Welcome to the Sorami repository, a collection of interconnected Generative AI projects developed during the Free GenAI Bootcamp 2025. The central hub for many of these tools is the lang-portal.
Here's a breakdown of the projects included:
agent.mp4
An AI agent designed for enhanced internet searching and comparison tasks. It leverages graph-based processing and various APIs (including email/web integration) to provide comprehensive results. The agent can:
- Perform complex web searches and data aggregation
- Compare results from multiple sources
- Assist with language-related tasks
- Integrate with the ASL project for cross-modal language processing
- Send formatted email reports with search findings
Focuses on American Sign Language (ASL) recognition and learning. This project combines computer vision and machine learning to:
- Process and analyze sign language from images and videos
- Train and utilize machine learning classification models for gesture recognition
- Provide interactive learning tools for ASL alphabets and numbers
- Potentially integrate with the
agentproject for comprehensive language tasks
Contains architectural diagrams and resources related to designing and building the Generative AI solutions within this repository. Includes:
- System architecture diagrams in Mermaid format
- Visual representations of component interactions
- Design patterns and best practices for GenAI applications
- Planning documents for future Sorami projects
Holds Kubernetes deployment configurations for containerizing and orchestrating the various Sorami applications. Features:
- Standard deployment manifests for all microservices
- Resource allocation and scaling policies
- Service discovery and networking configurations
- Persistent storage solutions for stateful components
The central portal connecting many Sorami projects. This core application serves as the main user interface and:
- Features a Go backend for high-performance API handling
- Implements a Next.js frontend for responsive user interactions
- Provides a unified interface to access all Sorami tools
- Manages shared data resources (e.g.,
words.db)
A listening comprehension tool, specifically geared towards JLPT preparation. This project:
- Implements Retrieval-Augmented Generation (RAG) to process audio transcripts
- Generates relevant questions based on audio content
- Provides difficulty-adjusted practice materials
- Tracks user progress through listening exercises
- Offers transcript search functionality for targeted practice
quiz.mp4
Generates practice questions, particularly for the JLPT and other language proficiency tests. The system:
- Utilizes LLM APIs to create contextually relevant quiz content
- Adapts question difficulty based on user performance
- Supports various question formats (multiple-choice, fill-in-blank, etc.)
- Saves question sets in structured JSON format for reuse
- Provides API endpoints for integration with other Sorami tools
sencon.mp4
A tool aiding in Japanese sentence construction with support for multiple language learning scenarios. Features include:
- Support for multiple Large Language Models (ChatGPT, Claude, Gemini, Llama)
- Japanese speech capabilities for pronunciation practice
- Grammar pattern exercises and corrections
- Context-aware sentence generation
- Vocabulary usage demonstrations in proper context
speech.mp4
speech.anaylsis.mp4
Analyzes spoken Japanese to provide feedback on pronunciation, intonation, and fluency. The system:
- Processes audio input to detect speech patterns
- Compares user pronunciation with native speech models
- Highlights areas for improvement in pronunciation
- Offers targeted exercises for difficult sounds
- Tracks progress over time with detailed metrics
2025-04-11.15-16-58.mp4
Helps language learners efficiently import and manage vocabulary lists. This tool:
- Extracts vocabulary from various source formats (text, PDF, websites)
- Organizes words by frequency, JLPT level, or custom categories
- Generates example sentences for each vocabulary item
- Integrates with the central language portal database
- Provides API endpoints for vocabulary retrieval
hello.mp4
An interactive web application for practicing Japanese writing, featuring word and sentence exercises with real-time feedback. Key features include:
- Word practice mode with random vocabulary generation
- Sentence practice with AI-generated example sentences
- Handwriting input and OCR re
cognition
- AI-powered grading and feedback on writing accuracy
- Progress tracking across different writing exercises
A tool for translating video content between languages, with special focus on educational materials. This project:
- Extracts audio from video files for processing
- Transcribes speech to text in the source language
- Translates content to target language
- Generates subtitles or dubbed audio
- Preserves timing and context across translations
To begin exploring the Sorami project collection:
- Clone this repository
- Navigate to the specific project directory you're interested in
- Follow the setup instructions in each project's individual README
- Most projects can be run using Docker Compose with the provided configuration files
The Sorami project collection leverages several key technologies:
- Large Language Models (Various providers)
- Python with specialized AI and NLP libraries
- Go for high-performance backend services
- Next.js for interactive frontends
- Docker and Kubernetes for containerization
- PostgreSQL and SQLite for data storage