Getting-started-with-langchain

Langchain provides components and abstraction that is useful to for quick prototypes/POCs, while also allowing vendor agnostic way to work with LLMs.

• By being vendor agnostic, users can interact with provider APIs in a consistent way.
• Provided convenient components include:
  • A text splitter function.
  • A PDF loader etc.
  • A lightweight InMemoryVectorStore implementation.

Langchain has many resources which makes it easy to get started.

Specifically, I suggest the following:

• Build a semantic search engine
  • Build a simple semantic search engine.
  • Familiarises you with LangChain’s document loader, embedding, and vector store abstractions.
• Build a simple LLM application with chat models and prompt templates
  • Build a simple LLM app.
• Build a Retrieval Augmented Generation (RAG) App
  • Build a simple RAG implementation.

Methods:

Invoking:

Runnable methods that result in the execution of the Runnable (e.g., invoke, batch, stream, astream_events) work with these input and output types.

• invoke: Accepts an input and returns an output.
• batch: Accepts a list of inputs and returns a list of outputs.
• stream: Accepts an input and returns a generator that yields outputs.

The input type and output type vary by component:

Component	Input Type	Output Type
Prompt	dictionary	PromptValue
ChatModel	a string, list of chat messages or a PromptValue	ChatMessage
LLM	a string, list of chat messages or a PromptValue	String
OutputParser	the output of an LLM or ChatModel	Depends on the parser
Retriever	a string	List of Documents
Tool	a string or dictionary, depending on the tool	Depends on the tool
(Documentation link)