The Shot Database
A structured database records every shot's configuration, outcome, and links to its telemetry, the human-and-machine index into the archive.
Every shot, indexed
Beyond the raw telemetry archive, the fabric keeps a structured shot database: one record per shot capturing its configuration, key outcomes, regime, events, and pointers into the archive, feature store, and vector database. It is how a person or a copilot finds the right shot without sifting petabytes of raw series.
What a shot record holds
- Machine and configuration: breeder or burner, target parameters, actuator settings.
- Outcome features: peak instantaneous Q, achieved profiles, whether a disruption or plug event occurred.
- Event log: precursors that fired, overrides, failsafe actions.
- Links: to raw telemetry, engineered features, and the shot's embedding.
Querying by physics
Because records are structured, the database answers physics questions directly: shots at a given breeder shape and current, shots where the burner's plug density crossed a threshold, shots that preceded a disruption. For fuzzier queries it hands off to the embedding search, which finds physically similar shots by nearest neighbor.
The organizing spine
The shot database is the spine that ties the fabric together: the archive holds the bytes, the feature store serves the features, the vector database enables semantic search, and the shot database is the index that connects them with governance and lineage attached. It serves both machines under one schema and is a design specification for the stack of machines under construction from Q2 2027.