v8.1.0

Changes

Backfilling now fetches more than one page

FediFetcher used to take whatever a single request returned when backfilling an account (usually 40 posts). The new max-posts-per-account option lets you ask for more:

python find_posts.py --max-posts-per-account=200

It defaults to 40, so nothing changes until you raise it.

Note

Higher values mean more requests to the remote server and to your own, and with backfill-with-context on, every extra post brings its replies too. Go up in steps and watch your run times. If you set this high, consider disabling backfill-with-context (Mastodon now auto-fetches context anyway) or exempt FediFetcher from your instance's rate limits.

Container images now carry more tags

stable tracks the latest stable release and is the recommended choice. latest continues to track the newest release of any kind, including RCs and betas. There are also floating major and minor tags alongside the fixed version tags — see Running FediFetcher from a container. Existing setups need no changes.

FediFetcher is now a proper Python package

pip install git+https://github.com/nanos/FediFetcher@v8.1.0
fedifetcher --server=mstdn.thms.uk --access-token=…

fedifetcher takes the same options as python find_posts.py, which still works and is what the container image, GitHub Action and wiki instructions all use. Installed runs default their state files (seen_urls, known_followings) to $XDG_STATE_HOME/fedifetcher rather than artifacts/, since the working directory is arbitrary. This should not affect existing installs, and you can continue to use --state-dir to define your own location.

Note

This mainly helps people packaging FediFetcher for platforms like NixOS. Most users are better served by the existing methods of running it.

Fixes

• PeerTube channels are backfilled correctly. A channel and its owning account are separate things in PeerTube. FediFetcher treated every PeerTube URL as an account, so backfilling a channel fetched the owner's videos instead. Both /video-channels/name and the short /c/name are now recognised.
• Profile URLs are read by the server that serves them. FediFetcher used to infer server flavour from URL shape, which breaks when two flavours share one — /c/name could be a Lemmy community or a PeerTube channel. It now asks the server what it is first. This is what made the PeerTube fix possible, and should mean fewer accounts silently skipped on non-Mastodon servers.

Full Changelog

• Type everything, and give the API payloads a shape by @nanos in #228
• Fix Peertube support by @nanos in #229
• Ask the server what it is before reading its URLs by @nanos in #230
• Backfill more than one page of an account's posts by @nanos in #232
• Add a proper python package by @nanos in #231
• keep a boost, so we can fetch the context of what it boosts by @nanos in #233
• Tag container images with stable, and with major and minor versions by @nanos in #234

Update instructions

For full instruction, please see the Updating FediFetcher documentation