🤖 LangChain & LangGraph — Agents, RAG & Multi-Agent Systems

<p align="center"> <img src="https://img.shields.io/badge/Python-3.11-blue?style=for-the-badge&logo=python&logoColor=white"/> <img src="https://img.shields.io/badge/LangChain-0.3-1C3C3C?style=for-the-badge&logo=langchain&logoColor=white"/> <img src="https://img.shields.io/badge/LangGraph-Agents-6C3483?style=for-the-badge"/> <img src="https://img.shields.io/badge/CrewAI-Multi--Agent-E74C3C?style=for-the-badge"/> <img src="https://img.shields.io/badge/Google-Gemini-4285F4?style=for-the-badge&logo=google&logoColor=white"/> <img src="https://img.shields.io/badge/Tavily-Search-FF6B35?style=for-the-badge"/> </p> <p align="center"> A structured collection of notebooks covering the full LangChain and LangGraph ecosystem — from prompt templates and chains to stateful multi-agent systems, RAG pipelines, and CrewAI orchestration. </p>

---

📂 Notebooks

| Notebook | Topics Covered | Key Concepts | |---|---|---| | Langchain.ipynb | LangChain foundations | Prompt templates, chains, LCEL, structured output, Pydantic, parsers | | LLM_Document_Loaders_And_RAG.ipynb | Document loaders + splitters | TXT, PDF, CSV, Web loaders; Character, Recursive, Code splitters; Chroma | | Rag_Implementation.ipynb | RAG from scratch | PDF loader, text splitting, Chroma vector store, RetrievalQA | | Langchain_Agents.ipynb | LangChain agents | Tool use, Tavily search, custom tools with @tool, ReAct agents | | Langgraph_Agents.ipynb | LangGraph agents | MessageGraph, StateGraph, planner-search-responder pipeline | | Chatbot_Graph.ipynb | Stateful chatbot | MemorySaver, multi-turn conversation, thread-based memory | | Post_Creation_Agent.ipynb | Multi-node graph | Writer → Reviewer agent loop, LinkedIn post generation | | Crew.ipynb | CrewAI multi-agent | Crew, Task, Agent orchestration, role-based AI agents | | Career_Agent.ipynb | Career AI agent | Resume + job analysis agent with Tavily real-time search |

---

🧠 Concepts Covered

1. LangChain Core (Langchain.ipynb)

• Prompt Templates: ChatPromptTemplate, PromptTemplate, multi-message templates
• LCEL (LangChain Expression Language): prompt | model | parser chains
• Structured Output: with_structured_output() with TypedDict, Pydantic, and JSON schema
• Output Parsers: StrOutputParser, JsonOutputParser, custom parsers
• Multi-model: Gemini, Mistral (via OpenRouter), OpenAI-compatible APIs

2. RAG Pipeline (Rag_Implementation.ipynb)

PDF → PyPDFLoader → RecursiveCharacterTextSplitter (chunk_size=1000, overlap=50)
    → HuggingFace Embeddings → Chroma VectorStore → RetrievalQA (k=8)

3. Agents & Tools (Langchain_Agents.ipynb)

• initialize_agent with zero-shot-react-description
• TavilySearchResults for real-time web search
• Custom tools with @tool decorator
• Tool selection and reasoning trace

4. LangGraph (Langgraph_Agents.ipynb, Chatbot_Graph.ipynb)

StateGraph / MessageGraph
    ├── Planner Node    → decides what to search
    ├── Search Node     → calls Tavily
    └── Responder Node  → generates final answer

• MemorySaver for persistent multi-turn memory
• thread_id for session isolation
• add_messages annotation for message accumulation

5. Multi-Agent Collaboration (Crew.ipynb)

• CrewAI: Agent, Task, Crew orchestration
• Role-based agents with backstory and goal definitions
• Multi-agent task delegation and output chaining

6. Post Creation Agent (Post_Creation_Agent.ipynb)

• Writer Node → writes LinkedIn post
• Reviewer Node → critiques and improves
• Cyclic graph with conditional edge back to writer

---

🚀 Setup

1. Clone

git clone https://github.com/Ashwin14101/langchain-langgraph-agents.git
cd langchain-langgraph-agents

2. Install dependencies

pip install langchain langchain-core langchain-community langchain-google-genai \
            langgraph crewai tavily-python chromadb sentence-transformers \
            pypdf python-dotenv

3. Set API keys

Create a .env file:

GOOGLE_API_KEY=your_google_api_key_here
TAVILY_API_KEY=your_tavily_api_key_here

Get your keys:

• Google AI Studio → https://aistudio.google.com/app/apikey (free)
• Tavily → https://tavily.com (free tier available)

4. In each notebook, replace the API key cell with:

from dotenv import load_dotenv
import os
load_dotenv()
API_KEY = os.getenv("GOOGLE_API_KEY")

5. Run notebooks

jupyter notebook

---

🛠️ Tech Stack

| Tool | Purpose | |---|---| | LangChain 0.3 | Prompt templates, chains, LCEL, structured output | | LangGraph | Stateful agent graphs, multi-node pipelines | | CrewAI | Role-based multi-agent orchestration | | Google Gemini | Primary LLM (via langchain-google-genai) | | Tavily | Real-time web search tool for agents | | Chroma | Local vector store for RAG | | HuggingFace | Embedding models |

---

📈 Learning Progression

Langchain.ipynb          ← Start here: prompts + chains
      │
Rag_Implementation.ipynb ← Add knowledge: PDF + vector search
      │
Langchain_Agents.ipynb   ← Add tools: agents + web search
      │
Langgraph_Agents.ipynb   ← Add control flow: graph-based agents
      │
Chatbot_Graph.ipynb      ← Add memory: stateful conversations
      │
Post_Creation_Agent.ipynb ← Add loops: writer-reviewer cycle
      │
Crew.ipynb               ← Add collaboration: multi-agent crews
      │
Career_Agent.ipynb       ← Final: full career AI agent

---

🤝 Contributing

Found a bug or want to add more notebooks? PRs welcome!

---

📄 License

MIT © Ashwin14101