What problem does it solve?
It helps you quickly create and iteratively improve a system prompt that gets LLMs to generate correct, validator-compliant code for your Langium DSL, especially when evaluation results reveal recurring failure patterns or missing language guidance.
Core Features & Use Cases
- System prompt bootstrapping: Generates an initial system prompt from your language descriptor when no prompt exists yet.
- Evaluation-driven refinement: Diagnoses gaps by inspecting evaluation runs and then guides targeted prompt improvements based on failure categories.
- Language-change resilience: Encourages updating descriptor context before refining so regenerated rules and constraints stay aligned with the prompt.
- Prompt quality hardening: Recommends adding validation rules, scoping/reference behavior, ordering by importance, and negative “common mistakes” examples to reduce syntactic and semantic failures.
Quick Start
Run lai-gen-sysprompt in your project and ask it to generate the missing system prompt from your descriptor so you can start evaluating immediately.