Veles

Veles operates as a local-first tool focused on building and maintaining project memory without external dependencies. Sessions are processed to extract facts and behavioral guidelines that the agent recalls automatically in future interactions. This approach reduces repetitive context setting while working within any chosen directory structure. The architecture remains provider agnostic allowing configuration for different models based on task type such as planning or summarization. Extensions come through installable skills that function as tools and modules that integrate into the agent lifecycle at specific points. A minimal core handles memory routing and tool registration with everything else added optionally. Deployment options include running as a persistent daemon accessible via API or attaching channels for external interfaces. Both organizational shared knowledge bases and individual personal assistants can leverage the same underlying memory and agent capabilities for varied workflows.