Databricks-Samples

Hands-on code samples for Databricks, MLflow, Model Serving, Foundation Models, and Agents. Companion repo for my YouTube channel.

---

Repository Structure

├── foundation-models/
│   ├── fm-api-intro.ipynb                    # Foundation Model API intro
│   ├── RAG/                                  # Vector Search + LangChain RAG
│   └── Agents/
│       ├── AgentBricks/                      # Knowledge Assistant & Info Extraction
│       └── BYO/
│           ├── agents-model-serving/         # Agent deployment on Model Serving
│           ├── Custom-MCP-Server-Agent/      # MCP server + LangGraph agent
│           └── ProductionAgentSeries/        # 6-part production agent series
│
├── mlflow/
│   ├── Intro/                                # MLflow intro notebook
│   ├── AgentServing/                         # MLflow Agent Server scaffold
│   └── ResponsesAgentInterface/              # ResponsesAgent wrapper + local serve
│
├── traditional-ml/
│   └── ModelServing/
│       ├── Built-In-MLflow/                  # Sklearn & Transformers serving
│       ├── BYO/                              # Custom PyFunc model serving
│       └── Multi-Model-Serving/              # Multi-model endpoint patterns
│
└── apps/
    └── mcp/                                  # MCP server + client test scripts

---

Foundation Models & Agents

Foundation Model APIs

• foundation-models/fm-api-intro.ipynb — Getting started with Databricks FM APIs

RAG

• foundation-models/RAG/ — Vector Search setup + LangChain RAG pipeline

AgentBricks

• AgentBricks/KnowledgeAssistant/ — Knowledge Assistant API sample
• AgentBricks/InformationExtraction/ — Information Extraction agent

BYO Agents

• agents-model-serving/ — Deploying agents on Databricks Model Serving
• Custom-MCP-Server-Agent/ — Custom MCP server with LangGraph agent

Production Agent Series

A 6-part series building a production-grade agent end-to-end: MCP tools, ResponsesAgent interface, Agent Server, Lakebase memory, MLflow evaluation, and full-stack deployment as a Databricks App.

foundation-models/Agents/BYO/ProductionAgentSeries/

Part	Focus
Part01	Custom MCP server with FastMCP + Databricks Apps
Part02	ResponsesAgent interface — framework-agnostic LangGraph wrapper
Part03	AgentServer with @invoke / @stream and SSE streaming
Part04	Lakebase memory — managed PostgreSQL + pgvector
Part05	MLflow tracing, metadata, and GenAI evaluation
Part06	Full-stack deployment — all parts combined as a Databricks App

---

MLflow & MLOps

• mlflow/Intro/ — MLflow intro notebook
• mlflow/ResponsesAgentInterface/ — LangGraph agent wrapped with MLflow ResponsesAgent, local serve + request examples
• mlflow/AgentServing/ — Minimal Agent Server scaffold

---

Traditional ML — Model Serving

• Built-In-MLflow/dbx-serving-sklearn.ipynb — Sklearn model on Databricks Model Serving
• Built-In-MLflow/transformers-dbx-serving.ipynb — Transformers model on Model Serving
• BYO/custom-model-pyfunc.ipynb — Custom PyFunc model serving
• Multi-Model-Serving/multi-model-serving-intro.ipynb — Multi-model endpoint patterns

---

Apps

• apps/mcp/ — MCP server and client test scripts

---

Video Links

Topic	Video
What is Databricks	Watch
What is Unity Catalog	Watch
MLflow Introduction	Watch
Model Serving Theory	Watch
Model Serving Hands-On (Sklearn)	Watch
Transformers on Model Serving	Watch
Foundation Model API Intro	Watch
RAG on Databricks (Theory)	Watch
RAG on Databricks (Hands-On)	Watch
Agentic Options on Databricks	Watch

---

Credits

• databricks/app-templates — Agent and app templates
• Databricks AI/ML Documentation — Foundation Models, MLflow, Model Serving
• Databricks Lakebase Documentation — Managed PostgreSQL for agent memory

---

Feedback & Contributions

Star the repo, open an issue, or fork and extend for your own projects.