Learning the Ropes for Your Next LangChain Project
Large language models entered the mainstream last year as tools for fun—and occasionally downright silly—experimentation. Who among us hasn’t challenged ChatGPT to invent a new knock-knock joke or compose a Shakespearean sonnet about puppies?
As the power of LLMs became increasingly apparent, so did their limitations. Machine learning practitioners and app builders quickly realized that there’s only so much you can do with models that can’t obtain up-to-date information and remain hermetically shut off from each other.
Luckily, we already have multiple tools and platforms that aim to overcome these constraints by connecting LLMs to external data sources and to other models, thereby opening them up to new and creative use cases. LangChain has emerged as one of the leading options in this space, and in the months following its launch, TDS authors have been exploring its abilities and pain points. You’ll find some of the best work on this topic in the lineup of practical, hands-on resources we’ve gathered this week—are you ready to roll up your sleeves?
- Part overview, part glossary, and part hands-on introduction, Leonie Monigatti’s beginner-friendly primer is the go-to article for anyone who’s just taking their first steps with LangChain. It includes clear explanations of all the main terms and concepts you’ll need to know.
- If you’re ready to sink your teeth even deeper, Dominik Polzer recently put together a comprehensive guide that provides helpful context around how LLMs work and why a framework like LangChain was needed to begin with, and goes on to detail the process of building an app, step by step.
- Developers have come up with numerous ways to combine models, APIs, and ML platforms with LangChain. Wen Yang’s tutorial covers OPL, a popular technical stack that brings together OpenAI models (think GPT-4), Pinecone (a vector-database tool), and LangChain, and shows how to leverage it to create a chatbot with specific-domain knowledge.
- What if you’ve already built a model-specific proof of concept and are now feeling stuck when you try to turn it into a bonafide app? Lily Hughes-Robinson is here to help with a concise explainer on the process of refactoring your project to make it LangChain-ready.
- If you’re not quite ready to stop tinkering, how about another detailed project walkthrough? Shuai Guo invites us to follow along the journey of building a language-learning app that relies on two chatbots communicating with each other.
Some of you might need a breather from all things LLM-related; we get it—and we’re here to help! Here are some excellent articles on other topics worth exploring:
- We kicked off a new month with a selection of thought-provoking articles at the intersection of data science, machine learning, and climate change.
- AI-generated audio typically requires massive computing resources. Christopher Landschoot’s debut TDS post shows how to train models and generate sounds with audio waveform diffusion—on your laptop.
- Multiple group analysis can be a powerful tool in healthcare, HR, marketing, and other fields. Laura Castro-Schilo’s latest contribution is a great place to start learning about ways to implement MGA in real-world scenarios.
- In the mood for a technical (yet accessible) deep dive? Peggy Chang delivers a solid read on integer encoding and compression in the context of open-source search engine Lucene.
- If you’ve been planning to expand your computer vision toolkit, Dhruv Matani and Naresh’s four-part series on image segmentation in PyTorch will motivate you to stop procrastinating and start coding.
- Approaching GPT models from the perspective of a data scientists who’s also a psychologist, Maarten Grootendorst takes a close look at “how these models behave, how we would like them to behave, and how we are nudging these models to behave like us.”
Thank you for supporting our authors! If you enjoy the articles you read on TDS, consider becoming a Medium member — it unlocks our entire archive (and every other post on Medium, too).
Until the next Variable,
TDS Editors
