Repository navigation
Releases
v0.1.6
The GitHub organization FreshLabDev is now asterfield-dev, and the bot moves
with it. Nothing a person sees in Telegram changes.
Security
- Built with Go 1.26.11, which fixes GO-2026-6859 in
net/http.
govulncheckfinds the bot reaching it through its HTTP client and health
server.
Changed
- Images are published as
ghcr.io/asterfield-dev/makeitmd. The
organization was renamed, and GitHub Container Registry does not redirect
the old name:ghcr.io/freshlabdev/makeitmdanswers 403. Point the
deployment's image at the new path; every earlier digest is there under it
too. - tg moves from
v0.0.1-alpha.7tov0.4.0-alpha.2, whose module path is
github.com/asterfield-dev/tg. A press without its message is now nil
rather than an empty message; the panel still ignores it. - Links, the module path and the docs point at
github.com/asterfield-dev.
Old links still redirect.
Known Limitations
- Released as a stable patch without a pre-release first, at the owner's
request: the change is the organization rename and the tg upgrade, checked
by the full test suite, not by a run on a test bot. v0.1.5was tagged but never published: its vulnerability scan stopped
the release on GO-2026-6859. This version is that one plus the Go update.
Container image
ghcr.io/asterfield-dev/makeitmd@sha256:3ca02bb1d5cb4cc72dd87d2dfff0b2bbb355adfbe725e44ffc1f963a4260f0f4
v0.1.4
makeitMD renders Markdown sent as a text file, renders formulas the way
ChatGPT, Gemini and Claude actually deliver them, and is built with a Go
release that closes ten advisories v0.1.2 is exposed to.
This promotes v0.1.4-alpha.5 unchanged. The five alphas ran on the
production bot, which has no test twin. Answers from ChatGPT, Gemini and
Claude were pasted from Telegram Desktop and from phones, and formula files
were sent; the replies were checked on Desktop, Android and iOS. The failure
path was exercised live too: four pastes that arrived as rich messages got the
explanation instead of silence. The details are in the alpha sections below.
There is no v0.1.3: that tag was pushed by mistake with unreviewed work,
nothing was published from it, and tags here cannot be deleted.
Added
- Markdown sent as a UTF-8 text file of up to 128 KB is rendered like a paste.
\(...\)and\[...\], ChatGPT's delimiters, render as formulas, and so
do the doubled ones its copy button produces.- Display formulas that the ChatGPT and Gemini phone apps copy as bare LaTeX,
Gemini's inside a code block, are recovered in pastes that delimit no
formula. Inline formulas ChatGPT's app deletes while copying cannot be. - A paste that lost its headings, lists and tables is followed by advice:
paste without formatting, send a file, or on a phone send from a computer.
Fixed
- A paste that arrives as a rich message, which Telegram Premium clients send,
gets an explanation instead of no answer at all. - A long paste from a phone is no longer stitched tail first.
- Inline code restored from client formatting keeps its edge backticks and
spaces.
Security
- Go 1.26.10 and
golang.org/x/text0.41.0:govulncheckfound ten reachable
advisories onv0.1.2and finds none here.
Container image
ghcr.io/freshlabdev/makeitmd@sha256:2870c308d10901b8a387e5c8b88c10ef5fe5af6219a78aa3d3a4a316e74a823c
v0.1.4-alpha.5
Found by pasting chatbot answers from phones into alpha.4. Still on the
production bot.
Fixed
- A long paste from a phone is no longer stitched tail first. A phone
sends the parts of a split paste at once, and Telegram gave the 2637-character
tail of a chatbot answer a smaller message id than its 4067-character start,
so the reply began in the middle. When all parts arrive in the same second
and exactly one is shorter than 3000 UTF-16 units without being last, it is
the tail and moves to the end. Anything else keeps message id order, so a
short note typed before a paste stays first. - Display formulas copied on a phone render. Chatbot apps such as ChatGPT
copy a display formula as bare LaTeX, with no delimiters; one app puts it in
a code block. In a paste whose source delimits no formula, a line or a
language-less code block that reads as a formula now becomes$$...$$: a
LaTeX command next to math, or a sub/superscript with a relation, and no
word of three or more letters once commands and braced arguments are
removed. A\begin{cases}split over lines and a formula a phone broke in
two are joined. On the test answers this recovered 7 of ChatGPT's display
formulas and 5 of another app's, and changed nothing in 17 earlier pastes.
