# Changelog of Telegram Channel Scraper: Monitor Posts & Keyword Search (`fetchwright/telegram-channel-scraper`) Actor

- **URL**: https://apify.com/fetchwright/telegram-channel-scraper/changelog.md
- **Full Actor documentation**: https://apify.com/fetchwright/telegram-channel-scraper.md

## Changelog

### 2026-11-16

- README: a short note and Apify's video on running it on a schedule.

### 2026-11-15

- The n8n workflow files are updated: they escape Slack markup in post text, so a post cannot ping @channel or mention people; also linked on GitHub.
- README now names the Actor "Telegram Channel Scraper", matching its Store title (monitor channels for new posts and keyword search).

### 0.1.46 - 2026-11-14

- The Actor is now named Telegram Channel Monitor & Keyword Search, and its listing and README lead with monitoring the channels you name and finding channels by keyword through a web search, with the Free-plan price and a paid-plan price. The README's first table also names the minimum-views filter, channel-change reports and the numeric channel ID, and states the condition of the only-new watch. The 0.1.44 entry is corrected: it said a too-deeply-nested result page is never taken for a verification page, but such a page can still be recognised as a verification page, for example by a CAPTCHA form, a reCAPTCHA widget, a /sorry/ form or a consent form. No change in behaviour or price.&#x20;

### 0.1.45 - 2026-11-13

- README: a "Use it in a pipeline" section with one step-by-step recipe for n8n (new posts to Slack), using Apify's own integration and your own Apify account. It says what it was checked against and that it was not run in a live account. No change to the Actor, its input or its output.

### 0.1.44 - 2026-11-13

- Reliability: very long input lists (keywords, search words, exclude words, cashtags) are de-duplicated in one pass; the lists, their order and spelling, and every input message are unchanged.
- Keyword search: a result page whose markup is nested too deeply to be read now counts as that one unreadable page ("the results could not be read"; that page is not charged; channels found on earlier result pages may still be processed and charged as usual) instead of stopping the run, and is not asked for again. A page that also carries a CAPTCHA form, a reCAPTCHA widget or a /sorry/ form is still recognised as a verification page first and stops the search; other verification checks still apply, including consent forms. (Wording corrected in 0.1.46.)

### 0.1.43 - 2026-11-13

- Keyword search: a search engine's verification, CAPTCHA or consent page is now also recognised when its form or widget is written in other valid ways (spaces around "=", single quotes or none, character references, a ">" inside a quoted attribute), so the run stops searching, as for any verification page. The new reading ignores text inside comments, scripts or another attribute's value; the earlier checks are kept unchanged beside it, so every page recognised before is still recognised.

### 0.1.42 - 2026-11-12

- README: the comparison table now shows that a public group gives a free profile row (title, description, member count). Field descriptions: media links are short-lived (fresh links still worked after an hour; links saved on 10 November had stopped working by 12 November), not "within hours".

### 0.1.41 - 2026-11-12

- New: Only the channel's own posts. Leaves out forwarded posts, which are not saved or charged and are counted in the run summary.

### 0.1.40 - 2026-11-12

- Free file details for each photo, video, voice message and sticker, read from the same page with no extra request: the file type (`fileFormat`), the size at which Telegram's web page shows single photos and videos (`previewWidth`, `previewHeight`), and for a video without a file link the reason the page states (`notOnWebReason`, such as "Media is too big"). Files keep Telegram's own size text (`sizeText`) when the page shows it.
- The README now says that Telegram's media links expire within hours and that the Actor stores no copies of media files.

### 0.1.39 - 2026-11-12

- Maintenance: page text is read with linear scans throughout, so an unusual or very long post can no longer slow a run down. Results are unchanged.

### 0.1.38 - 2026-11-12

- Fix: in Markdown post text, a line that starts with spaces or a tab is no longer shown as a code block.

### 0.1.37 - 2026-11-12

- New: Post text format. Each post's text can also be returned with its formatting (bold, italic, code, quotes, links and line breaks), as HTML (`textHtml`) or Markdown (`textMarkdown`), at no extra cost. The plain text field is unchanged.

