Granite Retrieval Agent provides developers with a flexible platform to build retrieval-augmented generative AI agents that combine semantic search and large language models. Users can ingest documents from diverse sources, create vector embeddings, and configure Azure Cognitive Search indexes or alternative vector stores. When a query arrives, the agent retrieves the most relevant passages, constructs context windows, and calls LLM APIs for precise answers or summaries. It supports memory management, chain-of-thought orchestration, and custom plugins for pre- and post-processing. Deployable with Docker or directly via Python, Granite Retrieval Agent accelerates the creation of knowledge-driven chatbots, enterprise assistants, and Q&A systems with reduced hallucinations and enhanced factual accuracy.

RecurSearch is an open-source Python library designed to improve Retrieval-Augmented Generation (RAG) and AI agent workflows by enabling recursive semantic search. Users define a search pipeline that embeds queries and documents into vector spaces, then iteratively refines queries based on prior results, applies metadata or keyword filters, and summarizes or aggregates findings. This step-by-step refinement yields higher precision, reduces API calls, and helps agents surface deeply nested or context-specific information from large corpora.

Embedditor is a cutting-edge, open-source tool designed as an efficient MS Word equivalent for embedding vectors. It offers a user-friendly interface for editing LLM vector embeddings, enabling users to upload, join, split, and edit content in various file formats. The aim is to optimize vector search capabilities, ensuring better performance and more precise search results. This tool provides significant flexibility and control over embedding processes, making it a valuable addition to any vector search and language model workflow.