Guardrails - Pydantic-first schema enforcement and safety checks: coerce LLM outputs into typed objects with validators, retries, and fallbacks

TruLens - LLM evaluation library providing reference-free and ground-truth metrics for RAG/agents

RAGxplorer - Tiny library and Streamlit UI to visualize chunking, retrieval, reranking, and answers (helping explain RAG behavior)

Cloudflare AI Search - Edge-hosted RAG search that indexes PDFs/HTML/Office and exposes a simple query API with hybrid + vector retrieval

GuideLLM - CLI to generate realistic, configurable benchmark workloads against inference endpoints to test latency/quality

MCP Inspector - Interactive UI to connect to any MCP server and test resources, prompts, and tools with OAuth and transport diagnostics

Langfuse Experiment Runner SDK - Programmatically run and compare prompt versions with schema-enforced structured outputs and scoring of workflows