Langchain

Framework modules chain prompts, parsers, retrievers, and tool executors into composable LLM applications. Swap models or vector stores without rewriting orchestration glue. Tracing hooks expose per-step latency for tuning agent loops.