
NeMo Guardrails is an open-source Python package for adding programmable guardrails around LLM calls. Use it to block unsafe, malicious, off-topic, or policy-violating user inputs, retrieved RAG content, tool usage, and model responses.
Defines structured data models with Pydantic BaseModel. Use when defining LLM outputs or inputs for RAG, LlamaIndex structured outputs, or when the user mentions Pydantic, structured output, or response schemas.
Crawls websites and extracts (scrapes) web page content including text, tables, lists, and images into clean LLM-friendly markdown. Use when crawling websites, scraping web content, extracting web page data for RAG/LLMs, or when the user mentions crawl4ai, web scraping, or deep crawling.
Used to supplement the data LLMs can use when answering questions by supplying them with custom data generated and managed by llama-index.
Builds simple user interfaces in Python. Use when creating chatbots, dashboards, or data apps with Streamlit—especially chat UIs with st.chat_message and st.chat_input, or when the user mentions Streamlit.
Connects to Supabase-hosted PostgreSQL and pgvector vector databases. Use when storing or querying vector embeddings, building RAG pipelines with Supabase, using vecs for similarity search, or when the user mentions Supabase, pgvector, or vector databases.