Vibe Coding Discoversolo_jev
View on GitHubPer-row semantic judgements over large relational tables on a decision model (Jev)
SOLO reorders rows and fields in structured datasets to increase prefix-cache reuse when sending per-row semantic decisions to an LLM. It includes a table/JSON execution client, layout diagnostics, benchmarks, and vLLM decision-model adapters.
Use Cases
Run semantic decisions over relational table rowsClassify customer requests against shared policiesReorder structured inputs to improve LLM prefix-cache reuseBenchmark decision-model throughput and cache effectiveness
Built With
- Language
- Python
- Frameworks
- vLLM
Tags
LLM inference optimization · prefix caching · structured data · decision models · table processing · layout optimization