This repository collection contains a set of practical Generative AI demos that I have built primarily for live, in-person presentations. The focus has been on creating clear, hands-on examples that demonstrate core Gen AI concepts in an accessible and interactive way.
Each demo is designed to showcase specific architectural patterns, tools, or techniques within the modern LLM ecosystem. They are intentionally lightweight, easy to run, and structured to support live walkthroughs, technical deep-dives, and audience Q&A sessions.
While originally developed for conference talks, workshops, and internal presentations, these demos can also serve as reference implementations or starting points for experimentation and further development.
| Repository | Description | Key Features |
|---|---|---|
| Awesome-RAG Agent Plugin | An Agent Plugin built from Awesome-RAG. It turns a human-facing RAG resource map into agent-usable skills for RAG architecture, evaluation, and production review directly inside the development workflow. | - Agent-readable layer on top of Awesome-RAG - Skills for RAG architecture advice, evaluation planning, and production review - Dedicated RAG Architect agent - Designed for use in editor-based agent workflows - Demonstrates the idea: the list is for humans, the plugin is for agents |
| Flowstate | A lightweight Python library for maintaining structured behavioral context for AI agents. Instead of relying on vector stores or document retrieval, it models user interaction as evolving state (facts, actions, intent) that the agent uses as background context. | - Structured context model (scope, facts, actions, intent) - Stateful interaction without RAG - Incremental context updates via setters and merges - Prompt-ready rendering for LLMs - Optional schema validation with Pydantic |
| LangGraph Agent Demo 2025 | A simple web search AI agent that decides when to do a web lookup based on user query. | - Simple LangGraph app with Django frontend - Web search classification using Pydantic models |
| LangChain LLM Classifications Demo | Demonstrates structured classifications using LLMs with customizable schemas. Utilizes few-shot prompting for accurate responses. | - Customizable classification schema - Few-shot prompting for guided outputs - Incident report processing with structured output |
| LangChain Chroma RAG Demo 2024 | Demonstrates a Retrieval-Augmented Generation (RAG) system using LangChain and ChromaDB. Supports custom data ingestion and real-time query streaming. | - Retrieval-Augmented Generation (RAG) - In-memory database with persistence - Real-time query streaming for user interactions |
| Gorilla Open Functions V1 Gist 2023 | Illustrates how to run Gorilla LLM's Open Functions, an early function calling model, locally with Hugging Face Transformers. | - Local inference with Hugging Face Transformers - Function calling |
| RetrieveGPT (Classic RAG Demo) | Early demonstration of a Retrieval-Augmented Generation (RAG) pipeline. Shows the core architecture of document ingestion, embedding generation, vector retrieval, and LLM answer generation. | - Classic RAG architecture - Document chunking and embeddings - Vector similarity search - Context grounding for LLM responses - Architecture diagram |