🔱 Artha Gemini Question

Artha Gemini Question is the pack's plainest AI node: ask a question, get an answer. It's a single-turn Q&A call to Gemini with no memory, no conversation, and no pretense of being a chatbot - the description says it straight: "the context will not be preserved as this node is not suitable for dialogue purposes." Every run is a fresh, stateless question with whatever you typed in the question field, and the answer comes back through a single output.

That sounds trivial, and in one sense it is. But in a node graph it's exactly what you want: deterministic, inspectable, per-call AI. There's no hidden state to confuse you, no conversation history to accidentally carry over between runs - the answer you get is purely a function of the question you asked.

What you set

• question - your question, multiline. This is the whole job.
• api_key, model (default gemini-2.5-flash), max_tokens (5000), temperature (0.7) - the pack's standard plumbing.

One output: answer, a plain string.

Under the hood it's a straightforward generate_content call with a fixed system prompt ("You are an intelligent ai assistant."). There's no image input and no extra context - just question in, answer out.

Where it earns its place

The statelessness is the feature. A few genuinely useful patterns:

• Conditional text in a workflow. Ask Gemini a yes/no or short-answer question, then wire answer into a text-switch or logic node to steer the graph. "Is this prompt suitable for a children's book?" → route accordingly.
• Batch interrogation. Same question, different batches, deterministic-ish answers you can compare.
• A cheap second opinion. Stuck between two prompt styles? Ask. It's a Gemini text call, which is the cheapest thing the pack does.

The honest limits: no conversation, no context beyond the question, and no image understanding (for that you want Artha Gemini Vision). And while it's cheap per call, it's not free - every execution hits the API. If you're asking the same question every run of a batch, you'll pay for each one.

The failure mode is the pack standard: an empty answer with the error printed to the ComfyUI console, so empty output means "read the console," not "try again." Key resolution is the usual chain - api_key field, then api.json in the pack folder, then the GEMINI_API_KEY environment variable - and api.json is the version that keeps your key out of shared workflow files.

Install

It rides along with the whole 🔱 pack:

cd ComfyUI/custom_nodes
git clone https://github.com/Cyrostar/ComfyUI-Artha-Nodes
cd ComfyUI-Artha-Nodes
pip install -r requirements.txt

Restart ComfyUI, or install "Artha" via ComfyUI Manager. Get a key from Google AI Studio - this node is just an API call, so the key is the whole setup.