A model given tools and a goal, deciding for itself which steps to take and in what order.
Reprocessing a past date after the fact, usually because data arrived late or was restated.
An AI system that writes, edits and tests code across multiple files with limited supervision.
The chain of dependent tasks whose total duration determines the earliest possible completion.
Four measures of software delivery performance: deployment frequency, lead time, change failure rate and recovery time.
Changing a schema in backward-compatible steps so old and new code can run simultaneously.
Tying generated statements to specific source records that support them.
Requiring a person to review or approve before machine output takes effect.
A property whereby repeating an operation produces the same result as performing it once.
A model trained to generate and interpret text, capable of reasoning over context it is given.
A service routing all model calls through one controlled point.
An open protocol letting AI systems connect to external tools and data sources through a common interface.
Degradation in model performance as real-world data diverges from what it was trained on.
One deployment serving many customers whose data must remain separated.
Reusing the processed form of a repeated prompt prefix instead of paying to process it again.
Database-enforced rules restricting which rows a query can return.
A target for how reliable a service should be, expressed as a measurable threshold.
A table that automatically retains the history of every row, allowing queries as at any past moment.
The unit of text a model processes, roughly a word fragment.
Writing a message to the same database transaction as the data change, then relaying it to a broker separately.