83

GitHub guide explains 2 way to authorize but looks neither of those works with the Release files.

as a result of:

curl -u 'username' -L -o a.tgz https://github.com/company/repository/releases/download/TAG-NAME/A.tgz

There always is something like

<!DOCTYPE html>
<!--
Hello future GitHubber! ...
user-id-14900042
  • 686
  • 4
  • 17
keiw
  • 1,260
  • 1
  • 9
  • 13

9 Answers9

88

To download release file from private repo, you can use Personal access token which can be generated at settings/tokens with Full control of private repositories scope.

Then download the asset with curl command (change with appropriate values):

curl -vLJO -H 'Authorization: token my_access_token' 'https://api.github.com/repos/:owner/:repo/releases/assets/:id'

or if you're using an OAuth app, use:

curl -u my_client_id:my_client_secret https://api.github.com/repos/:owner/:repo/releases/assets/:id

where:

  • :owner is your user or organisation username;
  • :repo is your repository name;
  • :id is your asset id, can be found in tag release URL, like:

    https://api.github.com/repos/:owner/:repo/releases/tags/:tag 
    
  • :token is your personal access token (can be created at /settings/tokens;

Note: Using access_token as a query param is deprecated.

See: Repositories API v3 at GitHub


Here is the Bash script which can download asset file given specific name of file:

#!/usr/bin/env bash
# Script to download asset file from tag release using GitHub API v3.
# See: http://stackoverflow.com/a/35688093/55075    
CWD="$(cd -P -- "$(dirname -- "$0")" && pwd -P)"

# Check dependencies.
set -e
type curl grep sed tr >&2
xargs=$(which gxargs || which xargs)

# Validate settings.
[ -f ~/.secrets ] && source ~/.secrets
[ "$GITHUB_API_TOKEN" ] || { echo "Error: Please define GITHUB_API_TOKEN variable." >&2; exit 1; }
[ $# -ne 4 ] && { echo "Usage: $0 [owner] [repo] [tag] [name]"; exit 1; }
[ "$TRACE" ] && set -x
read owner repo tag name <<<$@

# Define variables.
GH_API="https://api.github.com"
GH_REPO="$GH_API/repos/$owner/$repo"
GH_TAGS="$GH_REPO/releases/tags/$tag"
AUTH="Authorization: token $GITHUB_API_TOKEN"
WGET_ARGS="--content-disposition --auth-no-challenge --no-cookie"
CURL_ARGS="-LJO#"

# Validate token.
curl -o /dev/null -sH "$AUTH" $GH_REPO || { echo "Error: Invalid repo, token or network issue!";  exit 1; }

# Read asset tags.
response=$(curl -sH "$AUTH" $GH_TAGS)
# Get ID of the asset based on given name.
eval $(echo "$response" | grep -C3 "name.:.\+$name" | grep -w id | tr : = | tr -cd '[[:alnum:]]=')
#id=$(echo "$response" | jq --arg name "$name" '.assets[] | select(.name == $name).id') # If jq is installed, this can be used instead. 
[ "$id" ] || { echo "Error: Failed to get asset id, response: $response" | awk 'length($0)<100' >&2; exit 1; }
GH_ASSET="$GH_REPO/releases/assets/$id"

# Download asset file.
echo "Downloading asset..." >&2
curl $CURL_ARGS -H "Authorization: token $GITHUB_API_TOKEN" -H 'Accept: application/octet-stream' "$GH_ASSET"
echo "$0 done." >&2

Before running, you need to set your GITHUB_API_TOKEN with your GitHub token (see: /settings/tokens at GH). This can be placed in your ~/.secrets file, like:

GITHUB_API_TOKEN=XXX

Example script usage:

./get_gh_asset.sh :owner :repo :tag :name

where name is your filename (or partial of it). Prefix script with TRACE=1 to debug it.


