multimodal_input

Overview ¶

Demonstrates multimodal prompt input via the opencodesdk.QueryContent one-shot helper.

This example shows three entry points:

1. opencodesdk.QueryContent — one-shot text + image prompt
2. opencodesdk.ImageFileInput — load an image from disk
3. opencodesdk.Text / Blocks — ergonomic block constructors

The legacy string-only opencodesdk.Query still works for plain text — QueryContent is only needed when you want to attach images, embedded resources, or other non-text content blocks.

Image support requires the agent to advertise the "image" prompt capability during ACP initialize; opencode advertises it when attached to a multimodal-capable model. If the capability isn't advertised, the SDK rejects image blocks with ErrCapabilityUnavailable before the prompt reaches opencode.

go run ./examples/multimodal_input