### 0.1.36 - 2026-11-12

- New: Report channel changes. A named watch now reports, for every listed channel, subscriber growth or loss (number and percent), a new title or description with the old text, and a name that now belongs to another channel, on the free channel row, in the run summary and in the alert. No extra charge.

### 0.1.35 - 2026-11-12

- Post and channel rows now carry the channel's numeric Telegram ID (channelId), which stays the same when a channel changes its username.

### 0.1.34 - 2026-11-12

- New Minimum views filter: posts with fewer views are left out, counted in the run summary and not charged.

### 0.1.33 - 2026-11-12

- Keyword search can keep only channels and public groups inside a subscriber range (Minimum and Maximum subscribers); the others give no rows, cost nothing and are listed per keyword in the run summary.
- Channel rows now show when the channel last posted (`lastPostAt`), at no extra cost.
- A public group, listed or found by keyword, now gives one free group row with its title, description and member count; its members and messages are never read.
- Channel rows of channels found by keyword say which keyword found them.

### 0.1.32 - 2026-11-12

- Every post row now carries its channel's subscriber count (`subscriberCount`), exact when the run has read it for that channel and otherwise the rounded figure Telegram shows, with `subscriberCountApproximate` saying which. No change to prices.

### 0.1.31 - 2026-11-12

- If Apify restarts a run or moves it to another server, the restarted run reads nothing more from Telegram, saves and charges nothing more, and ends with a partial status that says how many rows were saved; nothing is saved or charged twice and the maximum posts per channel holds. Start the run again for what is missing.
- A row whose save could not be confirmed is never written a second time and is charged only if the dataset shows it was saved. Otherwise the run stops there (`uncertain_write` in the run summary) and writes no new row; in new-posts-only mode the next run returns that post, so none is skipped.
- A charge that reports an error after its row was saved never makes a post be returned again: the row counts as saved, the charge is listed under `chargeErrors`, and the run starts no new work.
- A first new-posts run that stops early (maximum charge, timeout, or one of the stops above) no longer loses the posts it did not reach: the next run returns them first, up to the first run's maximum.

### 0.1.30 - 2026-11-12

- README: the "only new posts" sentences now state their condition and exceptions (one run of a watch at a time, settings unchanged; a run killed before it saved its memory can return a post again; changing search text or keywords, content types, cashtags, excluded words or end date starts a new watch), and "you pay only for posts" now names the $0.00005 start fee.

### 0.1.29 - 2026-11-12

- README and input help now say plainly what new-posts-only mode guarantees: with one run of a watch at a time, the posts it has recorded as returned are skipped and not charged; two overlapping runs of the same watch, or a run killed before it saved its memory, can return a post again.

### 0.1.28 - 2026-11-12

- Views per subscriber (`viewsPerSubscriberPercent`) uses the exact subscriber count when the run has read it for that channel, otherwise the rounded figure shown with the posts (1.05K). The channel row's `subscribers` is exact where Telegram's public page shows the full figure, and a rounded figure never replaces an exact one. No extra page reads and no change to prices.

### 0.1.27 - 2026-11-12

- Wording only: the keyword search texts now say exactly what a run does (matching posts are read for the first channels found per keyword, 10 by default; the period filter is the search engine's own).

### 0.1.26 - 2026-11-12

- New: find Telegram channels and posts by keyword. Add words or phrases under **Find channels by keyword** (`findByKeyword`), with or without channels, and each keyword returns the public Telegram channels and posts that web search has found for it, one row each with title, snippet, link, rank and the date shown, from the country, language and period you choose. A page is returned only after its channel is checked and found to be a public channel. By default the Actor then reads the matching posts of up to 10 channels per keyword, as ordinary post rows marked with `foundByKeyword`; a found channel's matching posts are read once per run.
- Found pages are free for now; posts read from the channels found are charged as posts at the normal price. Keywords that find nothing or cannot be searched are listed and never charged. The run summary reports, per keyword, the result pages read, results, channels found and read, posts saved and what stopped it.
- If Telegram asks a run for a verification, it is not answered: the run stops reading Telegram, keeps everything saved, and lists what it did not finish in the run summary: listed channels under `unfinishedChannels`, channels found by keyword under that keyword's `channelsNotRead`.
- **Maximum posts per channel** now holds for the whole run, also when a run is restarted by the platform and new posts have appeared in the meantime.
- The input field `keywords` keeps its meaning when channels are given (words to search inside them). Given with no channel and no channel dataset, it is read as **Find channels by keyword**, and the log says so.

