volcengine/openviking is a context database for AI agents that bundles agent memory, knowledge RAG (finding documents and using them as evidence in answers), and skills into a single file system. It is a repository released by Volcengine (volcengine), and its latest release is v0.5.1.1. However, the given material contains nothing about what changed in that release. It is written in Python and licensed under AGPL-3.0.
Contexts are stored in a viking:// virtual file system so they can be browsed and edited like files, with per-directory summaries read first to judge what is relevant. The README states that it provides semantic search scoped to a project or memory subtree, session archiving, and a CLI and an SDK. The arrangement that bundles knowledge, memory, and skills into a single file system is the core of this project. It has 40459 stars.
That said, the given material contains no figures that would let one gauge search accuracy or processing performance. The project description calls itself a self-evolving context database, but does not explain what process that refers to. Of the scoring items, only momentum has a score of 7.8 out of 10; there are no freshness, technology, or quality scores. The repository was created on 2026-01-05, so the project is still young. AGPL-3.0 is a license whose source-disclosure conditions also apply when a modified version is provided over the network, so if you plan to build a service on top of it, you should check the conditions first. It is worth a look for the purpose of gathering agent context in one place, but what v0.5.1.1 changed cannot be known until the release notes are checked.