learning-langchain — Mini-projects

A compact collection of progressive LangChain mini-projects to learn prompt, chaining, memory, and agents.

Quick setup

• Create and activate a Python virtual environment.
• Install dependencies (each example has its own requirments.txt):

pip install -r requirments.txt

• Add API keys: copy the provided .env.example in an example folder and set OPENAI_API_KEY or ANTHROPIC_API_KEY.

Progressive mini-projects (test knowledge by doing)

• Basic prompt: basic-llm-prompt — simple LangChain prompt template + Chat model. Run python main.py inside the folder.
• Role-based explainer: role-based-explainer — generate explanations in different voices (teacher, pirate, lawyer, kid). Run python main.py.
• Simple 2-step chain: simple-chain — build a 2-step chain (explain simply → give real-world example). Run python main.py.
• Conversation memory: conversation-memory — multiple examples showing in-memory history, RunnableWithMessageHistory, and a checkpointer. Try python main.py, python langchain_main.py, or python checkpointer_main.py.
• Mini RAG: mini-RAG — Document Q&A (load a PDF or text, build a FAISS index, ask questions). Run python main.py inside the folder.
• Basic agent: basic-agent — starter agent examples and small demos. See folder for scripts and run instructions.
• Foundations & notebooks: lca-lc-foundations — longer-form notebooks and exercises for deeper learning. Open notebooks in notebooks/.

Notes

• Examples are intentionally small and runnable; they keep secrets in memory by default. For persistence, replace in-memory stores with a DB-backed store.
• Filenames use requirments.txt (project typo) — install from each example's file when running that example.
• If you hit message-format errors (e.g., TypeError: 'AIMessage' object is not subscriptable), extract replies defensively by checking whether messages are dicts or message objects (see conversation-memory README for a snippet).

Want me to run any example now to verify it starts? Use the folder name and I can run it here.