### 0.1.25 - 2026-11-11

- When Telegram limits how fast a run may read, the run now slows down for all channels for the rest of the run and carries on. Channels it still cannot read are reported with the reason and not charged.

### 0.1.24 - 2026-11-11

- Two ready-made n8n workflows are linked from the README: new Telegram posts to Slack on a schedule, one with the Apify node and one with n8n's built-in nodes only.

### 0.1.23 - 2026-11-11

- Input written for other Telegram channel actors now works as it is: field names such as `maxMessages`, `maxMessagesPerChannel`, `maxPosts`, `maxItems`, `telegram_url`, `oldestMessageDate`, `newestMessageDate`, `lastNDays`, `keywords`, `search`, `incremental` and `sinceLastRun` are read as this Actor's own, and the log says how. A field given under this Actor's own name wins.

### 0.1.22 - 2026-11-11

- The run's storage now describes its records (the run summary), so they show as named tabs in the run's Storage view, and the main input's help says it is required unless a dataset is used. No change to results or prices.

### 0.1.21 - 2026-11-11

- The listing now leads with what sets the Actor apart for scheduled use: no Telegram account or session is involved, so there is nothing to expire. No change to results or prices.

### 0.1.20 - 2026-11-11

- Listing text and image updated. No change to results or prices.

### 0.1.19 - 2026-11-11

- An alert that got no answer in time is no longer sent again, so a slow receiver cannot produce the same alert twice. The run's status message now says when an alert was not delivered.

### 0.1.18 - 2026-11-11

- The run summary now says for every channel how much of what you asked for was read (`coverage`: complete, source_search, capped, scan_limit, partial or unavailable), so an empty or short result is not mistaken for "nothing was posted".
- An alert that fails to deliver is tried three times, and the run summary says whether it was sent (`notification.deliveryStatus`) and after how many tries. The posts are saved either way.
- Paid stars are reported on their own (`paidStars`) next to `reactionsWithoutPaidStars`, so channels can be compared on ordinary reactions.

### 0.1.17 - 2026-11-11

- A channel pasted with quotes, brackets, a list marker or closing punctuation ("durov", - @durov, https://t.me/durov.) is now read as that channel. Before, the entry was reported as not a valid channel name.

### 0.1.16 - 2026-11-10

- Several channels pasted on one line ("durov, telegram" or "@durov @telegram") are now read as separate channels.
- More ways of writing a channel are understood: the app link (tg://resolve?domain=durov), the short form durov.t.me and a web.telegram.org address.

### 0.1.15 - 2026-11-10

- A long run now reports its progress every few seconds in the status message and the log: the channel being read and how many posts are saved so far.

### 0.1.12 - 2026-11-10

- New: alerts. Paste a Slack or Discord incoming-webhook address (or any HTTPS address) and each run posts a short message: posts per channel, the first three with the start of their text and their link, and a link to all rows. With new-posts mode and a keyword or cashtag filter this is a keyword alert. A quiet run sends nothing unless you ask for it.

### 0.1.11 - 2026-11-10

- Documentation: the legal answer now points to Telegram's content licensing terms, including their limits on AI training use.

### 0.1.10 - 2026-11-10

- New option **Only posts containing any of**: up to 20 keywords in one run. Results come back newest first in one list; a post that matches several keywords is returned and charged once, and the new `matchedKeywords` field lists them.
- New option **Only posts mentioning these cashtags**: keep only posts about the tickers you follow, such as BTC or NVDA. Other posts are skipped, never charged, and counted in the `SUMMARY` record.
- Every run's `SUMMARY` record now has `cashtagCounts` per channel: each ticker and how many saved posts mention it.
- Round video messages are recognised, with their own content type, thumbnail, duration and video link.
- Channels opened through a second username now return their posts.

