I've joined the POSSE and publish as much as I can to justin.searls.co and syndicate it elsewhere. I'm already using feed2toot to cross-post to Mastodon, but for my image posts in particular, I wanted to cross-post them to Instagram, so I made this thing that reads from an Atom XML feed and generates Instagram posts. It's meant to be run on a schedule (e.g. cron job) to regularly check the feed, and does its best to avoid double-posts by keeping track of post URLs that have already been processed
First step: take a deep breath and prepare to budget half a day of frustration to this. The number of hoops you need to jump through to post to Instagram are myriad:
- Convert your Instagram account to a professional (either "creator" or "business" types work with this gem)
- Create a Facebook Page with your Facebook account and link it to your Instagram account
- Create a Facebook developer account
- Create a Facebook app and grant it these permissions (you don't actually need all of these but figuring out the exact set of them is a pain and they sure seem to change frequently, so YMMV):
instagram_basic
instagram_manage_comments
instagram_manage_insights
instagram_content_publish
business_management
pages_show_list
pages_read_engagement
pages_manage_metadata
pages_manage_posts
public_profile
- Generate an access token for yourself; the easiest way is probably the Graph API Explorer tool
- With that access token, (set here to a
FACEBOOK_ACCESS_TOKEN
env var), find the right Facebook Page ID you linked to your Instagram account. You cancurl
it like this:curl -X GET "https://graph.facebook.com/v18.0/me/accounts?access_token=$FACEBOOK_ACCESS_TOKEN"
(I recommend piping the results to jq, so install that and tack on| jq
) - With that Facebook Page ID (set here to
FACEBOOK_PAGE_ID), find your Instagram account ID. Here's a curl command:
curl -X GET "https://graph.facebook.com/v18.0/$FACEBOOK_PAGE_ID?fields=instagram_business_account&access_token=$FACEBOOK_ACCESS_TOKEN"` - Note down your access token, your Instagram account ID, your App ID, and your App Secret (the last two can be retrieved from the "App Settings" -> "Basic" from your app's dashboard)
- You're ready to read the Instagram Graph API, Getting Started doc, and the Content Publishing Guide
To get an idea of what this gem is doing under the hood, namely it will:
- Trade whatever access token you hand it for a refreshed long-lived token, and then save that updated/refreshed token to your feed2gram configuration (long-lived tokens expire after 60 days and must be refreshed before they expire or else you need to generate a new one; keep this in mind if you don't plan to run
feed2gram
continuously or if the configuration file isn't writable) - Load your Atom feed and scan it for entries that contain a
<figure>
element (only the first<figure>
will be read). See notes on formatting your feed - For each such entry, create an image container (this is when the Facebook backend downloads the image and processes it)
- If the
<figure>
contains multiple<img>
tags, create a carousel container that references all the subordinate image containers - Once the container is created, publish it to Instagram
- Success or failure, save a cache entry that indicates the URL of the entry was processed so we don't repeatedly post (or fail to post) the same thing over and over again
$ gem install feed2gram
Next, create a configuration file in YAML to tell feed2gram everything it needs to run. Make sure this file is writable, as the gem will refresh the facebook access token on each run:
feed_url: https://example.com/photos.xml
facebook_app_id: 1234
facebook_app_secret: 5678
instagram_id: 9000
access_token: EAADXD
If the above were saved as my_feed2gram.yml
, we could then run the app from
the command line:
$ feed2gram --config my_feed2gram.yml
In addition to overwriting the access_token
in your configuration
file, a my_feed2gram.cache.yml
will also be created (or updated) in the same
directory. This file is used internally by feed2gram to keep track of which
entry URLs in the atom feed have been processed and can be ignored on the next
run.
For available options, run feed2gram --help
:
$ feed2gram --help
Usage: feed2gram [options]
--config PATH Path of feed2gram YAML configuration (default: feed2gram.yml)
--cache-path PATH Path of feed2gram's cache file to track processed entries (default: feed2gram.cache.yml)
--limit POST_COUNT Max number of Instagram posts to create on this run (default: unlimited)
--skip-token-refresh Don't attempt to exchange the access token for a new long-lived access token
--populate-cache Populate the cache file with any posts found in the feed WITHOUT posting them to Instagram
feed2gram uses the first <figure>
element to generate each Instagram post. That <figure>
can contain one or more <img>
tags and one <figcaption>
tag, which will be used as the post's image(s) and caption, respectively.
Some things to keep in mind:
- A
<figure>
may specify adata-post-type
with a value ofreels
,stories
, orpost
(if unspecified, the type defaults topost
)- If
data-post-type
is set tostories
orreels
, exactly one image or video must be included. Ifpost
, then multiple (up to ten) images and videos can be included and will publish as a carousel post - Posting stories (i.e.
