GLYPHAI - Multimodal Agentic Chat Platform

Glyph is an open-source platform that enables AI-powered multimodal interactions through standalone tools and an intelligent agentic orchestration layer.

Features

• Manim Integration - Generate mathematical animations and visualizations from natural language prompts using the Manim library
• RAG (Retrieval Augmented Generation) - Upload and query documents with semantic search powered by Supabase vector store and HuggingFace embeddings
• Flowchart Generation - Create visual flowcharts and diagrams automatically
• Agentic Orchestration - LangChain-based agent that intelligently routes requests to the appropriate tool

Architecture

• Backend: Python with Flask, LangChain, HuggingFace endpoints
• LLM Models: GLM-4.7 (code generation), Zephyr-7B-beta (basic text queries)
• Vector Store: Supabase for document embeddings
• Rendering: Manim for video generation, Mermaid for flowcharts
• Storage: AWS S3 for generated media files

Tech Stack

• Python > 3.11
• LangChain & LangChain-HuggingFace
• Langgraph (Agentic orchestration)
• Flask
• HuggingFace Inference API
• Supabase (vector database)
• AWS S3
• Manim

LangGraph Orchestration

The backend agent uses a LangGraph StateGraph to route each request through a small execution graph instead of a single monolithic handler.

• classifier_node decides whether the request should go to text generation or Manim video generation.
• planning_node creates a normalized scene plan for video requests.
• scene_executor_node generates, renders, and uploads one scene at a time.
• ffmpeg_node downloads completed scene videos, concatenates them, and uploads the final stitched output to AWS S3.
• text_node runs the RAG text path and preserves both generated content and research metadata in the final response.

Getting Started

# Backend setup
cd backend
pip install -r requirements.txt

# Set environment variables
cp .env.example .env
# Edit .env with your API keys

# Run the server
python src/main.py

Environment Variables

• HUGGINGFACEHUB_API_TOKEN - HuggingFace API key
• SUPABASE_URL - Supabase project URL
• SUPABASE_KEY - Supabase API key
• S3_STORAGE_LINK: Optional - URL for the AWS S3 bucket.
• aws_access_key_id - AWS access key
• aws_secret_access_key - AWS secret key
• aws_region - AWS region
• PARALLEL_API_KEY - Parallel-AI's api key