✔️Comprehensive access interfaceDingoDB provides comprehensive access interfaces, supporting various flexible access modes such as SQL, SDK, and API to meet the needs of different developers. Additionally, it introduces Table and Vector as first-class citizen data models, providing users with efficient and powerful data processing capabilities.

✔️Built-in data high availability

DingoDB provides fully functional and highly available built-in configurations without the need to deploy any external components, which can significantly reduce users' deployment and operation and maintenance costs and significantly improve the efficiency of system operation and maintenance.

✔️Fully automatic elastic data sharding

DingoDB supports dynamic configuration of data shard size, automatic splitting and merging, realizing efficient and friendly resource allocation strategies, and easily responding to various business expansion needs.

✔️Scalar-vector hybrid retrieval

DingoDB supports both traditional database index types and various vector index types, providing a seamless scalar and vector hybrid retrieval experience, reflecting industry-leading retrieval capabilities. In addition, it also supports fusion of scalars, vectors and text. Distributed transaction processing.

✔️Built-in real-time index optimization

DingoDB can build scalar and vector indexes in real time, providing users with unconscious background automatic index optimization. At the same time, it ensures no delays during data retrieval.

✔️Cold-Hot Tiered Retrieval for Massive Datasets

DingoDB provides disk-based vector search capabilities to minimize memory consumption, and supports dynamic switching between different indexes based on data scale requirements.

• All DocumentationDingoDB Docs

• InstallHow to install and deployDocker orAnsible

• UsageHow to use DingoDBUsage

• VS CodeWe recommendVS Code to develop the DingoDB codebase.

✔️POSIX Compliance

DingoFS delivers a native file system-like operational experience, enabling seamless system integration.

✔️AI-Native Architecture

Deeply optimized for large language model (LLM) workflows, efficiently managing massive training datasets and checkpoint workloads.

✔️S3 Protocol Compatibility

DingoFS supports standard S3 interface protocols for streamlined access to filesystem namespace resources.

✔️Fully Distributed Architecture

DingoFS's metadata Service (MDS), data storage layer, caching system, and client components all support linear scalability.

✔️Exceptional Performance

Combines SSD-level low-latency responsiveness with object storage-grade elastic throughput capacity.

✔️Intelligent Caching Acceleration System

DingFS implements a three-tier caching topology (memory/local SSD/distributed cluster) to deliver high-throughput, low-latency intelligent I/O acceleration for AI workloads.

• All DocumentationDingoFS Docs

1. Setup Dingo-eureka and Dingo-sdkIf you installed the software using aDocker container, the container already includes pre-integratedDingo-eureka andDingo-sdk, no additional installation is required.

2. Install jemalloc

wget https://github.com/jemalloc/jemalloc/releases/download/5.3.0/jemalloc-5.3.0.tar.bz2tar -xjvf jemalloc-5.3.0.tar.bz2cd jemalloc-5.3.0&& ./configure&& make&& make install

3. Download dep

git submodule syncgit submodule update --init --recursive

4. Build Etcd Client

bash build_thirdparties.sh

5. Build

mkdir buildcd buildcmake ..make -j 32

• Install DependenciesWe recommend Rocky and Ubuntu to develop the DingoFS codebase.
• GCC 13We recommend using GCC 13 as the primary compiled language.

✔️ Mirror Acceleration

Cache the resources downloaded for the first time. When the client makes a subsequent request, the data will be read from the cache and returned, greatly improving the download rate.

✔️ Convenient Access

There is no need for scientific internet access or complex network configuration. Simply deploy the DingoSpeed service and use it as the proxy address to easily complete the download.

✔️ Traffic Reduction and Load Alleviation

Download once and use multiple times, reducing the traffic waste caused by repeated downloads, which is efficient and saves traffic.

✔️ Localized Management

Cover the entire process of local compilation, deployment, monitoring, and usage of the mirror service, bringing an excellent and flexible experience. It avoids reliance on external networks and public mirror repositories, significantly improving the system's response speed and data security.

The project uses the wire command to generate the required dependency code. Install the wire command as follows：

# Import into the projectgo get -u github.com/google/wire# Install the commandgo install github.com/google/wire/cmd/wire

Wire is a flexible dependency injection tool that completes dependency injection at compile time by automatically generating code. In the dependency relationships between various components, explicit initialization is usually used instead of passing global variables. Therefore, using Wire to initialize the code can effectively solve the coupling between components and improve code maintainability.

This project uses go mod to manage dependencies and requires Go version 1.23 or higher. It uses makefile to manage the project and requires the make command.

# 1. Install dependenciesmake init# 2. Generate codemake wire# 3. Compile the executable file for the current system versionmake build# 4. Compile the Linux executable file on macOSmake macbuild# 5. Add a license to each filemake license

💡 Deploy the compiled binary file and execute ./dingo-hfmirror to start the service. Then set the environment variable HF_ENDPOINT to the mirror site (here it ishttp://localhost:8090/).

Linux:

export HF_ENDPOINT=http://localhost:8090

Windows Powershell:

$env:HF_ENDPOINT = "http://localhost:8090"

From now on, all download operations in the Hugging Face library will be proxied through this mirror site. You can install the Python library to try it out:

pip install -U huggingface_hub

from huggingface_hub import snapshot_downloadsnapshot_download(repo_id='Qwen/Qwen-7B', repo_type='model',local_dir='./model_dir', resume_download=True,max_workers=8)

Alternatively, you can use the Hugging Face CLI to directly download models and datasets. Download GPT2:

huggingface-cli download --resume-download openai-community/gpt2 --local-dir gpt2

Download a single file:

huggingface-cli download --resume-download --force-download  HuggingFaceTB/SmolVLM-256M-Instruct config.json

Download WikiText:

huggingface-cli download --repo-type dataset --resume-download Salesforce/wikitext --local-dir wikitext

You can view the path ./repos, where the caches of all datasets and models are stored.