NLWeb Prompt Customizer
Configures and optimizes NLWeb LLM prompts and Schema.org-specific behaviors for agentic commerce workflows.
This skill empowers developers to fine-tune the NLWeb framework's LLM interactions by managing structured XML configuration files. It provides deep control over prompt templates, JSON output contracts through the returnStruc property, and Schema.org type inheritance. By leveraging this skill, developers can improve RAG synthesis, refine tool routing, and implement domain-specific logic for e-commerce agents, ensuring high-quality, deterministic AI responses tailored to specific site requirements.
Características Principales
0132 GitHub stars
02Customizes <promptString> templates and <returnStruc> JSON schemas
03Optimizes LLM answer quality through decontextualization and ranking tuning
04Supports localization of prompt text while maintaining strict output parsing
05Manages Schema.org type inheritance and per-type prompt overrides
06Provides debugging strategies for multi-step agentic LLM call sites
Casos de Uso
01Resolving LLM output failures by tightening returnStruc contracts and prompt instructions
02Localizing e-commerce AI agents for international markets without breaking JSON parsing
03Improving product ranking and search relevance for specific Schema.org domains