Generate Text and Images with Gemini in Jinba Flow

Use GEMINI_INVOKE to get a text response from Gemini, GEMINI_INVOKE_WITH_IMAGE to analyze images, or GEMINI_GENERATE_IMAGE to generate images.

Interact with Gemini with prompts or by sending images. The Gemini tool also allows you to send prompts and receive structured responses.

• GEMINI_INVOKE
  • Send prompts to Gemini AI
• GEMINI_INVOKE_WITH_IMAGE
  • Send images to Gemini AI for analysis
• GEMINI_GENERATE_IMAGE
  • Generate images from text descriptions using Gemini image generation models (Nano Banana)

Connect with a Gemini API key

For further details, click here.

If you do not have an API key, you can still run Gemini tools by using Jinba API credit. If you’d like to use your own API key, you can obtain one from the official Gemini website.

Note: Treat API keys as sensitive information and never commit them to public repositories.

Get a response from Gemini about an image

- id: analyze_image
  name: analyze_image
  tool: GEMINI_INVOKE_WITH_IMAGE
  config:
    - name: version
      value: gemini-flash-latest
    - name: token
      value: "{{secrets.GEMINI_API_KEY}}"
  input:
    - name: prompt
      value: Please analyze this image in detail. If there is any text, extract it as well. Please provide as much detail as possible.
    - name: base64_image
      value: YOUR_IMAGE_URL_OR_BASE64
- id: structure_information
  name: structure_information
  tool: OPENAI_INVOKE
  needs: [analyze_image]
  config:
    - name: version
      value: gpt-5.2
  input:
    - name: prompt
      value: |
        Please organize the following image analysis results into a structured format.

        ==Image Analysis Results==
        {{steps.analyze_image.result.content}}

        ==Instructions==
        1. Extract and organize important information by category
        2. Format the output suitable for spreadsheets
        3. Output in the following JSON array format:
        [
          ["Category", "Item", "Content"],
          ["Text", "Detected Text 1", "..."],
          ["Object", "Detected Object 1", "..."],
          ...
        ]

Generate an image with Gemini

GEMINI_GENERATE_IMAGE generates an image from a text description and returns a downloadable image URL (result.image_url, valid for 24 hours). Five models are available: gemini-3.1-flash-image (Nano Banana, fast and efficient, the default), gemini-nano-banana-2.1 (Nano Banana 2.1, its successor at half the image price), gemini-3.1-flash-lite-image (Nano Banana Lite, same image price as Nano Banana 2.1), gemini-3-pro-image (Nano Banana Pro, high-fidelity professional output) and gemini-2.5-flash-image (retiring 2027-03-15 — steps that select it keep running on it, and are priced at its rate, until that date; from then on they run on gemini-3.1-flash-lite-image and are priced at that model’s rate). The aspect_ratio input accepts 1:1 (Square, the default), 3:4 (Portrait), 4:3 (Standard), 9:16 (Mobile Portrait), or 16:9 (Landscape).

- id: generate_image
  name: generate_image
  tool: GEMINI_GENERATE_IMAGE
  config:
    - name: model
      value: gemini-3.1-flash-image
  input:
    - name: prompt
      value: A watercolor illustration of Mount Fuji at sunrise, with cherry blossoms in the foreground
    - name: aspect_ratio
      value: "16:9"

Notes

• GEMINI_INVOKE, GEMINI_INVOKE_WITH_IMAGE, and GEMINI_INVOKE_WITH_FILE accept a temperature config value (0–2, default 0.7). It is only sent to Gemini models older than 3.6 (for example gemini-2.5-flash). Gemini 3.6 and later fix their sampling parameters, so for those models and for the auto-advancing -latest aliases (gemini-flash-latest, the default) the value is omitted from the request and has no effect.