The Community Edition v1.4.3 of axoviq-ai/synthadoc was released on October 20, 2026. This version is a release that includes security hardening and improvements. Anthropic prompt caching was applied to reduce the cost and time to first response for repeated queries, and the release notes put the input cost reduction at 80–90%. The high-risk and medium-risk items found in an external static analysis audit were also resolved in this version.
Synthadoc is an open-source LLM knowledge compilation engine that turns original documents into structured wikis. The resulting wiki has a local-first structure and can be managed and improved on its own without tools. The repository description presents this approach as a transparent, human-readable alternative to traditional RAG (retrieval-augmented generation). The repository currently has 1593 stars.
Prompt caching is a way of storing repeated input content and reusing it. In queries where the same input recurs, cost and time to first response drop together. Static analysis is an inspection that finds problems by scanning only the source without executing the code, and this release handled the items that came out of that audit.
That said, 80–90% is only a figure based on input cost as presented in the release notes, and the available records do not show which query lengths and patterns it was measured on. No figures on response speed or accuracy are given either, and there is no information about the license or supported languages. This update is better viewed as a version that may lower cost and latency in environments where the same queries are sent repeatedly.