TILDE

ColBERT is a fast and accurate retrieval model, enabling scalable BERT-based search over large text collections in tens of milliseconds. It relies on fine-grained contextual late interaction: it encodes each passage into a matrix of token-level embeddings. At search time, it embeds every query into another matrix and efficiently finds passages that contextually match the query using scalable vector-similarity (MaxSim) operators. These rich interactions allow ColBERT to surpass the quality of single-vector representation models while scaling efficiently to large corpora. The toolkit includes components for retrieval, reranking, evaluation, and response analysis, facilitating end-to-end workflows. ColBERT integrates with Pyserini for retrieval and provides integrated evaluation for multi-stage pipelines. It also includes a module for detailed analysis of input prompts and LLM responses, addressing reliability concerns with LLM APIs and non-deterministic behavior in Mixture-of-Experts.

Vectara is LLM-powered search-as-a-service. The platform provides a complete ML search pipeline from extraction and indexing to retrieval, re-ranking and calibration. Every element of the platform is API-addressable. Developers can embed the most advanced NLP models for app and site search in minutes. Vectara automatically extracts text from PDF and Office to JSON, HTML, XML, CommonMark, and many more. Encode at scale with cutting edge zero-shot models using deep neural networks optimized for language understanding. Segment data into any number of indexes storing vector encodings optimized for low latency and high recall. Recall candidate results from millions of documents using cutting-edge, zero-shot neural network models. Increase the precision of retrieved results with cross-attentional neural networks to merge and reorder results. Zero in on the true likelihoods that the retrieved response represents a probable answer to the query.