AI FrameCoach SF6

FrameCoach SF6 is an AI-powered coaching assistant for Street Fighter 6. It uses semantic search and a local large language model (LLM) to provide accurate, competitive-level answers about moves, frame data, strategies, and character-specific tools.

---

Features

• Searchable database of frame data and character guides
• AI-generated answers via a local LLM (Gemma)
• Example questions you can ask:
• "What are Zangief’s best anti-airs?"
• "How plus is JP’s OD Amnesia on hit?"
• "What is the startup on Cammy’s crouching heavy punch?"

---

Tech Stack

• Sentence Transformers for semantic vector embeddings
• ChromaDB as a vector database
• Flask for the backend API
• Ollama to run local LLMs (e.g., mistral, gemma)
• JSON and TXT files with move data and guides (converted and ingested)

---

AI Prompt Design

The LLM is instructed to behave like a world-class Street Fighter 6 coach, with a focus on:

• Competitive insights
• Accurate use of frame data
• Concise and helpful explanations
• No fluff, no speculation

---

Requirements

Make sure your system has the following installed:

Python 3.10 or higher

pip (Python package installer)

Ollama (for running local LLMs)

---

Setup Instructions

1. Install Ollama
   https://ollama.com/download
   Once installed, pull the model you want to use (e.g., gemma or mistral):

   ollama pull gemma:2b
   or 
   ollama pull mistral
   

   You can test if it's working with:

   ollama run gemma:2b
   

2. Clone the Project & Set Up Python
   git clone https://github.com/NathanaelDousa/AI-FrameCoach-SF6 cd AI-FrameCoach-SF6

   (Optional) Create a virtual environment: python -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate

3. Install Dependencies
   pip install flask flask-cors sentence-transformers chromadb requests

4. Prepare the Database (Ingest Data) First, make sure your move and guide files are stored as plain .txt files (one file per character or topic) inside the data/ folder. Then run: python ingest_texts.py

   This script will:

   • Load all .txt files in the data/ folder
   • Convert them into vector embeddings
   • Store them in a local ChromaDB database

5. Run the Server
   Start your Flask app: python app.py

   You should see something like:

   • Running on http://127.0.0.1:5000 Open your browser and go to: http://localhost:5000

6. Ask Questions
   You can now ask questions like:

   • What are Zangief’s best anti-airs?
   • How plus is Cammy’s crouching light punch?
   • Tell me how to use Manon’s command grab The app will:

   Embed your question Retrieve the top relevant documents using ChromaDB Feed both the context and your question into the local LLM (Gemma or Mistral) Return a clean and accurate response

Notes You can change the model name in app.py:

MODEL = "gemma"  # or "mistral", etc.

Make sure Ollama is running while using the app.

If you add or update files in data/, run ingest_texts.py again.