Hi HN,
Personal notes on a constrained DSL to make LLMs generate reliable decision logic for quant (no arbitrary Python, no hallucinations). Schema-driven prompts, validation loop, deterministic execution.
Article + diagrams + public schema: https://github.com/myinvestpilot/ai-architecture/blob/main/d...
Repo (complete 4-article series, EN/ZH): https://github.com/myinvestpilot/ai-architecture
Feedback welcome!
Show HN: Constrained DSL for Reliable LLM Decisions
github.com