ChatGPT's app deletes inline formulas while copying; those cannot come back. - The paste advice tells phone users what to do. ⌘⇧V does not exist on a
phone, and chatbot apps there lose headings, lists and tables while copying,
so the advice now says to send from a computer.
Container image
ghcr.io/freshlabdev/makeitmd@sha256:6d59297c56891ed35d11f26c1b467fd565dfec0f000ea82fac286130f6c02ae9
v0.1.4-alpha.4
Found by pasting a Claude answer copied in the iPhone app: the bot did not
answer at all. Still on the production bot.
Fixed
- A message that arrives already formatted gets an answer. Since Bot API
10.1 a Telegram client can send a paste as a rich message: blocks, with no
text field and no Markdown source. The bot ignored such messages without a
word. It now explains that it cannot read them and suggests a .md file or a
paste without formatting. Nothing is stored, and the log records only how
many blocks of each type arrived, to find out what clients send before
deciding whether to render them.
Container image
ghcr.io/freshlabdev/makeitmd@sha256:0ef22eda2647ab8514bd373575e89398e03619e8b2b00b0fa3bedff07141a235
v0.1.4-alpha.3
Found by pasting a ChatGPT answer into alpha.2 with ⌘⇧V. Still on the
production bot.
Fixed
- ChatGPT formulas pasted as plain text render. ChatGPT's copy button
escapes its Markdown once more, so the plain text holds\\[ 1\\,000 \\]
and\\(PV\\). alpha.2 read the doubled backslash as an escaped backslash
and left every formula of the test answer as text. Doubled delimiters are
now formulas too, with Markdown escapes such as\_and\&removed from
the content. They are converted only when at least one holds LaTeX syntax,
so a loneC:\\(x86\\)stays text.
Container image
ghcr.io/freshlabdev/makeitmd@sha256:a70dd4923beb0c3e4e50aa8cd081521c22069b608e733a0aa2c6a3329baf95d8
v0.1.4-alpha.2
Found by pasting answers from three chatbots into alpha.1: ChatGPT's formulas
did not render, and an ordinary paste lost tables and lists before the bot saw
them. Still on the production bot.
Added
- Formulas written as
\(...\)and\[...\]render. That is how
ChatGPT delimits them, and Markdown reads\(as an escaped parenthesis, so
Telegram showed(7^{-\frac{2}{3}})as text instead of a formula. Outside
code they now become$...$and$$...$$; a multi-line\[...\]becomes
one line. An escaped backslash and a bracketed reference such as\[1\]keep
their meaning. In a ChatGPT answer pasted on 20 October all fifteen formulas
were affected. - Advice after a paste that lost its structure. An ordinary paste of a
chatbot's answer hands Telegram the clipboard's HTML, and the client drops
headings, lists, tables and LaTeX backslashes before the bot sees the text.
A long message that arrives with a quote or text link and no heading, table
or bullet line left now gets one more message after the reply: paste without
formatting (⌘⇧V, Ctrl+Shift+V) or send a file. How it works says the same.
Of the three chatbot answers pasted on 20 October, all three would have got
it; the same answer pasted with ⌘⇧V would not.
Container image
ghcr.io/freshlabdev/makeitmd@sha256:bbcb3dd7dc3d75e76bfa2618c6880f45e7b54aad1d044b670be6ed45ff62ce7b
v0.1.4-alpha.1
makeitMD renders Markdown sent as a text file, and is built with a Go release
that closes ten advisories the running v0.1.2 is exposed to.
makeitMD has no test bot, so this alpha runs on the production bot. It becomes
v0.1.4 once formulas from docs/examples/formulas.md sent as a file render
correctly on Telegram Desktop, Android and iOS, or is rolled back to v0.1.2.
There is no v0.1.3. That tag was pushed by mistake on 20 October with work
that had not been reviewed, nothing was published from it, and tags here
cannot be deleted.