In case you wonder why curl fails sometimes with (as mentioned in other answer):

Only one auth mechanism allowed; only the X-Amz-Algorithm query parameter, Signature query string parameter or the Authorization header should be specified.

when running like:

curl -vLJ -H 'Authorization: token <token>' -H 'Accept: application/octet-stream' https://api.github.com/repos/:owner/:repo/releases/assets/<id>

this is because you're specifying multiple mechanism at the same time, so S3 server doesn't know which one to use, therefore you have to choose only one, such as:

  • X-Amz-Algorithm query parameter
  • Signature query string parameter (X-Amz-Signature)
  • Authorization header (Authorization: token <token>)

and since GitHub redirects you from asset page (when requesting application/octet-stream), it populates credentials automatically in query string and since curl is passing over the same credentials in the request header (which you've specified), therefore they're conflicting. So as for workaround you can use access_token instead.

kenorb
  • 155,785
  • 88
  • 678
  • 743
  • 1
    Thanks for the script @kenorb, I will use it to allow my ansible playbook to grab release .deb files from a private repo. Much appreciated. – Russell Ormes Feb 22 '17 at 12:01
  • 1
    Thanks for the script. If the `jq` JSON utility is installed, extracting the asset id can be greatly simplified with `ASSET_ID=$(echo "$response" | jq --arg name "$name" '.assets[] | select(.name == $name).id')` or the likes. – Max Jun 28 '17 at 08:47
  • 1
    NOTE: using access_token as query param is deprecated: https://developer.github.com/changes/2020-02-10-deprecating-auth-through-query-param/#changes-to-make – joonas.fi Feb 19 '20 at 12:01
  • 1
    @joonas.fi Thanks for the suggestion, I've updated my answer. – kenorb Feb 19 '20 at 12:10
  • 15
    `Full control of private repositories scope` for getting just the release data? Is there any other scope I can use, coz I just don't want to authorize everything but `only releases` – Ecto Ruseff Feb 12 '21 at 04:19
  • 1
    agree with @EctoRuseff, it is scary the access level to do this :o – eglasius Mar 12 '21 at 10:08
  • Great script, many thanks! Maybe `awk` should also be added to the dependency check? BTW: Do you have it published on some public code repository? – Benjamin W. Bohl Jan 12 '23 at 09:00
  • @EctoRuseff Thankfully there is a new BETA on GitHub that's called Fine-grained tokens which solves exactly this issue (selective repos, selective permissions (read-only, write, ...) – Programmer Feb 16 '23 at 14:07
  • 1
    How do I get asset_id? – panoet May 29 '23 at 23:25
27

We had to download release assets from private GitHub repos fairly often, so we created fetch, which is an open source, cross-platform tool that makes it easy to download source files and release assets from a git tag, commit, or branch of public and private GitHub repos.

For example, to download the release asset foo.exe from version 0.1.3 of a private GitHub repo to /tmp, you would do the following:

GITHUB_OAUTH_TOKEN="your token"
fetch --repo="https://github.com/foo/bar" --tag="0.1.3" --release-asset="foo.exe" /tmp
Yevgeniy Brikman
  • 8,711
  • 6
  • 46
  • 60
Josh Padnick
  • 3,157
  • 1
  • 25
  • 33
  • 5
    Just adding a comment here that fetch is a standalone executable, and does exactly what the original poster is requesting. I've incorporated it into my team's build process as well. – Steve Hollasch May 26 '17 at 05:52
  • 2
    ONLY answer that actually worked for me. Tried to get one specific asset from a github release at a private repo, and only this one did the trick. – jimh Sep 24 '19 at 23:49
  • I had to actually specify the `--github-oauth-token=${GITHUB_OAUTH_TOKEN}` flag. Works fantastically then. – M. Rodo Jun 22 '23 at 07:04
10

Seems both authentication methods only work for the API endpoints. There's an API endpoint for downloading release assets (docu):

GET /repos/:owner/:repo/releases/assets/:id

But one will need to know the asset's numerical ID for that. I asked their support whether they could add an API endpoint to download release assets by name (like they have for tarballs).


Update: Response from Github support:

We don't offer any API like you're suggesting. We try to avoid brittle URLs that would change if the tag or filename of the release asset change. I'll take your feedback to the team to discuss further.

