Social networks are excellent places to discover useful information and terrible places to find it again.
You may remember that someone explained a subtle PostgreSQL problem, linked to a useful security report, or posted the perfect answer to a programming question. You may even remember which community it came from. But after a few days of new posts, notifications, and open tabs, the original discussion has disappeared into the timeline.
Bookmarks help when you remember to create one. Platform search helps when it supports the words and time range you need. Neither gives you one place to search across everything you have read.
Hister takes a different approach. It indexes the content that reaches your browser or crawler and stores it in a search engine you control. With dedicated extractors for Bluesky, Mastodon, Reddit, X, and Discourse, it can understand the difference between a timeline, an individual post, and a complete discussion.
The result is a private, searchable memory of the social web you actually saw.
Why a Social Feed Is Not a Normal Web Page
A traditional article is usually one page with one title and one body. A social feed is a container that may hold dozens of unrelated posts. New items appear as you scroll, old items disappear, and the page changes without navigating to a new URL.
If a search engine treats that feed as one large document, several problems appear. A result points back to the feed rather than to the post. Text from different authors is mixed together. The next update can replace the content that was indexed previously. A search result may tell you that a phrase existed somewhere in a timeline without identifying who wrote it.
Hister extractors sit between page capture and indexing. They recognize the structure of supported sites and turn the page into documents that match the way people think about its content.
Bluesky, Mastodon, and X feeds are split into individual posts. Each post gets its own URL and searchable text. Reddit and Discourse use a different model: the complete discussion is kept together so the original post and its replies remain useful as a unit.
This distinction is small, but it changes search from “these words appeared on a feed” into “this person wrote this post in this conversation.”
Bluesky Profiles, Feeds, and Threads
The Bluesky extractor handles profile pages, custom feeds, and post threads.
Every available post becomes a separate Hister document with its canonical bsky.app URL, author, handle, publication time, and a type: bluesky metadata value.
Bluesky publishes structured information for profile and individual post pages,
so Hister can often understand those pages with its normal HTTP crawler. Custom
feeds are rendered inside the browser and require either the chromedp or bidi crawler backend. The browser extension can
also submit the rendered page you are already viewing.
When structured data and rendered content describe the same post, Hister combines them using the canonical post URL. Rendered text is preferred when it contains richer content, while missing author or publication details can still come from the structured data.
To search only Bluesky posts, use:
metadata.type:blueskyAdd remembered words just as you would in any other Hister search:
metadata.type:bluesky "database migration"Mastodon Across the Fediverse
Mastodon creates an extra challenge because there is no single Mastodon domain. The same timeline structure appears across thousands of independently operated instances.
The Mastodon extractor detects the software from information in the page rather
than relying on a fixed hostname. It splits every toot found on the page into an
individual document and marks it with type: toot metadata.
Federation also means that a toot from one server may appear through another server. When possible, Hister rewrites that federated link to the original account server. Search results then point to the canonical source instead of an instance that happened to display a copy.
The browser extension watches for page changes, so newly loaded toots can be submitted while you use your normal timeline. You do not have to remember to save each interesting post manually.
Search Mastodon content with:
metadata.type:tootFor a detailed walkthrough, see Never Lose a Toot Again.
Tweets Without the Timeline Noise
The X and Twitter extractor follows the same document model as Bluesky and
Mastodon. It recognizes feeds, profiles, and individual tweet pages, then
creates one document for every visible tweet. Each document can include its
canonical X URL, author, handle, and publication time, together with type: tweet metadata.
Links inside tweets require special care. The visible page may use a t.co address even when the browser already knows the original destination. Hister
uses the expanded destination when it is available. If a link cannot be
resolved safely, Hister keeps it as plain text instead of inventing a target.
This makes the indexed text more useful. Searching for a domain or product name can find the original link rather than only finding Twitter’s redirect service.
Search tweets with:
metadata.type:tweetReddit Threads as Complete Documents
A Reddit post is usually valuable because of the conversation below it. The best explanation, correction, or link may be several replies deep. Splitting every comment into an isolated result would lose that context.
The Reddit extractor therefore keeps the post and all comments already present on the page in one document. It preserves reply depth in both indexed text and the preview. Metadata can include the author, subreddit, publication time, score, flair, post identifier, and number of extracted comments.
Only individual Reddit post pages are matched. Listings, profiles, and search pages are ignored because they are temporary collections rather than complete discussions. Hister does not request hidden comments or expand content that was not present in the submitted page.
Search indexed Reddit discussions with:
metadata.type:redditYou can combine that filter with the words you remember:
metadata.type:reddit "garbage collector" latencyDiscourse Topics With Their Solutions Intact
Discourse powers support forums and community sites across a wide range of projects. These topics often contain information that never reaches the formal documentation: migration fixes, configuration details, design explanations, and solutions confirmed by maintainers.
The Discourse extractor keeps each topic as one searchable document. It combines the original post with the regular replies already available on the page, while preserving post numbers, authors, timestamps, reply targets, likes, reactions, categories, tags, and accepted solutions when that information is available.
System actions and hidden posts are excluded. Category lists, tag pages, profiles, and search results are also ignored. The extractor focuses on the topic page because that is the stable unit someone will want to retrieve later.
Search Discourse topics with:
metadata.type:discourseOne Search Across Every Community
The real value appears when you stop remembering which platform contained the answer.
You can search every supported social document type with one query:
metadata.type:(bluesky|tweet|toot|reddit|discourse)Add a subject to search across all five sources:
metadata.type:(bluesky|tweet|toot|reddit|discourse) sqlite backupAdd a date filter and sorting when you remember seeing it recently:
metadata.type:(bluesky|tweet|toot|reddit|discourse) added:<7d sort:dateIf that expression is too long for daily use, create an alias such as !social in the Rules page and map it to the metadata expression. Your query then becomes:
!social sqlite backupNormal phrase matching, exclusions, labels, dates, and semantic search continue to work. Semantic search is particularly helpful when you remember the idea of a post but none of its exact vocabulary.
Start Building Your Searchable Social Memory
Install Hister, follow the quickstart guide, and connect the browser extension. Then use Bluesky, Mastodon, Reddit, X, and Discourse as you normally would. Supported pages are recognized automatically by the extractor chain.
You can inspect every available extractor and its exact matching behavior in the extractor documentation. If there is another community site you rely on, extractor contributions are welcome in the Hister repository.