Skip to content

v1.0.0 — Initial Release

Latest
Latest

Choose a tag to compare

View all tags
@mieitza mieitza released this 01 Jun 08:17
· 4 commits to main since this release
v1.0.0
a21ade9

AetherDB v1.0.0

AI-native database with semantic search, GraphRAG, hybrid search, full-text search, and edge deployment on Cloudflare Workers.

Highlights

  • 3 search modes: semantic (vector), hybrid (vector + knowledge graph), full-text (FTS5)
  • GraphRAG: auto entity/relationship extraction via Workers AI (GLM-4.7-Flash)
  • Edge-first: runs on Cloudflare Workers + D1 + Vectorize + Workers AI
  • MCP server: 12 tools for AI assistant integration
  • Client SDKs: TypeScript and Python

Features

Core

  • Document CRUD with content deduplication and versioning
  • Bulk import (up to 1000 docs per batch)
  • Per-token rate limiting (60 RPM default)
  • Read-only SQL query proxy

Search

  • Semantic search — auto-embedding via Workers AI (bge-base-en-v1.5, 768-dim), indexed in Cloudflare Vectorize
  • Hybrid search — merges vector similarity with knowledge graph traversal: score = (1 - graph_weight) * vector_score + graph_weight * graph_score
  • Full-text search — SQLite FTS5 with MATCH syntax, snippets, BM25 ranking
  • Metadata filtering — filter vector search by string/number/boolean fields

Knowledge Graph

  • 7 entity types: person, organization, concept, location, event, technology, other
  • 6 relationship types: related_to, works_at, located_in, part_of, created_by, uses
  • Multi-hop bidirectional traversal via recursive CTEs
  • Async extraction (<1s response, entities appear in ~5s)

Sync

  • CRDT sync with Hybrid Logical Clock timestamps
  • Push/pull protocol for offline-first operation

Integrations

  • MCP server (12 tools) for Claude and other AI assistants
  • TypeScript and Python client SDKs
  • OpenAPI 3.1 specification

Quick Start

cd cf-worker
npm install
cp wrangler.toml.example wrangler.toml
# Edit with your Cloudflare account_id and database_id

npx wrangler d1 create aetherdb
npx wrangler vectorize create aetherdb-vectors --dimensions=768 --metric=cosine
npx wrangler secret put API_TOKEN
npx wrangler deploy

See CONTRIBUTING.md for full setup.

Tech Stack

Component Technology
Edge Cloudflare Workers + D1 + Vectorize + Workers AI
Embeddings Workers AI (bge-base-en-v1.5)
Entity Extraction Workers AI (GLM-4.7-Flash)
Core Rust 2024 edition (1.85+) / Axum / libSQL
Orchestration Docker Compose / Kubernetes (Helm)
Assets 2
Loading