Welcome to GenZ's GenAI Repository : a complete collection of my hands-on learning, implementations, and experiments in modern Generative AI engineering.

This repository combines practical implementations of:

• LangChain
• LangGraph
• LangSmith
• Model Context Protocol

The goal of this repository is to deeply understand how modern AI applications are designed, orchestrated, monitored, and deployed using production-ready frameworks and workflows.

I explored everything from prompt engineering and RAG systems to agentic workflows, MCP servers, memory handling, observability, tool calling, persistence, and human-in-the-loop systems.

---

LangChain

Topics Covered

Introduction to LangChain

• What is LangChain
• Benefits and ecosystem
• Model agnostic development
• Memory and state handling
• AI application use cases

---

Models

• LLMs vs Chat Models
• OpenAI integration
• Anthropic integration
• Google models
• HuggingFace models
• Open source models
• Embedding models

---

Prompts

• PromptTemplate
• Dynamic prompts
• ChatPromptTemplate
• Role based prompting
• Few shot prompting
• MessagesPlaceholder

---

Structured Output

• JSON output
• TypedDict
• Pydantic models
• Structured response generation
• with_structured_output()

---

Output Parsers

• StrOutputParser
• JSONOutputParser
• StructuredOutputParser
• PydanticOutputParser

---

Chains

• Simple Chains
• Sequential Chains
• Parallel Chains
• Conditional Chains

---

Runnables

• RunnableSequence
• RunnableParallel
• RunnablePassthrough
• RunnableLambda
• RunnableBranch

---

LCEL

• LangChain Expression Language
• LCEL pipelines
• Runnable composition
• Custom workflows

---

Document Loaders

• TextLoader
• PyPDFLoader
• DirectoryLoader
• WebBaseLoader
• CSVLoader

---

Text Splitters

• Length based splitting
• Structure based splitting
• Document based splitting
• Semantic chunking

---

Vector Stores

• FAISS
• Chroma
• Vector stores vs vector databases
• Similarity search
• Embedding storage

---

Retrievers

• Vector Store Retriever
• Wikipedia Retriever
• MMR Retriever
• Multi Query Retriever
• Contextual Compression Retriever

---

RAG

• RAG architecture
• Indexing
• Retrieval
• Augmentation
• Generation
• Hybrid retrieval
• Reranking
• Evaluation with Ragas
• LangSmith integration

---

Tools and Tool Calling

• Built in tools
• Custom tools
• Structured tools
• BaseTool
• Tool binding
• Tool execution

---

Agents

• AI Agents
• ReAct pattern
• Agent Executor
• Custom agent creation
• Multi step reasoning
• Tool using agents

---

What I Learned

• Building modular LLM applications
• Designing scalable RAG systems
• Creating tool calling workflows
• Managing prompts, memory, and retrieval pipelines
• Working with vector databases and embeddings
• Structuring AI outputs for production systems

LangGraph

Topics Covered

Introduction to LangGraph

• What is LangGraph
• Why LangGraph
• Stateful AI workflows
• Graph based orchestration

---

Core Concepts

• Nodes
• Edges
• State management
• Graph execution flow
• START and END nodes

---

Workflow Patterns

• Sequential workflows
• Parallel workflows
• Conditional workflows
• Iterative workflows
• Dynamic routing
• Multi agent workflows

---

State Management

• Shared state
• Typed state
• State updates
• State reducers
• Custom state handling

---

Memory

• Short term memory
• Conversation memory
• Context persistence
• Stateful chat workflows

---

Persistence

• Checkpointing
• Persistent execution
• Resume interrupted workflows
• Durable state management

---

Streaming

• Token streaming
• Real time graph updates
• Streaming responses

---

Tools Integration

• Built in tools
• Custom tools
• Tool calling workflows
• External API integration
• Function calling agents

---

MCP Integration

• MCP client in LangGraph
• Tool orchestration using MCP
• AI assistant workflows with MCP
• External system communication

---

Human in the Loop

• HITL workflows
• Manual approval systems
• Human feedback integration
• Interrupt and resume execution

---

Observability

• Graph debugging
• Execution tracing
• Monitoring workflows
• State inspection
• LangSmith observability

---

RAG in LangGraph

• Retrieval workflows
• RAG pipelines
• Vector store integration
• Query routing
• Context augmentation
• Retrieval optimization

---

Agents

• ReAct agents
• Tool using agents
• Autonomous workflows
• Multi step reasoning
• Agent orchestration

---

What I Learned

• Building stateful AI systems
• Designing complex agent workflows
• Managing graph based execution
• Implementing production ready AI orchestration
• Creating persistent and observable workflows
• Integrating tools and MCP clients
• Handling human approvals in AI systems

LangSmith

Topics Covered

What is LangSmith

A developer platform by LangChain for:

• Debugging
• Testing
• Monitoring LLM applications

Focus

• Observability
• Evaluation

---

Why LangSmith

LLM applications are non deterministic.

Problems

• Hard to debug prompts
• Difficult to track failures
• No proper execution visibility

LangSmith Solves

• Visibility into LLM pipelines
• Performance tracking
• Debugging complex chains and agents

---

Topics Covered

Core Concepts

• Projects
• Runs
• Traces

---

Observability

• Inputs and outputs tracking
• Token usage
• Latency tracking
• Failure debugging

---

Monitoring

• Success rate tracking
• Response time monitoring
• Error analysis

---

Alerting

• Failure alerts
• Latency alerts
• Threshold based notifications

---

Prompt Engineering

• Prompt experimentation
• Prompt comparison
• Prompt versioning

---

Evaluation

• Dataset creation
• Annotation workflows
• Response evaluation

---

User Feedback

• Feedback collection
• Human evaluation workflows

---

Collaboration

• Shared traces
• Team debugging
• Workflow sharing

---

LangSmith Integrations

• LangChain tracing
• LangGraph observability
• Agent workflow debugging
• RAG evaluation

---

What I Learned

• Debugging AI workflows
• Monitoring production LLM apps
• Evaluating AI responses
• Tracking prompt performance

Model Context Protocol

Topics Covered

Introduction to MCP ( FastMCP )

• What is MCP
• Why MCP
• MCP architecture
• MCP Lifecycle
• Client server communication
• AI tool interoperability

---

MCP Core Concepts

• MCP clients
• MCP servers
• Resources
• Tools
• Context sharing

---

MCP Servers

• Local MCP servers
• Remote MCP servers
• Server configuration
• Multi server setup
• Server communication workflows

---

MCP Client Implementations

• MCP client creation
• Tool discovery
• Resource access
• Prompt handling
• Multi server connections

---

Resources

• Static resources
• Dynamic resources
• File based resources
• Resource retrieval workflows

---

MCP with LangGraph

• MCP integration workflows
• Tool orchestration
• Stateful AI workflows

---

MCP with LangChain

• Tool calling integration
• Agent workflows
• Context aware orchestration

---

• converting FastAPI Application to FastMCP

What I Learned

• Building MCP based AI systems
• Creating local and remote MCP servers
• Integrating MCP with LangGraph and LangChain
• Managing scalable tool orchestration

---

Regards,
Prathamesh Bhavsar