### 0.1.9 - 2026-11-10

- New-posts memory is now saved strictly in order, so an interrupted run can no longer leave an older memory behind and return a few posts twice later.
- Short run timeouts: the time kept free at the end of a run now scales with the run's length. A run whose timeout is too short to read even one page stops at once with a clear message instead of returning nothing.
- Cashtags at the end of a sentence (`$ETH.`) are now found; environment variables such as `$PATH` and text inside code blocks are no longer taken for cashtags.
- Changing a watch's search text, content types, excluded words or end date now shows a clear notice in the run log, because it starts a new watch whose first run returns only the newest posts.
- After a run is moved to another server, posts Telegram does not show on the web are no longer counted as saved posts in the summary.
- Clearer documentation of engagement rates (views per subscriber can exceed 100% for widely forwarded posts) and of how named monitors are shared between tasks.

### 0.1.8 - 2026-11-10

- New images in the documentation.
- New option **Posts older than**: set the end of a date window, alone or with **Posts newer than** or **Only the last N days**, for example October only. Posts outside the window are never charged, and the run starts near your end date instead of reading every newer post first.
- New option **Exclude posts containing**: skip posts that mention any word or phrase you list, such as a giveaway or an ad. Excluded posts are never charged and are counted in the `SUMMARY` record.
- Flat columns for spreadsheets on every post: `mediaCount`, `firstPhotoUrl`, `photoUrls`, `videoUrls` and `documentUrls`. The Overview table now shows the first photo of each post.
- Clearer messages for a date that cannot be read, an end date before the start date, and an empty exclude keyword.

### 0.1.7 - 2026-11-10

- Documentation rewritten for clarity: who it is for, why choose it, use cases by role.

### 0.1.6 - 2026-11-10

- The final status message is now always stored, also for runs of a few seconds.

### 0.1.4 - 2026-11-10

- New fields on every post: `cashtags` (such as BTC for $BTC), `reactionsTotal`, `reactionsPer1000Views`, `viewsPerSubscriberPercent` and `viewsApproximate`.
- Posts Telegram does not show on the web are no longer charged: they arrive as free rows marked `notViewableOnWeb`, and the Text-only filter leaves them out. Channel events such as "Channel created" are skipped.
- Stickers have their image link and their own content type.
- New-posts-only memory is now kept per saved task and per content filter and search text, so two watches on one channel never use up each other's posts. Watches from earlier versions without a content filter carry on.
- A run moved to another server by Apify is built to continue where it was, without saving or charging a post twice (tested; not yet observed on the platform); new-posts memory is saved as the run goes.
- Runs are built to finish their summary before the run timeout, even when Telegram stops answering (tested; not yet observed on the platform).
- A scheduled run with no new posts no longer repeats the channel row.

### 0.1.3 - 2026-11-10

- New-posts-only mode no longer loses posts when a channel has more new posts than one run returns: the next run returns the rest first, then anything newer. Each post is returned once. Monitors from earlier versions keep working.
- Several channels are read at once (new input **Channels read at once**, default 5, up to 10). Each channel's posts stay newest first, and your maximum charge is still never exceeded.
- When your maximum charge is reached, the post that reached it is delivered without charge and named in the `SUMMARY` record.
- If a later page of a channel fails, the posts already saved are kept and the run reports where it stopped.

### 0.1.0 - 2026-11-10

- First release: posts from public Telegram channels with text, date, views, photos, videos, files and audio, polls, link cards, reactions, forwards, replies, edited flag and author signature, newest first.
- One free row per channel with title, description, exact subscriber count where Telegram shows it, content counts, verified badge and avatar.
- Search inside channels, date limits (newer than a date, last N days), content-type filter, and a new-posts-only mode with named monitors.
- Channels that are missing, private or hidden from the web return a free diagnostic row; the run still succeeds.
- Clean stop at your maximum charge or before the run timeout, with unfinished channels listed in the `SUMMARY` record.
- Channels can be read from another run's dataset.
