memory-setup

Persistent memory is enabled by setting memorySearch.enabled = true in your agent config, specifying a vector provider like voyage, and defining sources. This allows the agent to retain context across multiple runs.

Vector search configuration requires setting a provider, defining memory sources, and adjusting parameters like indexMode, minScore, and maxResults. This controls how the agent retrieves relevant context from stored memories.

Long-term memory requires creating a MEMORY.md file and a dedicated memory/ directory. The directory should be organized into subfolders for logs, projects, groups, and system data.

Yes, daily logs are supported as part of the memory workflow. When stored in the structured memory directory, they help the agent recall past interactions and maintain consistency across conversations.

memorySearch requires configuring enabled, provider, sources, indexMode, minScore, and maxResults. These parameters define the vector search behavior and determine how context is retrieved during agent sessions.

The best way is to enable memorySearch with a vector provider like voyage, set sources to memory and sessions, and create a structured MEMORY.md file with a memory directory for persistent storage.