qri is a global dataset version control system (GDVCS) built on the distributed web

Breaking that down:

• global so that if anyone, anywhere has published work with the same or similar datasets, you can discover it.
• Specific to datasets because data deserves purpose-built tools
• version control to keep data in sync, attributing all changes to authors
• On the distributed web to make all of the data published on qri simultaneously available, letting peers work on data together.

If you’re unfamiliar with version control, particularly the distributed kind, well you're probably viewing this document on github — which is a version control system intended for code. Its underlying technology – git – popularized some magic sauce that has inspired a generation of programmers and popularized concepts at the heart of the distributed web. Qri is applying that family of concepts to four common data problems:

1. Discovery Can I find data I’m looking for?
2. Trust Can I trust what I’ve found?
3. Friction Can I make this work with my other stuff?
4. Sync How do I handle changes in data?

Because qri is global and content-addressed, adding data to qri also checks the entire network to see if someone has added it before. Since qri is focused solely on datasets, it can provide meaningful search results. Every change on qri is associated with a peer, creating an audit-able trail you can use to quickly see what has changed and who has changed it. All datasets on qri are automatically described at the time of ingest using a flexible schema that makes data naturally inter-operate. Qri comes with tools to turn all datasets on the network into a JSON API with a single command. Finally, all changes in qri are tracked & synced.

Qri is comprised of many specialized packages. Below you will find a summary of each package.

Package	Go Docs	Go Report Card	Description
api			user accessible layer, primarily made for communication with our frontend webapp
cmd			our command line interface
config			user configuration details, includes peer's profile
lib			takes arguments from the cmd and api layer and forms proper requests to call to the action layer
p2p			the peer to peer communication layer of qri
repo			the repository: saving, removing, and storing datasets, profiles, and the config
dataset			the blueprint for a dataset, the atoms that make up qri
registry			the blueprint for a registry: the service that allows profiles to be unique and datasets to be searchable
starlib			the starlark standard library available for qri transform scripts
qfs			"qri file sytem" is Qri's file system abstraction for getting & storing data from different sources
ioes			package to handle in, out, and error streams: gives us better control of where we send output and errors
jsonschema			used to describe the structure of a dataset, so we can validate datasets and determine dataset interop

The following packages are not under Qri, but are important dependencies, so we display their latest versions for convenience.

Package	Version
ipfs

To build qri you'll need the go programming language on your machine. We require at least go version 1.13 to build qri.

If you are new to using go, you should set your PATH environment variable to include the location that your go tool installs binaries to, which is usually ~/go/bin. For more information about environment variables see this tutorial.

Building then depends upon your operating system:

Having go installed is enough, proceed to building.

To start, make sure that you have enabled Developer Mode. A library that we depend on needs it enabled in order to properly handle symlinks. If not done, you'll likely get the error message "A required privilege is not held by the client".

You should not need to Run As Administrator to build or run qri. We do not recommend using administrator to run qri.

For your shell, we recommend using msys2. Other shells, such as cmd, Powershell, or cygwin may also be usable, but msys2 makes it easy to install our required dependencies. IPFS also recommends msys2, and qri is built on top of IPFS.

Building depends upon having git and make installed. If using msys2, you can easily install these by using the package manager "pacman". In a shell, type:

pacman -S git make

Assuming you've also installed go using the official Windows installer linked above, you will also need to add go to your PATH by modifying your environment variable. See the next section on "Environment variables" for more information.

Due to how msys2 treats the PATH variable, you also need to add a new environment variable MSYS2_PATH_TYPE, with the value inherit, using the same procedure.

Once these steps are complete, proceed to building.

On a Raspberry PI, you'll need to increase your swap file size in order to build. Normal desktop and server linux OSes should be fine to proceed to building.

One symptom of having not enough swap space is the go install command producing an error message ending with:

link: signal: killed

To increase your swapfile size, first turn off the swapfile:

sudo dphys-swapfile swapoff

Then edit /etc/dphys-swapfile as root and set CONF_SWAPSIZE to 1024.

Finally turn on the swapfile again:

sudo dphys-swapfile swapon

Otherwise linux machines with reduced memory will have other ways to increase their swap file sizes. Check documentation for your particular machine.

In your terminal, navigate to some directory that you want to work within. Let's say we're using a directory in our home folder called "go-code".

$ cd go-code
$ git clone https://github.com/qri-io/qri

Once this repository is cloned, enter it and install:

$ cd qri
$ go install

If this is your first time building, this command will have a lot of output. That's good! Its means it's working :) It'll take a minute or two to build.

After this is done, there will be a new binary qri in your ~/go/bin directory if using go modules, and $GOPATH/bin directory otherwise. You should be able to run:

$ qri help

and see help output.