Function createMultimodalIndexerFromResolver

Create a MultimodalIndexer that reuses providers from the voice pipeline's SpeechProviderResolver and an optional VisionPipeline.

This is the recommended way to instantiate a multimodal indexer in applications that also use the voice pipeline — it ensures both subsystems share the same STT and vision providers instead of requiring separate configuration.

Error handling

• If the resolver has no configured STT provider, resolveSTT() will throw. This function catches that error and simply leaves the STT slot empty — audio indexing will throw at call time, not at construction time. This makes the factory safe to call even when STT is not configured.

• If neither visionPipeline nor visionProvider is provided, image indexing will throw at call time.

Throws

If embeddingManager or vectorStore is missing (propagated from MultimodalIndexer constructor).

Example

// Full setup: shared STT + vision
const indexer = createMultimodalIndexerFromResolver({
  resolver: speechResolver,
  visionPipeline: pipeline,
  embeddingManager,
  vectorStore,
});

// Vision-only (no audio indexing)
const visionIndexer = createMultimodalIndexerFromResolver({
  visionPipeline: pipeline,
  embeddingManager,
  vectorStore,
});

// STT-only (no image indexing)
const audioIndexer = createMultimodalIndexerFromResolver({
  resolver: speechResolver,
  embeddingManager,
  vectorStore,
});
Copy