🤖 Gemini API is a ComfyUI node in the ArchAi3d/Edit/VLM category from ComfyUI-ArchAi3d-Qwen. Inputs: name, prompt, model, model_override. Outputs: STRING. Run…

Professional AI Interior Design Toolkit for ComfyUI. Custom Qwen-VL nodes with 4 design modes: Text-to-Design, Mood Board, Reference Control, and Room Cleaning. Includes 5 optimized nodes with advanced strength controls and 23 preferred aspect ratios.

ComfyUI-ArchAi3d-Qwen

name

prompt

model

model_override

system_prompt

api_key

image1

image2

image3

image4

temperature

top_p

top_k

max_tokens

seed

thinking_budget

use_cache

ArchAi3D_Gemini

response

🤖 Gemini API

Name	Type	Default	Description
name	STRING	gemini_prompt	Identifier name for this input (used by web interface)
prompt	STRING	Describe this image in detail.	The prompt to send to Gemini
model	COMBO	gemini-2.5-flash	Select Gemini model (can be overridden by model_override)
model_overrideopt	STRING		Connect ArchAi3D Gemini Model node here to override model selection
system_promptopt	STRING		System instructions to guide model behavior
api_keyopt	STRING		API key (saved to config after first use)
image1opt	IMAGE		—
image2opt	IMAGE		—
image3opt	IMAGE		—
image4opt	IMAGE		—
temperatureopt	FLOAT	1.000–2	Temperature (0=deterministic, 1=balanced, 2=creative)
top_popt	FLOAT	0.950–1	Top-p nucleus sampling (0.95 recommended)
top_kopt	INT	401–100	Top-k token selection
max_tokensopt	INT	81921–65536	Maximum output tokens
seedopt	INT	00–2147483647	Random seed for reproducibility (0 = random)
thinking_budgetopt	INT	00–24576	Thinking token budget for Gemini 2.5/3 models (0=off, 1024-24576=on). Model uses these tokens to reason before responding.
use_cacheopt	BOOLEAN	true	Use cached response if inputs haven't changed (saves API calls)

🤖 Gemini API

Inputs (17)

Outputs (1)