Lang

LangChain Academy - courses

Overviews

• LangChain

• LangGraph

• LangSmith LangChain is built upon LangGraph. Debug with LangSmith.

Integrations - by component

• Models - chat

• Models - embedding

• Tools & toolkits

• Text splitters

• Vector stores

• Doc loaders

• Key-value stores

semantic search tutorial

tool decorator

• does: configure tool to attach raw docs as artifacts to each ToolMessage.

• purpose: lets us access document metadata

---

LangChain

LangChain | Core components

<image>

1. Agents: a framework to orchestrate tool-use via LLM's reasoning. They run tools in a loop to achieve a goal.

2. Models: reasoning engines that take input ⟶ gen output

3. Messages and tools

   1. Messages (memory): units of context that carry context & conversation state

   2. Tools: callable functions/APIs that retrieve external data

4. Short-term memory, Streaming, Structured output

   1. Short-term memory: in-session context stores that retain recent interaction history

   2. Streaming: for displaying output progressively (even before a complete response is ready)

   3. Structured output: format for model output (eg JSON) for the LLM

LangChain | Quickstart | Build a real-world agent

Quickstart Create these to build an agent:

1. write prompts

2. tools: integrate w external data

   1. depends on runtime context 

   2. interacts with agent memory

3. model config: for consistent responses

4. response format for predictable results

5. memory for context across chats

6. invoke agent