[!NOTE] This README was generated by SKILL, get the ZH version from here.

A READ-ONLY RAG DATABASE SERVICE — DROP FILES, SEARCH INSTANTLY

A Go RAG database service with dual keyword and semantic search, filesystem watcher auto-indexing, and OpenAI embedding vector cache

Table of Contents

• Features
• Architecture
• License
• Author

Features

go install github.com/agenvoy/kuradb/cmd/app@latest · Documentation

• Keyword + Semantic Dual Search — Combines Chinese tokenizer keyword matching with OpenAI embedding vector similarity search for both precision and semantic understanding.
• Filesystem Watcher Auto-Indexing — Drop files into a watched directory and they are automatically parsed, chunked, and embedded — no manual indexing needed.
• Read-Only API Security Boundary — Only query endpoints are exposed externally; all writes flow through the watcher → parser → SQLite unidirectional pipeline, preventing external tampering.
• Vector Cache & Query Cache — In-memory cosine similarity search paired with OpenAI query embedding cache delivers near-zero latency for repeated queries.
• Single Binary Deployment — A statically compiled Go binary with embedded SQLite, zero external dependencies, deployable with a single go install.

Architecture

Full Architecture

graph TB
    Client[Client] --> API[HTTP API]
    API --> Health[Health Check]
    API --> List[Database List]
    API --> Semantic[Semantic Search]
    API --> Keyword[Keyword Search]
    Watcher[File Watcher] --> Parser[File Parser]
    Parser --> SQLite[(SQLite)]
    SQLite --> Embedder[OpenAI Embedding]
    Embedder --> Vector[Vector Cache]
    Vector --> Semantic
    SQLite --> Keyword

License

This project is licensed under the MIT LICENSE.

Author

邱敬幃 Pardn Chiu

hi@pardn.io
https://www.linkedin.com/in/pardnchiu

©️ 2026 邱敬幃 Pardn Chiu