on, the 👤/🤖 log is rebuilt as user/model turns and passed to start_chat(history=...), so it holds a real conversation rather than the stateless single calls the OpenAI node makes.

One validation quirk to know: the node checks that your key starts with "AI". Gemini keys look like AIzaSy..., so that's usually fine - but if it doesn't match you get "❌ Invalid or missing Gemini API Key." instead of a call.

The inputs and outputs that matter

• api_key - your AIzaSy... key from aistudio.google.com. Same caveat as ever: it's saved in the workflow JSON, so don't hand out workflows with the key inside.
• user_message - the instruction or prompt draft, multiline.
• model - a hardcoded dropdown of just two: gemini-2.5-flash-lite (default) and gemini-exp-1206.

reset clears the in-session history. Outputs are last_reply and chat_history, both STRING - the second is the newline-joined log with 👤/🤖 markers, handy for display or saving.

Installing - the one trap

Install the pack normally (Manager: search OmiXdev Custom Nodes, or the clone steps below), then install the requirements:

cd ComfyUI/custom_nodes
git clone https://github.com/omixmaxdimo/OmiXDev.git
cd OmiXDev
pip install -r requirements.txt

And here's where it bites: the pack's requirements.txt lists google-genai, but the node's code imports google.generativeai. Those are two different packages - google-genai ships the google.genai module, not the one this node needs. A fresh install can load everything fine, then throw ModuleNotFoundError: No module named 'google.generativeai' the first time the node is imported. If you see that:

pip install google-generativeai

then restart. Ten-second fix, just annoying it's not in the requirements file.

Troubleshooting

• The stale model. gemini-exp-1206 was an experimental 2024 release, and Google retires experimental models. If you pick it and get a model-not-found error, that's why - stick with gemini-2.5-flash-lite.
• Shared history again. Like the rest of the pack, chat_history is class-level, so all your OmixGeminiPro instances share one log per session and it clears on restart. Hit reset if threads start bleeding together.
• Errors are silent strings. Exceptions are caught and returned as last_reply with a ❌ prefix. Keep a text preview on the output so a failed call doesn't invisibly feed your prompt encoder.

Pick this one when you want Google's cheap flash model doing your prompt drafting. The install hiccup is a one-time annoyance; after that it just works.