Each demo in this section is a finished piece built on SurrealDB Agent Memory: a corpus ingested into a Context, and pages drawn from what the memory holds. Every demo opens with the result, embedded so you can use it on the page, and follows it with how it was built, so you can adapt the same steps to your own data.
Demos
Calgary 1912-1913 explorer
Walk the city of Calgary built from two volumes of Henderson's Calgary Directory, where every answer a resident gives comes from a scope you can inspect.
Robin Hood character graph
Howard Pyle's The Merry Adventures of Robin Hood ingested one chapter at a time, and drawn as the cast known at each point in the book.
Pompeii election notices
127 campaign notices from the walls of Pompeii, and chat answers about who backed whom, with every citation linked to its notice.
The Domesday Book, 1086
52 entries of the Domesday survey for one Norfolk hundred, and chat answers on lordship, wealth and mills, each checked against the data.
Vulcan, Alberta: an article's history
24 years of edits to one Wikipedia article in Agent Memory, and agents answering questions about what the article used to say. A small model with Agent Memory scores at least as well as a large one reading the history, for less than half the cost.
Rome, 753 BC to AD 1453
Every year from the founding of Rome to the fall of Constantinople on a map, with rulers and control of places read back as value histories, and an export that turns any period into a prompt for an LLM.
Night Owls: twenty years of forum memory
Seven agents on a simulated bulletin board react to five real events from 1999 to 2019. Each post shows what Agent Memory held on its author at that moment, and private notes from 1999 still come back in 2019.
Sundance Lake Loop: who can read what on a pipeline project
A fictional pipeline project in six chapters. Each employee's assistant reads Agent Memory through its own scoped key, remembers its owner's conversations, and leaves a note when its owner says something the listener should not hear.