A Programing Language for Data and Configuration!
summary
Proposes a new programming-language-style schema that packs type-safety, versioning, packaging, IDE support into data/config specs, with real newtypes, composable regex types, and logical constraints (e.g. percent ∈ [0,100], low<high). Prototypes emit a human-readable polyglot literal format, generate valid mocks by calling an SMT solver (Z3 SMT solver) to synthesize responses (it even produced a snow forecast for Phoenix), and ties to LLM workflows to auto-generate examples and enforce payment hard limits on agent-callable APIs; prototypes run and validate as described.
AI-written summary · about
chapters (13)
- 0:00Data configuration challenges
- 1:45Data errors impact
- 4:34Solving data problems
- 6:19Weather service data model
- 8:48BOSI data description language
- 10:10Enforcing data constraints
- 12:21Human-readable literal format
- 13:45Beyond API specifications
- 14:29Automated mocking
- 16:55AI programming with specs
- 18:38Guardrails for AI agents
- 20:49Conclusion and future
- 22:13Q&A
related talks
13mGradually Structured Data
21mDataEngBytes 2023 - SYD-T3-04 - Abhinav Goyal
50mThe Babel Fish Data: A Case Study
41mDatomic Auto Schemas
24mFrom Code To Models-as-Data: GEMS, a High-Level Language for Energy System Modelling
05mLightning Talk: Schema Inference
14mMaking Data Quality an integral part of developing Machine Learning and Data Products
40mFunctional Programming for Machine Learning and Data Pipelines