Show HN：Satya – 適用於偏鄉學校的離線優先 AI 輔導員 (Phi-1.5 與 RAG)

Hacker News·4 個月前

Hacker News 上的「Show HN」貼文介紹了 Satya，這是一款專為偏鄉學校教育設計的離線優先 AI 輔導員。它利用 Phi-1.5 模型和檢索增強生成 (RAG) 技術來實現其功能。

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

To see all available qualifiers, see our documentation.

A learning companion.

License

Uh oh!

There was an error while loading. Please reload this page.

aa-sikkkk/satya

Folders and files

Latest commit

History

Repository files navigation

Satyá: Learning Companion

An intelligent learning companion built with RAG-powered content discovery and Microsoft's efficient Phi 1.5 model. Delivers education that scales from offline rural classrooms to connected urban schools, all while running smoothly on the hardware you already have.

Smart enough to handle complex questions. Efficient enough to work anywhere. Simple enough for anyone to deploy.

Overview

Satya reimagines AI-powered education for the real world. This comprehensive learning platform is designed for students in Nepal, bringing intelligent tutoring to any environment. Whether in a connected classroom, a rural school, or a home with unreliable internet, Satya supports every learner with RAG-enhanced content discovery and the efficient Phi 1.5 model.

Mission & Vision

Our Mission

To democratize AI-powered education by making intelligent tutoring accessible to every student, regardless of their location, internet connectivity, or hardware resources.

Satya exists to bridge the digital divide in education. While AI transforms learning in well-connected urban centers, millions of students in rural areas remain excluded. We believe every student deserves access to intelligent, personalized learning assistance—not just those with high-speed internet and modern devices.

The Educational Crisis

Important

2.9 billion people worldwide lack reliable internet access. In Nepal alone, 60% of students study in rural areas with limited connectivity and outdated hardware.

Current barriers to AI education:

The result: Educational inequality widens as AI advances accelerate in privileged areas while underserved communities fall further behind.

Our Solution: Offline-First AI Education

Satya breaks down these barriers through radical accessibility:

Offline-First Architecture
Low-Resource Optimization
Intelligent RAG System
Single Model Efficiency
Community-Driven Content

Impact & Reach

Target Beneficiaries:

Measurable Outcomes:

Design Philosophy

Note

Every technical decision in Satya prioritizes accessibility over performance, simplicity over features, and offline capability over cloud convenience.

Core Principles:

Why This Matters

Education is a fundamental right, not a privilege. AI-powered learning should be accessible to every student, not just those in well-connected urban centers.

Satya proves that intelligent, personalized education doesn't require expensive infrastructure. With thoughtful engineering and community collaboration, we can deliver AI tutoring to the students who need it most—those currently excluded from the AI revolution.

This isn't just about technology. It's about educational justice.

Key Features

Student-Facing Features

Tip

The RAG system searches both textbooks and notes collections automatically, providing comprehensive answers from multiple sources.

Teacher-Facing Features

Note

Use scripts/ingest_content.py for all content ingestion. It replaces all previous ingestion scripts.

System Architecture

High-Level Architecture

Important

Architecture has been updated in version 2.0. Single Phi 1.5 model replaces previous multi-model approach.

Component Architecture

Implementation (scripts/ingest_content.py)

Processing Flow:

Implementation (system/rag/rag_retrieval_engine.py)

Retrieval Flow:

Tip

The system provides progressive status updates to make the 10-12 second retrieval feel faster.

Single Phi 1.5 Model Handler (ai_model/model_utils/phi15_handler.py)

Model Configuration:

Technical Specifications

Dependencies

Performance Targets

Note

On 3rd gen i3 CPU with 4GB RAM, expect 10-12 second TTFT (Time To First Token). This is normal for quality RAG on CPU-only systems.

File Structure

Installation

Prerequisites

Step-by-Step Installation

Linux/macOS:

Windows:

Tip

For OCR support, install optional dependencies:

Download the Phi 1.5 GGUF model:

Model Sources:

Important

Place the model file in satya_data/models/phi_1_5/ and ensure it has a .gguf extension.

Quick Start

CLI Mode

GUI Mode

Note

First run will take 5-10 seconds to load the model. Subsequent runs are faster.

Content Management

Adding Educational Content

Important

Use the universal ingestion script for all content types. It replaces all previous ingestion scripts.

Process all content (textbooks + notes):

Process only textbooks:

Process only notes:

Auto-detect (recommended):

Force OCR on all PDFs:

Never use OCR (text-only):

Content Organization

Textbooks:

Notes:

Tip

The system auto-detects grade from folder structure and subject from filename.

Verification

Documentation

Usage Guide

Student Interface

The system supports subject-related questions:

Answer Generation Process:

Tip

The progressive status updates make the 10-12 second retrieval feel faster by showing what's happening.

The system only shows confidence warnings when needed:

API Reference

Model Handler

Location: ai_model/model_utils/model_handler.py

RAG Retrieval Engine

Location: system/rag/rag_retrieval_engine.py

Troubleshooting

Common Issues

Warning

Ensure model file exists in satya_data/models/phi_1_5/ with .gguf extension.

Solutions:

Note

On i3 CPU with 4GB RAM, 10-12 second TTFT is normal for quality RAG retrieval.

Optimization:

Solutions:

Note

This is normal - model inference runs in background thread. Wait for completion.