Lighter than MySQL, Stronger than SQLite: Finally, Someone Got the AI Database Right
# Rethinking Databases for the AI Era
To **free AI developers from the burden of complex data infrastructure** and let them focus on algorithms and application logic, a **new kind of database** is needed — one designed for agility, lightness, and rapid iteration.
## The Current Landscape
Recently, Turso founder Pekka shared an insightful perspective:
> SQLite is considered the ideal database for AI agents because it is lightweight and fits many AI agent scenarios — but it still needs to evolve.
This sparked lively discussions.
Some showcased projects like *Super SQLite*, while others debated various evolved SQLite-based products and their pros/cons.
The **need for lightweight databases** extends far beyond AI agents.
Traditional databases — once ideal for massive web and mobile applications — now struggle with:
- **High deployment complexity**
- **Heavy resource consumption**
- **Over-engineering that hurts efficiency**
Moreover, many modern AI apps run **locally** on smartphones or IoT devices, with tightly coupled data and models.
The classic client-server remote database model often fails to meet **embedded or offline** requirements.
---
## 0x01 — The Origins of *seekdb*
OceanBase began with transaction and payment workloads for **Taobao** and **Alipay**, later serving banks and other industries. See: [An open-source project polished over 15 years](https://mp.weixin.qq.com/s?__biz=MzA5MzYyNzQ0MQ==&mid=2247516865&idx=1&sn=b00aec296af264d486d5220f0da8082e&scene=21#wechat_redirect).
But in the AI era, **SQLite is too limited** (especially for combining vectors with SQL), while traditional DBs remain heavy and complex.
**OceanBase created seekdb** — a free, open-source, **lightweight database with AI search capabilities** that runs effortlessly on personal laptops.
> GitHub: [github.com/oceanbase/seekdb](https://github.com/oceanbase/seekdb)
---
## 0x10 — What is seekdb?
**seekdb** is OceanBase’s **out-of-the-box, AI-ready lightweight database** for developers, supporting:
- **Unified vector, full-text, and multi-modal search**
- **Hybrid query capabilities** for building AI apps faster
It inherits OceanBase’s high-performance engine from Taobao & Alipay, **compatible with MySQL**, but optimized for **AI data search**.
**Product capability comparison:**
> Note: MySQL 8.0 removed embedded capabilities
---
## Why seekdb Matters for AI Creators
Tools like seekdb fit seamlessly into AI creation pipelines.
Paired with platforms like [AiToEarn官网](https://aitoearn.ai), developers and creators can:
- Build & store AI data efficiently
- Publish across multiple channels
- Analyze results and monetize AI content
---
## Key Application Scenarios
### 1. Retrieval-Augmented Generation (RAG)
- Built-in **document parsing**, **vector embedding**, **reranking** and **LLM** integration
- Supports **vector/full-text/scalar hybrid search**
- Delivers **Doc In, Data Out** in one database instance
### 2. AI-Assisted Programming
- Vector & full-text indexing for code repos
- Powerful semantic search for code completion
### 3. AI Agent Platforms
- Quick startup & embedded deployment
- High-frequency CRUD operations with real-time queries
- Vector/full-text/hybrid search with flexible metadata
### 4. Semantic Search Engines
- E-commerce product search, multimedia retrieval
- Hybrid Index hides vector complexity from users
### 5. MySQL Modernization + AI Upgrade
- MySQL-compatible with AI capabilities
- Suitable for edge IoT, teaching, OLTP, HTAP, or AI workloads
---
## Product Architecture
**Deployment Modes:** Client/Server & Embedded (integrates directly into Python apps)
**Storage Layer:** ACID transactions, unified row-column LSM-Tree engine with compression
**Index Layer:**
- Vectors, text, JSON, GIS, arrays
- HNSW/IVF vector indexes
- BM25 full-text search
- Hybrid Index for semantic search
- JSON, primary/secondary, and GIS indexes
**Compute Layer:**
- Hybrid searches (vector + text + scalar)
- In-database AI functions for inference
- Full ACID, MVCC, advanced optimizer, vectorized execution
**Application Interface:**
- Fully MySQL-compatible
- SQL + SDK for hybrid search
- Integrates with LangChain, LlamaIndex, Dify
- MCP Server for AI ecosystem integration
---
## Core Features
- **Ready-to-use**: Single-node architecture, deployable via yum, Docker, Windows/macOS, Python embedded
- **Runs on small configs**: E.g., 1 CPU core + 2GB RAM
- **Vector Search**: Up to 16K dimensions
- **Full-Text Search**: BM25 scoring, multiple tokenizers
- **Hybrid Search**: Single SQL query for multiple modes
- **Built-in AI Functions**: `AI_COMPLETE`, `AI_PROMPT`, `AI_EMBED`, `AI_RERANK`
- **JSON Dynamic Schema** with multi-value indexes
- **Immediate search after data ingestion**
- **HTAP Capabilities**: OLTP + real-time analytics
---
## Quick Deployment
> For detailed requirements, see `deploy-seekdb-testing-environment` in the docs.
### **1. yum Installation**
**Add Mirrors:**sudo yum-config-manager --add-repo https://mirrors.aliyun.com/oceanbase/OceanBase.repo
**Install:**sudo yum install seekdb obclient
**Config (`/etc/oceanbase/seekdb.cnf`):**port=1234
base-dir=/var/lib/oceanbase
data-dir=/var/lib/oceanbase/store
redo-dir=/var/lib/oceanbase/store/redo
datafile_size=2G
datafile_next=2G
datafile_maxsize=50G
cpu_count=4
memory_limit=2G
log_disk_size=2G
**Control Service:**sudo systemctl start seekdb
sudo systemctl status seekdb # check if Service is ready
obclient -h127.0.0.1 -uroot -P1234 -A oceanbase
sudo systemctl stop seekdb
sudo systemctl enable seekdb # auto-start on boot
---
### **2. Embedded Python Usage**import seekdb
seekdb.open()
conn = seekdb.connect()
cursor = conn.cursor()
cursor.execute("select version();")
print(cursor.fetchall())
conn.close()
Output:[('1.2.34-OceanBase NewProduct-v5.6.7.8',)]
---
### **3. Other Deployment Options**
- **Docker image** available
- **Desktop management tool**
See: [oceanbase.ai/docs](https://oceanbase.ai/docs)
---
## 0xFF — Final Thoughts
Last year, running OceanBase on my MacBook Pro sometimes failed due to memory limits. It was **powerful but distant**.
Now, seekdb brings **financial-grade technology** *within reach*.
It’s light, fast, developer-friendly, and AI-ready.
> GitHub: [github.com/oceanbase/seekdb](https://github.com/oceanbase/seekdb)
And for AI creators, combining seekdb with [AiToEarn官网](https://aitoearn.ai) enables **generation → publishing → analytics → monetization** across platforms (Douyin, Kwai, WeChat, Bilibili, Xiaohongshu, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, X).
---
**- END -**
