Datasets:
You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
by downloading this dataset, you agree with the dataset license
Log in or Sign Up to review the conditions and access this dataset content.
Bluesky scrape
hundreds of millions of public posts from almost 6 million Bluesky accounts since November 2022, read from the firehose and updated monthly.
from datasets import load_dataset
posts = load_dataset("tiagozip/bluesky", "posts", split="train", streaming=True)
profiles = load_dataset("tiagozip/bluesky", "profiles", split="train")
format
| config | rows | notes |
|---|---|---|
| posts | 300M+ | one file per month, data/posts/YYYY-MM.parquet |
| follows | 95M+ | follow and unfollow events, one file per month |
| profiles | 5.8M+ | profile snapshot, one row per account |
posts
| column | type | notes |
|---|---|---|
uri |
string | at://did/app.bsky.feed.post/rkey |
did |
string | author, joins to profiles.did |
text |
string | |
created_at |
timestamp | as claimed by the client, can be wrong |
cid |
string | |
reply_to |
string | parent uri, null when not a reply |
repost_count |
int32 | counter at fetch time, not current |
like_count |
int32 | same |
quote_count |
int32 | same |
media_urls |
list[string] | these may be deleted |
follows
follower, subject (both DIDs), rkey, op (create or delete), created_at, seen_at.
profiles
did, handle, display_name, description, avatar, banner, followers,
follows, posts, pinned_post_uri, pinned_post_cid, created_at, indexed_at, hydrated.
what this is not
not complete. the firehose was not recorded without gaps, and older months are thinner than recent ones.
engagement counts are snapshots. every counter is whatever it read when the post was stored. do not treat these as final values.
bridged accounts are absent. accounts bridged in from the fediverse are left out.
license
disclaimers
access can be revoked at any time and all data is provided as-is. this dataset contains personal information about real people, a lot of whom are not public figures. this dataset is not a reflection of the views of the author or the institutions they are affiliated with. please be respectful. every post in this dataset was publicly readable at the time it was fetched.
opting out
you may opt out at optout.tiago.zip or by adding "#noindex" to your bluesky bio.
citation
@misc{bluesky,
title = {bluesky: a corpus of public Bluesky posts},
author = {tiagozip},
year = {2026},
url = {https://huggingface.co/datasets/tiagozip/bluesky}
}
- Downloads last month
- 63