Load an Image URL into a Jinba Flow Workflow

Use INPUT_IMAGE_FILE to fetch an image from a direct URL. The value input accepts one URL or an array of URLs. The result is base64 image data: a string for one URL, or an array for multiple URLs. A one-element array also returns a single string.

Fetch an image

- id: load_image
  tool: INPUT_IMAGE_FILE
  input:
    - name: value
      value: "https://example.com/photo.png"
    - name: description
      value: "Photo to analyze"

Replace the URL with a reachable direct image URL. A successful step returns a base64 data URL in result; pass that result to an image-capable tool. description is optional. use_as_input defaults to true for API and MCP inputs, and optional defaults to false.

Analyze the loaded image

Pass the returned image data to an image-capable AI tool, as in the original pipeline example:

- id: analyze_image
  tool: GEMINI_INVOKE_WITH_IMAGE
  needs: [load_image]
  config:
    - name: version
      value: gemini-flash-latest
  input:
    - name: prompt
      value: "Describe this image and any visible text."
    - name: base64_image
      value: "{{steps.load_image.result}}"

Check steps.analyze_image.result.content for the description. Configure a Gemini key or use available Jinba credit.

Check image input failures

If no URL is supplied, the step fails with “Image file URL is required.” If fetching fails, check that the URL is reachable with HTTP or HTTPS and actually returns an image; an HTTP error status fails the step. An image larger than 20 MiB, or a download that takes longer than 5 minutes, also fails the step (see Download limits). When optional is true and no value is supplied, result is null; downstream steps must handle that case.

​Fetch an image

​Analyze the loaded image

​Check image input failures

Fetch an image

Analyze the loaded image

Check image input failures