Xavi
  • 20,111
  • 14
  • 72
  • 63
njam
  • 1,233
  • 14
  • 16
  • I'm looking for a easy way to get the latest asset. As long as I've understood I need to get the list of releases, pick the id of the latest, get the release with the asset ids and download finally the asset. Is there a way to download the asset by the tag name? – Simon Jul 04 '14 at 07:53
  • There's sort of an example of the technique here: http://brantiffy.axisj.com/archives/215, https://github.com/brant-hwang/get-git-private-repo-latest-release-zip. You can probably tweak the code to use tag name instead of ID, etc. – David Sep 22 '15 at 19:21
6

I found out the answer in this comment: https://github.com/request/request/pull/1058#issuecomment-55285276

curl is forwarding the authentication header in the request to the AmazonS3 bucket, where the Github release assets are stored. Error response from S3:

<?xml version="1.0" encoding="UTF-8"?>
<Error>
 <Code>InvalidArgument</Code>
 <Message>
   Only one auth mechanism allowed; only the X-Amz-Algorithm query parameter, Signature query string parameter or the Authorization header should be specified
 </Message>
 <ArgumentName>Authorization</ArgumentName>
 <ArgumentValue>token <yourtoken> </ArgumentValue><RequestId>4BEDDBA630688865</RequestId> <HostId>SXLsRKgKM6tPa/K7g7tSOWmQEqowG/4kf6cwOmnpObXrSzUt4bzOFuihmzxK6+gx</HostId>
</Error>

One line wget solution:

wget --auth-no-challenge --header='Accept:application/octet-stream' https://<token>:@api.github.com/repos/:owner/:repo/releases/assets/:id -O app.zip

Try:curl -i -H "Authorization: token <token>" -H "Accept:application/octet-stream" https://<token>:@api.github.com/repos/:owner/:repo/releases/assets/:id, for some more details. Add -L to see the S3 error message.

phedoreanu
  • 2,588
  • 3
  • 25
  • 29
  • Can you explain how you would generate the `` in your solutions? Sorry if it's a dumb question. – Dominic Sayers Apr 27 '15 at 17:49
  • 1
    Think I've answered my own question. Github API authentication is documented here: https://developer.github.com/v3/#authentication – Dominic Sayers Apr 27 '15 at 17:52
  • if you use curl, the workaround is to put the authentication elsewhere than as a header. Like as query string parameter method (per the API docs, "URL?access_token=TOKEN") . Can you also do "http://TOKEN:@api.github.com/..." with curl? – David Sep 22 '15 at 19:17
6

Here is curl & jq one;)liner:

CURL="curl -H 'Authorization: token <auth_token>' \
      https://api.github.com/repos/<owner>/<repo>/releases"; \
ASSET_ID=$(eval "$CURL/tags/<tag>" | jq .assets[0].id); \
eval "$CURL/assets/$ASSET_ID -LJOH 'Accept: application/octet-stream'"

Change parts surrounded with <> with your data. To generate auth_token go to github.com/settings/tokens

If you like to login with password use this (note it will ask for password twice):

CURL="curl -u <github_user> https://api.github.com/repos/<owner>/<repo>/releases"; \
ASSET_ID=$(eval "$CURL/tags/<tag>" | jq .assets[0].id); \
eval "$CURL/assets/$ASSET_ID -LJOH 'Accept: application/octet-stream'"
tworec
  • 4,409
  • 2
  • 29
  • 34
  • 1
    You can replace `tags/` with `latest` to get the latest release. – Miguel Oct 29 '20 at 19:01
  • The latest release seems to need just `/latest` now, i.e. `"$CURL/latest"` – fooquency Nov 21 '21 at 22:07
  • The new Github token system seems to require the -u method, i.e. `-u username:token` – fooquency Nov 21 '21 at 22:07
  • You assume that your target will always be the first item in the `.assets` array. That may not be the case. That's why others use `jq` with `select()` to search for the correct `.name`. – Cliff Aug 22 '23 at 05:14
6

Use gh release download to easily script downloads. gh auth login first to authorise.

