A curated collection of practical AI and LLM examples using Keras,LangChain,OpenAI, Azure OpenAI, and LangGraph. This repository demonstrates how to build intelligent agents, integrate custom tools, work with APIs, and run local models like Mistral using Ollama.
| 🔧 Project Name | 📋 Description |
|---|---|
| 📷 basic-cnn | Basic CNN model to perform image classification |
| 🔁 basic-rnn | Simple RNN model to predict next word |
| 🔤 tiny-gpt-model | Tiny GPT model for learning Transformer architecture |
| 🤖 local-llm | Run local LLM using Ollama |
| 🧮 basic-agent | Basic calculation agent using LangChain and tool |
| 🛰️ basic-agent-tracing-langsmith | Calculation agent with LangSmith tracing enabled |
| 📡 basic-rag | Basic RAG system to query weather data over a date range |
| 🔁 basic-agent-langgraph | Agent using LangGraph, structured tool, and logging |
| 🌦️ fastapi-mcp-api | Weather API built using FastAPI as MCP server |
| 🔌 test-mcp-client | Basic MCP client for testing interactions |
| 🔄 basic-agent-2-agent | Sample using Agent-to-Agent (A2A) communication protocol |
| 🗺️ basic-multi-agent-system | Simple multi-agent architecture using LangGraph |
This repository began as a personal learning journey — a place to experiment, break, fix, and deeply understand the core building blocks of AI and modern software systems.
Instead of just reading blogs or watching tutorials, I wanted to learn by doing — by writing real code, training models from scratch, connecting agents, and seeing firsthand how ideas like RNNs, Transformers, LangChain, and local LLMs actually work in practice.
Each project here represents a small “aha!” moment — and a belief that the best way to grow is to keep building, keep iterating, and stay curious.
If any of these examples help you learn something new (or spark your own experiments), this repo has done its job. 🚀
Some images or sample data used in this project were found on the internet and are assumed to be in the public domain or used under fair use for non-commercial purposes.
Portions of the code, architecture ideas, and debugging help were assisted by ChatGPT, used as a learning partner and coding assistant throughout the project.
💡 Like it? Don't forget to star ⭐️ it! 😊