Skip to main content

Class: LLMVisionProvider

Defined in: packages/agentos/src/core/vision/providers/LLMVisionProvider.ts:119

Vision provider that delegates to a cloud LLM via generateText().

Satisfies the narrow IVisionProvider contract used by the MultimodalIndexer, allowing any vision-capable LLM to serve as the image description backend.

Example​

const provider = new LLMVisionProvider({ provider: 'openai' });
const indexer = new MultimodalIndexer({
  embeddingManager,
  vectorStore,
  visionProvider: provider,
});

Implements​

Constructors​

Constructor​

new LLMVisionProvider(config): LLMVisionProvider

Defined in: packages/agentos/src/core/vision/providers/LLMVisionProvider.ts:141

Create a new LLM vision provider.

Parameters​

config​

LLMVisionProviderConfig

Provider configuration specifying which LLM to use.

Returns​

LLMVisionProvider

Throws​

If config.provider is not specified.

Example​

const provider = new LLMVisionProvider({
  provider: 'anthropic',
  model: 'claude-sonnet-4-20250514',
});

Methods​

describeImage()​

describeImage(image): Promise<string>

Defined in: packages/agentos/src/core/vision/providers/LLMVisionProvider.ts:175

Generate a text description of the provided image using a cloud vision LLM.

The image is sent as a base64 data URL in a multimodal message to the configured provider. The LLM's response is returned as-is.

Parameters​

image​

string

Image as a URL string (https://...) or base64 data URL (data:image/png;base64,...).

Returns​

Promise<string>

Detailed text description of the image content.

Throws​

If the LLM call fails.

Throws​

If the LLM returns an empty response.

Example​

const description = await provider.describeImage(
  'data:image/png;base64,iVBORw0KGgoAAAA...'
);
console.log(description);
// "A golden retriever playing fetch on a sandy beach..."

Implementation of​

IVisionProvider.describeImage