Agentic Blueprints 9

Splits a hard question into sub-questions, researches them in parallel to keep the wall-clock low, then synthesizes one brief with citations.

LATS (Language Agent Tree Search)

io.github.langchain-ai/lats

LangGraph

Monte-Carlo tree search over agent actions: generate candidates, reflect and score them, expand the best branches until one solves the task.

A supervisor decides who works next — a researcher who searches or a coder who edits the repo — and assembles the result when the work is done.

Retrieval that doesn't trust its first search: it grades the results, rewrites the query when they're weak, and only answers once the context is good.

Multi-Agent Collaboration

io.github.langchain-ai/multi-agent-collaboration

LangGraph

A researcher agent and a chart-generator agent divide a task and route work between themselves until the answer is a finished chart.

Reflexion

io.github.langchain-ai/reflexion

LangGraph

An answerer that critiques itself with citations: draft, self-review for missing and superfluous content, search to fill gaps, revise.

Plan-and-Execute (LangGraph tutorial)

io.github.langchain-ai/plan-and-execute-tutorial

LangGraph

The original plan-and-execute notebook: a planner writes the step list, an executor works through it, and a replanner revises as results come in.

Adaptive RAG

io.github.langchain-ai/adaptive-rag

LangGraph

Routes each question to the right retrieval strategy, grades what comes back, and falls back to web search when the index can't answer.

Corrective RAG

io.github.langchain-ai/corrective-rag

LangGraph

Grades retrieved documents for relevance and, when they fail the grade, falls back to web search before answering.