LangChain with Python Bootcamp

Learn to create and manage basic model input prompts and outputs in LangChain, connect to OpenAI or other LMS, and use prompt templates, few-shot prompts, and output parsing and serialization.

Create an OpenAI account, obtain an API key, and securely store it as an environment variable; learn billing basics and the free $18 credit for first three months.

Learn to work with large language models using LangChain, comparing text completion LM and chat models, and integrate them via OpenAI and other services with prompts and a system prompt.

Explore how LangChain enables chat models by structuring system, human, and AI messages, adjust parameters like temperature, presence penalty, and max tokens, and use in-memory caching to save replies.

Explore prompt templates in LangChain with Python to configure inputs for text completion and chat models, using input variables, from template, and format to create system, human, and AI prompts.

Apply long chain syntax with chat models and prompt templates to create a travel idea function that outputs itineraries from an interest and budget.

Build a travel itinerary function in LangChain with Python Bootcamp by using system and human templates, compile into chat prompt, and run a chat OpenAI model with interest and budget.

Learn to build few-shot prompt templates in LangChain with Python, using histories of input-output pairs and system/human/AI messages to scaffold simple, readable responses for complex texts.

Learn how to convert language model outputs into precise Python data types using LangChain parsers, format instructions, and parse methods, including date time and comma separated list examples.

Fix misformatted outputs in LangChain by using a strong system prompt with format instructions, or apply the output fixing parser to reprocess results from the date time output parser.

Leverage the Pedantic output parser and Pydantic for type validation to convert language model replies into a Python object, such as a scientist with name and discoveries.

Master serialization by saving prompt templates to json files and loading them later with load prompt. Share and reuse prompts with input variables across projects.

Explore the langchain with python bootcamp's models inputs and outputs project by building a history quiz bot. Implement methods to generate a date question and format the date time answer.

Explore a LangChain with Python bootcamp project solution: using system and human prompts to generate quiz questions, with format instructions and a date time parser to validate answers.

Connect your model to data sources by loading pdf, csv, and html documents, transforming them into vector embeddings, storing and querying them with your own models using different retrievers.

Learn how LangChain's built-in document loaders ingest csv, html, and pdf files into internal document objects for embeddings or model interactions, and how to handle required libraries.

Explore document loader integrations that connect with third-party services like Google Cloud, Wikipedia, and AWS S3, and summarize Hacker News posts with a LangChain chat model.

Build a function that loads a Wikipedia article via the Wikipedia loader, then queries a chat model with the article context to answer a question about a historical figure.

Use a Wikipedia loader to fetch a person's article, cap the docs loaded, and answer questions via a formatted OpenAI chat prompt with the article text as context.

Explore document transformers that split long text into chunks for embeddings, using character and token splitters. See how 8000 token limits guide chunking for vector stores like Chroma DB.

Learn how to generate text embeddings with Lang Chain using OpenAI embeddings, store vectors in a vector store, and perform similarity searches with document loaders and CSV data.

Learn to persist embeddings with a vector store using chroma, load and split documents, embed chunks, and perform similarity searches with LangChain.

Convert a vector store to a retriever via the as retriever method, enabling get relevant documents and other retrieval calls for multi query retrieval and context compression.

Use a multi query retriever with a chat model to generate query variations and retrieve relevant documents from a vector store, illustrated with an MK Ultra example.

Explore context compression with large language models to distill and tailor answers from vector-store documents, using a compression retriever and lm chain extractor for concise, relevant summaries.

Engage in a code-along data connections exercise to build a legal research assistant bot that answers questions with compressed excerpts from the US Constitution.

Build a US Constitution help bot by loading documents, splitting into 500-word chunks, and embedding with OpenAI in a chroma DB. Retrieve relevant amendment content with a contextual compression retriever.

Explore how LangChain chains connect successive large language model calls, using built-in types like sequential, router, and transform chains to work with vector stores, document similarity, and question answering.

Learn to use the LM chain object as a wrapper around a chat or LM call, connect it to a model and a prompt, then run a single input.

Learn to build a simple sequential chain by linking LM chain objects in a Python list to pass input and generate a blog post from an outline.

Learn to use LangChain's sequential chain: connect three LM chains to generate a review summary, weaknesses, and a tailored improvement plan while accessing all internal outputs.

Learn to route inputs to multiple LM chains using the LangChain LLMRouterChain, by creating beginner and advanced templates and a multi-prompt router to handle magnets and Feynman diagrams.

Explore how the transform chain lets you inject custom Python transformations into a chain, convert a Yelp review to lowercase, and then summarize it with a language model.

Explore OpenAI function calling with LangChain to produce strict json outputs via a json schema and implement a scientist object from a dictionary.

Discover LangChain's built-in mathchain to answer math questions by evaluating string expressions in python with eval, building a math tutor bot and improving accuracy across models.

Explore additional and pre-built chains, including QA models for document questions, connect to a vector store like Chroma via link chain, and implement SQL and summarize chains with minimal code.

Practice sequential chains that detect language, translate customer emails to English, and summarize the translated email, returning a dictionary with the language, translated email, and summary.

Demonstrates building a multi-step LangChain workflow to detect the email language, translate a Spanish customer email to English, and generate a short summary via sequential chains and prompt templates.

Learn how LangChain memory options track and store chat history, including full conversations, limited interactions, or token-based storage, plus summarize memory and load past chats.

Explore the chat message history object as a building block and create an instance to add user and AI messages. Use this history with a language model to continue conversations.

Conversation buffer memory attaches to a conversation chain to track historical user and AI messages, enabling saving, loading, and continuing conversations with memory.

Explore conversation buffer window memory to store and recall the most recent K interactions, with K as the window size; load memory variables to fetch that window.

Connect a conversation summary buffer memory to a large language model in LangChain to generate concise summaries of human AI dialogue, with a configurable max token limit.

Discover how LangChain agents fuse reasoning and acting to connect tools like Wikipedia, calculators, and Google search, using an observation, thought, and action cycle to solve tasks.

Learn the basics of creating an agent in LangChain with Python bootcamp: choose an agent type, load tools, connect a large language model, and run a zero-shot React description agent.

Learn to extend LangChain agents with tool integrations and a Python agent that executes Python code, including setup, API keys, and practical examples.

Learn to build custom tools for LangChain agents by decorating functions and crafting clear docstrings to guide tool use and API calls.

Learn to build conversation agents by attaching memory to a conversational agent and providing tools, enabling memory recall and targeted tool use.