Build a Lovable Clone: Python AI Coding Agent Workshop with PydanticAI and OpenAI SDK

Описание: AI Bootcamp is my live, hands-on course that focuses on building production-ready AI agents step by step. We focus on building AI systems that are measurable, testable, and observable. Join the next iteration here: https://maven.com/alexey-grigorev/fro...

In this hands-on workshop, we build an autonomous coding agent from scratch that mimics the functionality of tools like "Lovable." We start with a basic Django template and build an agent capable of planning, coding, and modifying the app using Python.

We explore multiple frameworks—starting with basic tool use, moving to the OpenAI Agents SDK, and finally implementing a production-grade workflow using PydanticAI with Anthropic’s Claude 3.5 Sonnet.

Links:

Workshop repo: https://github.com/alexeygrigorev/wor...
Workshop about agents: https://maven.com/p/3b1afc/hands-on-w...
Course: https://maven.com/alexey-grigorev/fro...

0:00 Introduction & Workshop Goals
0:09 Presenter Background & Experience
1:08 Example: What is "Lovable"? (The Goal)
2:18 The Plan: Building a Python/Django Coding Agent
3:43 Tools Setup: OpenAI API Keys & Budget
5:02 Environment: Using GitHub Codespaces
11:20 Concept: Agents vs. Chatbots (Adding Tools)
16:48 Defining System Prompts & User Prompts
24:00 Setting up the Django Project Template
29:01 Defining Agent Tools (Read/Write Files, Search)
33:36 Implementing & Testing Tools in Jupyter
37:27 Crafting the "Developer Prompt" (Crucial Step)
42:16 Running the Agent with ToyAIKit
47:49 Recap & Break
50:42 Intro to OpenAI Agents SDK
53:48 Building a Joke Agent with Agents SDK
58:08 Porting the Coding Agent to Agents SDK
1:04:07 Intro to PydanticAI (Production Grade Framework)
1:06:50 Creating the Agent with PydanticAI
1:10:56 Switching to Anthropic Claude 3.5 Sonnet (Better Code)
1:15:34 Test Case: Building an Anki Cards App
1:16:16 Testing Z.AI Models (Reasoning Models)
1:25:45 Course Overview & Learning Path

In this video, we cover:

Agent Theory:The difference between Chatbots and Agents (Tools & Reasoning).
The Tech Stack: Python, Django, OpenAI, and GitHub Codespaces.
Tool Creation: Giving the AI access to the file system (Read, Write, Grep, Bash).
Prompt Engineering: crafting the "Developer Prompt" that forces the agent to plan before coding.
Frameworks: Comparing OpenAI Agents SDK vs. PydanticAI.
Model Swapping: Why we switched from GPT-4o Mini to Claude 3.5 Sonnet for better coding results.

Tech Stack Used:

Language: Python
Frameworks: Django, PydanticAI, OpenAI Agents SDK
Models: GPT-4o Mini, Claude 3.5 Sonnet, Z.AI (GLM-4.5)
Environment: GitHub Codespaces, Jupyter Notebook

Who is this for?
Python developers and Data Scientists who want to understand how coding tools like Lovable or Cursor are actually built under the hood, and how to implement their own agentic workflows.

#AIAgents #Python #Django #PydanticAI #OpenAI #Claude3 #LLM #CodingAgent