Aktilot

Aktilot is a self-hosted, open-source RAG platform for chatting with your documents on your own infrastructure. It ingests files such as PDF, Word, and plaintext and builds searchable knowledge bases using hybrid BM25 plus vector retrieval, combining keyword and semantic search. Every response answers with sources, exposing the source documents, chunk positions, and relevance scores, and the platform surfaces full retrieval traces and timing data for observability. It is durable by design, using Temporal workflows for fault tolerance and automatic retries, and supports multiple LLM providers including local models. Projects are isolated with separate vector stores per knowledge domain, and deployment is Docker-based so no data leaves your servers. Aktilot is best for engineering teams and organizations handling sensitive documents that need data sovereignty and cited answers rather than a hosted SaaS. It is open-source with no per-seat or per-query fees, though running it requires your own infrastructure. Because it is self-hosted, teams take on setup, hosting, and maintenance in exchange for keeping data on-premises.

Best for engineering teams that need self-hosted, cited RAG over sensitive documents with full data sovereignty.