Index

Index turns your scattered documents into a searchable knowledge layer. Files are chunked, embedded with a high-dimensional model, and stored in a vector index so answers come back grounded in your own source material rather than a model's guesswork.

Each query runs retrieval first, then generation: the most relevant passages are pulled and handed to the model with citations back to the original document. Re-indexing is incremental, so adding or editing a document refreshes only what changed instead of rebuilding the whole corpus.

Retrieval-augmented answers typically land in the 200ms-800ms range per query once the index is warm, and grounding cuts hallucinated answers sharply versus a raw prompt. Quality depends far more on chunking and source hygiene than on model choice, and Index makes both observable.