docs: add modalities configuration for custom providers

[slycrel]

docs: add modalities configuration for custom providers

Summary

Adds documentation for the modalities property that enables vision and multimodal support in custom OpenAI-compatible providers.

Problem

The modalities configuration property exists in the codebase and works correctly, but is not documented anywhere. This causes users to:

• Believe vision/multimodal support is unavailable for custom providers
• Try incorrect properties like supportsImageInput or capabilities.vision
• Create duplicate issues asking how to enable image support
• Potentially abandon OpenCode for other tools

Solution

Added a new "Multimodal Support" section to the custom provider documentation that:

• Documents the modalities property with a clear example
• Lists all supported input/output modalities (text, audio, image, video, pdf)
• Explains validation behavior when modalities are not configured
• Provides a practical example showing image input configuration

Changes

• File: packages/web/src/content/docs/providers.mdx
• Location: Added after the existing custom provider example, before the Troubleshooting section
• Format: Follows the existing documentation style with code examples and tips

Testing

• ✅ Configuration property exists in schema: packages/opencode/src/provider/models.ts:53-58
• ✅ Used in capability detection: packages/opencode/src/provider/provider.ts:769-773
• ✅ Covered by tests: packages/opencode/test/provider/provider.test.ts:1254-1257
• ✅ Verified working with real-world custom provider (TaxHawk API)

Example Configuration

{
"provider": {
"myprovider": {
"npm": "@ai-sdk/openai-compatible",
"models": {
"my-vision-model": {
"name": "My Vision Model",
"modalities": {
"input": ["text", "image"],
"output": ["text"]
}
}
}
}
}
}

Related

• PR Feature/OpenAI compatible reasoning #5531 mentions multimodal support for @ai-sdk/openai-compatible
• Test coverage: test/provider/provider.test.ts:1241-1280

Impact

This documentation will:

• Reduce support burden from users asking about vision support
• Improve discoverability of existing features
• Help users configure multimodal providers correctly on first try
• Prevent duplicate issues and confusion

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Based on my search, I found one potentially related PR:

PR #5531: Feature/OpenAI compatible reasoning
#5531

This PR is mentioned in the current PR's description as relating to multimodal support for @ai-sdk/openai-compatible. It appears to be a complementary feature addition for the same provider ecosystem.

However, this is not a duplicate—it addresses a different feature (reasoning capabilities) rather than documentation of modalities configuration. The current PR #9898 is a documentation-only change that documents existing functionality.

No duplicate PRs found that are currently addressing the same documentation scope of modalities configuration for custom providers.

[slycrel]

Related issue: #9897

[slycrel]

Closes #9897

[github-actions]

Closing this pull request because it has had no updates for more than 60 days. If you plan to continue working on it, feel free to reopen or open a new PR.

[danmilon]

This documentation update is still useful.