OpenCode

Let OpenCode send images to asAi.

OpenCode needs its own model metadata before it sends an image. The generator creates an @ai-sdk/openai-compatible provider section for https://serveras.click/v1, with attachment and modalities set from the live catalog.

Connect your key first

1. Run /connect in OpenCode.
2. Choose Other.
3. Enter provider ID asai, then paste your asai_sk_... API key.

Never paste a real key into opencode.json. Keep it in OpenCode's credential flow. Create or manage a key under API keys.

Generate your model config

Choose a model set and select Generate config.

Why image input needs this config

For every image-capable model, the generator sets attachment: true and modalities.input to ["text", "image"]. It sets image input to false for text-only models. It also preserves the exact branded asAi model ID, context limit, and customer output cap.

By default, select asai/gpt-6-sol under One model. Use All image models when you want every current vision option. Use Full catalog only when you want text-only and vision models in the same OpenCode provider.

Verify it works

1. Save your merged config and start a fresh OpenCode session.
2. Select an asai/... image model and attach a small image.
3. Ask a question that depends on visible pixels, such as a word printed in the image.
4. Check Logs if you need to confirm a request reached asAi.

Troubleshooting

• If OpenCode says a model cannot read images and no request appears in Logs, regenerate the config and make sure the selected model has image input.
• Do not copy image capability settings to a text-only model. Check the live Models catalog first.
• Start with a small image. Base64 image data increases request size.