RAG Architecture: GenAI Pokémon Go Strategy Assistant

Overview

This projects is an end-to-end architecture designed to enhance Pokémon Go UX. Here demonstrates a sequential process involving multiple Large Language Model (LLM) gpt-4 agents, leveraging a Retrieval-Augmented Generation (RAG) system to deliver personalized responses.

Features

• End-to-End Architecture: Integration of multiple LLM agents using a sequence-by-sequence process.
• Personalized Responses: Utilization of a Retrieval-Augmented Generation (RAG) system to tailor interactions.
• Pokémon-Focused Application: Designed as a complementary strategy assitant for Pokémon users.

Tech Stack

• Python 3.10: The backbone of our project, ensuring a modern and efficient development environment.
• FastAPI: A modern, fast (high-performance), web framework for building APIs with Python 3.7+ based on standard Python type hints.
• Faiss Vector Database: Utilization Meta's High-Performance Vector Database to optimize memory usage and speed, with excellent GPU architecture adaptability.
• Streamlit Application: A user-friendly interface to interact with the system by a QA-chatbot.
• Langchain Library: A library to facilitate the communication between the GPT-4 and the LLM Agents.
• OpenAI API: Integration with the OpenAI API to leverage the GPT-4 model.
• Pokepy: A Python wrapper for the Pokémon API, providing a simple way to access Pokémon data through Pydantic model input schemas.

Architecture

For an in-depth understanding of the architecture and to access the full article, please visit my website.

Getting Started

To delve into the architecture and explore the application, please ensure you have Python 3.10 installed, along with the Langchain library and Faiss by Meta.

Important notes:

System - The entire setup process is designed for Unix-based systems, so you can use the next Makefile commands to run locally your system, otherwise, you can replicate the commands of the Makefile in your terminal.

Environment Variables - each one of the next commands will source the variables defined in you .env file, so double check that your OpenAI API Key is in place. If you want to make permanent changes, you would need to add these variables to your shell initialization script (like .bashrc, .bash_profile, or .zshrc for Unix/Linux/MacOS, or Environment Variables on Windows).

Create a Virtual Environment

Create a virtual environment using the following command:

make create_venv

Activate the Virtual Environment

Activate the virtual environment using the following command:

source .venv/bin/activate

Install the Dependencies

Install the dependencies using the following command:

make install_dev

Install the Test Dependencies

Install the test dependencies using the following command:

make install_test

Run API Server

Run the API server using the following command:

make api_server

Display the Swagger UI documentation to see the available endpoints at: http://localhost:8000/docs

Display UI

Run the Streamlit app using the following command:

make display_ui

Usage examples

The following are examples of the different types of interactions that the system can handle, which are categorized by the intent type and the structure type.

Intent Types

defense_suggestion: An unknown Pokémon has appeared and the user needs a suggestion to use a Pokémon against it. The user must provide the name of the Pokémon as part of the input.

information_request: The user wants to know more about a Pokémon. In the input, the user must provide one of the following input text structures: (1) the name of the Pokémon with a request for information, (2) a description of a Pokémon in natural language without explicitly mentioning the name of the Pokémon with the intention of guessing what the Pokémon is, OR (3) a question about the Pokémon with the Pokémon name explicitly mentioned, asking for a specific attribute such as: Evolution, Diet, Habitat, Base Stats, etc.

squad_build: The user wants to build a squad of Pokémon based on the opponent's Pokémon types. The user must provide a list of Pokémon names as part of the input.

None: Return GPT-4 own response. The user request does not fall into any of the above categories.

Prompt examples:

Intent Type	Structure Type	Example
defense_suggestion	any	I stumbled upon a wild Grovyle lounging in the park! Which Pokemon should I choose for an epic battle to defeat it?
information_request	pokemon_names	Alright, Pokédex! It's time to find out everything about Snorlax and Pikachu!
information_request	natural_language_question	Do you know in what kind of habitats I can find a Psyduck?
information_request	natural_language_description	Can you guess which Pokémon is a dual-type Grass/Poison Pokémon known for the plant bulb on its back, which grows into a large plant as it evolves
squad_build	any	Time to challenge the Fire Gym Leader! He's got a tough team with a Ninetales and Combusken, but I need your help to build a squad
None	any	I love Pretzels

Response examples:

Prompt: Do you know in what kind of habitats I can find a Lucario?

Answer with JSON mode enabled:

{
  "header": "Lucario can be found in Cave, Forest, and Mountain habitats.\n",
  "body": [
    " ### Lucario\n\n**ID**: 448\n\n**Stats**: - HP: 70\n - Speed: 90\n - Attack: 110\n - Special-Attack: 115\n - Defense: 70\n - Special-Defense: 70\n\n**Height**: 1.2m  **Weight**: 54.0kg\n\n**Types**: fighting, steel\n\n**Abilities**: steadfast, inner-focus, justified\n\n**Damage Relations**:\n - no damage to: ghost\n- half damage to: flying\n- double damage to: normal\n- half damage from: rock\n- double damage from: flying\n"
  ],
  "sprites": {
    "Lucario": [
      "https://raw.githubusercontent.com/PokeAPI/sprites/master/sprites/pokemon/back/448.png",
      "https://raw.githubusercontent.com/PokeAPI/sprites/master/sprites/pokemon/back/shiny/448.png",
      "https://raw.githubusercontent.com/PokeAPI/sprites/master/sprites/pokemon/448.png",
      "https://raw.githubusercontent.com/PokeAPI/sprites/master/sprites/pokemon/shiny/448.png"
    ]
  },
  "intent_type": "information_request",
  "intent_structure": "natural_language_question"
}