How langchain-ai/langchain-academy works
Purpose and Execution Model
This is a LangChain Academy course repository for learning LangGraph through progressively advanced, runnable examples. Primary entry points are Jupyter notebooks for guided lessons and LangGraph Studio graph modules for interactive local execution. The supported runtime is Python 3.11–3.13.
Paths: module-0/basics.ipynb, module-1/agent.ipynb, module-6/assistant.ipynb
Curriculum Structure
Modules are organized as a capability progression rather than one integrated application:
• Module 0 establishes environment and basic concepts.
• Modules 1–5 build graph agents, state, human intervention, parallelism, and memory.
• Module 6 focuses on deployment and production interaction patterns.
Notebook variants demonstrate each concept independently; Studio implementations are the executable graph equivalents where present.
Paths: module-1/simple-graph.ipynb, module-3/time-travel.ipynb, module-4/research-assistant.ipynb
Graph Agent Foundations
Module 1 defines the baseline LangGraph patterns: linear chains, simple state graphs, tool/agent behavior, routing, and agent memory. Its Studio graphs are likely the first reusable executable targets for langgraph dev.
Paths: module-1/studio/simple.py, module-1/studio/agent.py, module-1/studio/router.py
State, Persistence, and Message Management
Module 2 centers on graph state schemas and reducers, including message trimming/filtering and persistent conversational state. The checked-in SQLite database supports the state/persistence examples, establishing a local storage boundary rather than requiring a remote database.
Paths: module-2/studio/chatbot.py, module-2/state_db/example.db, module-2/state-reducers.ipynb
Human-in-the-Loop Control
Module 3 demonstrates execution interruption and recovery: static/dynamic breakpoints, streamed interruption, human edits to graph state, and time travel. These examples define a control boundary where graph execution can pause, expose state for review, then resume from altered or historical checkpoints.
Paths: module-3/studio/agent.py, module-3/studio/dynamic_breakpoints.py, module-3/edit-state-human-feedback.ipynb
Parallel and Hierarchical Workflows
Module 4 covers fan-out/fan-in and decomposition patterns: parallelization, map-reduce, research workflows, and subgraphs. The research assistant is the principal external-data workflow and uses Tavily for web search; results are coordinated through LangGraph rather than a standalone service layer.
Paths: module-4/studio/parallelization.py, module-4/studio/map_reduce.py, module-4/studio/research_assistant.py
Long-Term Memory and Deployment
Module 5 expands from session state into durable memory models: agent memory, generic stores, collection schemas, and profile schemas, with shared configuration guiding those graph variants. Module 6 provides deployment-oriented graph code plus Docker Compose example configuration and covers connection/creation, assistant behavior, and double-texting.
Paths: module-5/studio/memory_agent.py, module-5/studio/memory_store.py, module-6/deployment/task_maistro.py
Runtime, Infrastructure, and External Services
Each Studio directory is a self-contained local LangGraph deployment unit: langgraph.json identifies graph exports, requirements.txt defines module dependencies, and .env supplies secrets. langgraph dev exposes a local API and connects it to LangGraph Studio. OpenAI supplies LLM inference; LangSmith provides tracing and Studio hosting/UI integration; Tavily is used for search-oriented lessons.
Paths: module-1/studio/langgraph.json, module-4/studio/.env.example, module-6/deployment/docker-compose-example.yml
Main entry points: agent.ipynb Module 1 agents · state-reducers.ipynb State reducers · edit-state-human-feedback.ipynb Human feedback · research-assistant.ipynb Research assistant
Components 20
Curriculum Modules
Independent notebook lessons progress from basics to deployment.
- Module 0 basics · Environment and LangGraph foundations.
module-0/basics.ipynb - Module 1 agents · Chains, state graphs, tools, routing, and memory.
module-1/agent.ipynb - State reducers · State schemas, reducers, and message management.
module-2/state-reducers.ipynb - Human feedback · State editing and interrupted execution recovery.
module-3/edit-state-human-feedback.ipynb - Research assistant · Parallel research and coordinated external search.
module-4/research-assistant.ipynb - Durable memory · Agent memory, stores, collections, and profiles.
- Module 6 assistant · Deployment and production interaction patterns.
module-6/assistant.ipynb
Studio Graph Units
Runnable LangGraph graph implementations.
- Foundation graphs · Simple, agent, and router executable graph patterns.
module-1/studio/agent.py - Persistent chatbot · Conversation state and reducer-based message handling.
module-2/studio/chatbot.py - Local state DB · Checked-in persistence for state examples.
module-2/state_db/example.db - Interruptible graphs · Dynamic breakpoints pause, expose, and resume graph state.
module-3/studio/dynamic_breakpoints.py - Parallel research graphs · Fan-out/fan-in, map-reduce, and research workflows.
module-4/studio/research_assistant.py - Memory graphs · Durable agent memory and generic store patterns.
module-5/studio/memory_store.py - Deployment graph · Assistant deployment-oriented workflow.
module-6/deployment/task_maistro.py
Local Runtime
Local development and interactive graph execution boundary.
- Studio manifest · Declares graph exports for local development.
module-1/studio/langgraph.json - langgraph dev · Runs a Studio unit locally and exposes its graph API.
- Docker Compose example · Example deployment infrastructure configuration.
module-6/deployment/docker-compose-example.yml
External Services
- OpenAI
- Tavily
- LangSmith / Studio · Tracing plus interactive Studio integration.
Connections 23
- Module 0 basics → Module 1 agents: builds toward
- Module 1 agents → State reducers: extends state
- State reducers → Human feedback: adds control
- Human feedback → Research assistant: scales workflows
- Research assistant → Durable memory: adds durability
- Durable memory → Module 6 assistant: prepares deployment
- Module 1 agents → Foundation graphs: implemented as
- State reducers → Persistent chatbot: implemented as
- Persistent chatbot → Local state DB: persists state
- Human feedback → Interruptible graphs: implemented as
- Research assistant → Parallel research graphs: implemented as
- Durable memory → Memory graphs: implemented as
- Module 6 assistant → Deployment graph: implemented as
- Studio manifest → langgraph dev: declares exports for
- langgraph dev → Foundation graphs: runs
- langgraph dev → Persistent chatbot: runs
- langgraph dev → Interruptible graphs: runs
- langgraph dev → Parallel research graphs: runs
- langgraph dev → Memory graphs: runs
- Docker Compose example → Deployment graph: deploys example
- Foundation graphs → OpenAI: uses LLM
- Parallel research graphs → Tavily: searches web
- langgraph dev → LangSmith / Studio: connects Studio and tracing