<figure data-post-type="stories">
) requires a business account, not a creator one (in which case a, "the user is not an Instagram Business," error will be returned)
- If
- If one
<img>
tag is present, a single photo post will be created. If there are more, a carousel post will be created - Because Facebook's servers actually download your image as opposed to receiving them as uploads via the API, every
<img>
tag'ssrc
attribute must be set to a publicly-reachable, fully-qualified URL - To post videos, stories, or reels, set the
data-media-type
attribute on the<img>
tag tovideo
orimage
(a media type ofimage
will be assumed by default if left unspecified). Note that whileimage
andvideo
media may be interspersed throughout a carousel - For carousel posts, the aspect ratio of the first image determines the aspect ratio of the rest, so be mindful of how you order the images based on how you want them to appear in the app
- Only one caption will be published, regardless of whether it's a single photo post or a carousel
- The caption limit is 2200 characters, so feed2gram will truncate it if necessary
- The API is pretty strict about media file formats, too, so you may wish to preprocess images and videos to avoid errors in processing:
- Images can't be more than 8MB and must be standard-issue JPEGs
- Videos are even stricter (best to just read the docs, including this bit on reels). Videos that appear in carousels seem to have additional no-longer-documented restrictions (in my testing, 9:16 videos routinely failed but 16:9, 1:1, 4:3, and 3:4 succeeded)
Here's an example <entry>
from my blog feed:
<entry>
<id>http://localhost:1313/shots/2023-10-17-08h04m28s/</id>
<title type="text">A tale of artificial intelligence in four acts</title>
<link href="http://localhost:1313/shots/2023-10-17-08h04m28s/" rel="alternate" type="text/html" />
<author>
<name>Justin Searls</name>
<email>website@searls.co</email>
</author>
<published>2023-10-17T12:04:28+00:00</published>
<updated>2023-10-17T12:04:28+00:00</updated>
<content type="html"><![CDATA[
<figure>
<img src="/shots/2023-10-17-08h04m08s-c913ad8.jpeg"/>
<img src="/shots/2023-10-17-08h04m08s-79dbb2d.jpeg"/>
<img src="/shots/2023-10-17-08h04m08s-8421af6.jpeg"/>
<img src="/shots/2023-10-17-08h04m08s-b172e07.jpeg"/>
<figcaption>
I was wondering if I should keep dragging my hacky little OpenAI API wrapper class from script to script, so:
1. Search [rubygems.org](https://rubygems.org) for "gpt"
2. Find one called `chat_gpt` described as "This is OpenAI's ChatGPT API wrapper for Ruby"
3. Click the "Homepage" link
4. The code repository is archived and contains the disclaimer "NOTE this code was written by ChatGPT and may not work"
Great job, everyone.
See more at http://localhost:1313/
</figcaption>
</figure>
]]></content>
</entry>
We publish a Docker image using GitHub
actions
tagged as latest
for every new commit to the main
branch, as well as with a
release tag tracking every release of the gem on
rubygems.org. The images are hosted here on GitHub's
container
registry
You can also use Docker to run this on your own automation platform like Proxmox or Kubernetes.
$ docker run --rm -it \
-v ./your_config_dir:/srv/config
ghcr.io/searls/feed2gram
To configure the container, there are just four things to know:
- A volume containing your configuration and cache files must be mounted to
/config
- By default, feed2gram will run with
--config /config/feed2gram.yml
, but you can customize this by configuring the command value as needed - By default, feed2gram is run as a daemon every 60 seconds, and that duration can be overridden
by setting a
SLEEP_TIME
environment variable to the number of seconds you'd like to wait between runs - If you'd rather run
feed2gram
as ad hoc as opposed to via the included daemon (presumably to handle scheduling it yourself), simply change the entrypoint to/srv/exe/feed2gram
I run this on my Synology DS 920+, using the DSM's Container Manager app.
There are just a few things to know to set this up:
At the time of this writing, the Action > Import > Add from URL
feature of the Container Manager's
"Image" tab does not support GitHub Container Registry URLs. However, if you connect via SSH:
$ sudo -s
# Enter your user password.
$ docker pull ghcr.io/searls/feed2gram:latest
Once downloaded, the image will appear in the app. From there, select
ghcr.io/searls/feed2gram
, hit Run, and complete the wizard, setting any custom
command line flags (once the container is created, this cannot be edited), as
well as choosing a location to mount the /config
volume and setting a
SLEEP_TIME
environment variable (these can be changed after the fact).
Look at your cache file (by default, feed2gram.cache.yml
) and you should see
all the Atom feed entry URLs that succeeded, failed, or were (by the --populate-cache
option) skipped. If you don't see the error in the log, try
removing the relevant URL from the cache and running feed2gram
again.
If you're seeing an embedded API error like this one:
The submitted image with aspect ratio ('719/194',) cannot be published. Please submit an image with a valid aspect ratio.
It means your photo is too avant garde for a mainstream normie platform like Instagram. Make sure all images' aspect ratios are between 4:5 and 1.91:1 or else the post will fail.