So to download the examples URL https://github.com/company/repository/releases/download/TAG-NAME/A.tgz use:

gh release download --repo company/repository TAG-NAME -p 'A.tgz'

Dwayne
  • 845
  • 7
  • 11
4

An easier working solution is to use .netrc to store credentials. This way, curl does not forward credentials to Amazon S3 Bucket

In ~/.netrc file (should be created with 0600 file permission):

machine api.github.com
login yourusername
password yourpassword

Then use the curl -n option to use .netrc:

curl -L -O -J -n -H "Accept:application/octet-stream" https://api.github.com/repos/:owner/:repo/releases/assets/:id
FF_Dev
  • 462
  • 5
  • 17
  • Too bad this option is a bit more hassle when you run on dynamic hosts/configs that don't have the .netrc file by default, like newly spawned VMs/environments for which you can't define a pre-configured image. Like CircleCI for example, I think. – David Sep 22 '15 at 19:19
  • @David Well, you can also write the 3 lines to the .netrc file programatically in the head of your script, it remains the easiest solution. – FF_Dev Nov 20 '15 at 15:09
  • It also works on windows, but the netrc file may be named ~/_netrc instead of ~/.netrc – FF_Dev Nov 20 '15 at 15:10
  • As a word of warning: it is not recommended to put the token into a `.netrc` file on a server. A potential hacker can leverage this to get full access to all the user's private repositories on GitHub. – pfrenssen May 01 '23 at 08:25
4

Here is a "one-liner" using wget for making HTTP requests and python for JSON parsing:

(export AUTH_TOKEN=<oauth-token>; \
 export ASSET_ID=$(wget -O - https://api.github.com/repos/<owner>/<repo>/releases/tags/<tag>?access_token=$AUTH_TOKEN | python -c 'import sys, json; print json.load(sys.stdin)["assets"][0]["id"]'); \
 wget --header='Accept:application/octet-stream' -O <download-name> https://api.github.com/repos/<owner>/<repo>/releases/assets/$ASSET_ID?access_token=$AUTH_TOKEN)

To use it, just replace <oauth-token>, <owner>, <repo>, <tag> and <download-name> with appropriate values.

Explanation:

  • The first statement (export AUTH_TOKEN=<oauth-token>) sets GitHub OAuth token which is used by the subsequent wget commands.
  • The second statement has two parts:
    1. The wget -O - https://api.github.com/repos/<owner>/<repo>/releases/tags/<tag>?access_token=$AUTH_TOKEN part gets GitHub release information from a tag name and prints it on stdout.
    2. The python -c 'import sys, json; print json.load(sys.stdin)["assets"][0]["id"]' part parses JSON from stdin and extracts the id of the (first) release asset.
  • The third statement (wget --header='Accept:application/octet-stream' -O <tarball-name>.tar.gz https://api.github.com/repos/<owner>/<repo>/releases/assets/$ASSET_ID?access_token=$AUTH_TOKEN)) gets a single GitHub release asset by id and stores it to a file.
  • The outer parentheses create a subshell and ensure exported environment variables are discarded afterwards.
tjanez
  • 2,175
  • 1
  • 21
  • 14
2

Like @dwayne I solved this using the gh CLI which you would also need to to install. As I was inside a docker container I needed to install curl, dirmngr, then the gh CLI and then download the release like below. Also shows how to authenticate using a personal token to the gh cli, and unzip the release files (a tarball in my case).

FROM debian:10.9 as base
RUN apt update \
 # Install package dependencies
 && apt install -y \
    build-essential \
    wget \
    dirmngr \
    curl

# Install GH CLI - see https://github.com/cli/cli/blob/trunk/docs/install_linux.md
RUN curl -fsSL https://cli.github.com/packages/githubcli-archive-keyring.gpg | gpg --dearmor -o /usr/share/keyrings/githubcli-archive-keyring.gpg && \
   echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" | tee /etc/apt/sources.list.d/github-cli.list > /dev/null && \
   apt update && \
   apt install -y gh

# auth to github with PAT and download release
ARG RELEASE_TOKEN
RUN echo $RELEASE_TOKEN | gh auth login --with-token && \
   cd /opt && \
   gh release download YOUR_TAG_HERE --repo https://github.com/SOME_ORG/SOME-REPO && \
   tar -xjf SOME-RELEASE-FILES.tar.bz2 -C /opt && \
   rm SOME-RELEASE-FILES.tar.bz2
alastairtree
  • 3,960
  • 32
  • 49