Added
- Markdown can be sent as a file. A UTF-8 text file of up to 128 KB gets
the same native rich reply as pasted Markdown, whatever its extension:.md,
.txt,.texor none. Text is recognised by its contents, not by the name or
MIME type Telegram reports. Captions are ignored, and a file is never joined
to a neighbouring message. An empty, binary or non-UTF-8 file, a PDF and a
file over the limit each get a short explanation instead. The download has
15 seconds includinggetFile, retries a transient failure up to three
times within that budget, and goes to a private temporary directory that is
removed afterwards. Accepted files follow the existing conversion retention;
rejected ones are not stored. In groups a file is not downloaded at all. - How it works explains files, formulas (
$...$,$$...$$and fenced
math) and Telegram's limits on a rich message. Formula source is passed to
Telegram unchanged;\(...\)and\[...\]are not rewritten.
Security
- Go moves from 1.26.6 to 1.26.10 and
golang.org/x/textfrom 0.39.0 to
0.41.0. Onv0.1.2govulncheckfinds ten advisories reachable from
makeitMD's code, nine in the standard library (among them GO-2026-6611,
-6612, -6613 and -6617 innet/httpand HTTP/2) and GO-2026-6629 in
x/text; on this version it finds none.
Fixed
- Inline code restored from a client's formatting kept its edge backticks and
surrounding spaces wrong: a span starting or ending with a backtick merged
with the delimiter, and a span with a space at both ends lost them. Both now
come back as the person typed them.
Changed
- Rendering errors now mention Telegram’s rich-message limits as well as
Markdown syntax. A byte download limit is not a rendered-character limit;
documents are never silently truncated or split. - A release waits for CI.
release.ymlrunsci.ymlon the tagged commit
and publishes nothing unless it passed. When CI cannot run,release.ymlcan
be started by hand with the tag and a reason; that run skips CI and prints the
reason at the top of the release notes. Seedocs/releases.md. - The release notes end with the image digest to pin, which
docs/releases.md
already said they did. - The image carries
org.opencontainers.image.versionand.revisionlabels,
sodocker inspectnames the build without starting it. - CI also runs weekly, so a new advisory against a dependency or the standard
library is noticed while nothing is being pushed. - Dependabot opens grouped version updates against
dev.
Container image
ghcr.io/freshlabdev/makeitmd@sha256:7420a6c99ee31f171a74835ed018aca4534ab5ee63a2a8915f5b5fa55865b2d0
v0.1.2
makeitMD speaks sixteen languages.
Every word it says moved out of the Go source into a translation catalogue, the
language is read from the shared hub so a choice made in any sibling bot is
honoured here, and there is a picker for changing it. Nobody has to do anything:
a person whose Telegram is Russian gets Russian without touching a setting.
Changed
docs/releases.mdnames the real cause of the403a digest pull used to
answer: the account running the deploy could not read the private package,
not anything about the digest. Pulling the tag first was a workaround for
that, and is no longer required.
v0.1.2-alpha.1
Added
- The panel speaks the languages the family shares, and lets you pick one.
Every sentence makeitMD says now comes from one embedded catalogue instead of
a Go constant, the language is resolved per person from the shared hub with
the Telegram client's own language as the fallback, and a Language screen
offers the same sixteen options, in the same order, that the sibling bots
offer. The choice is stored once for the whole family, so making it here makes
it everywhere; Follow Telegram withdraws it and lets the client decide
again. Only English is written so far -- the other fifteen locales read in
English until they are translated, which is why they can arrive one at a time.
Telegram's command menu is published per language for the same reason: a
translated panel opened from an English menu speaks two languages at one
person.
Changed
- Every screen opens with a title. The root screen used to start with bare
prose while every other panel in the family started with a bold title and a
one-line hint, which made makeitMD read as a different product from the bot
next to it. The greeting itself is unchanged, and the About card keeps the
shape it already had.
v0.1.1
makeitMD gets a panel. The product boundary moved deliberately: /start now opens
two tabs instead of answering with a sentence and no way forward. Still no
settings.
Changed
/startnow opens a panel instead of answering with one sentence. Two
tabs and no more: How it works, and an About card carrying the version and
commit the running process reports on/healthz, the rendering it uses, the
repository, the licence and the admin.Backis the only navigation, every
tab edits the same message rather than posting another one, and there is still
nothing to configure. The greeting itself is unchanged.- Commands are published per scope instead of globally. The private list
is/start — Open the makeitMD panel; the group list is one/start
registered as ephemeral, so a person on a Bot API 10.3 client who types it in
a group gets a private link back to the direct chat and the group sees nothing
at all. An older client cannot send an ephemeral command, so the bot has no
message it is entitled to answer privately and stays silent rather than
replying in front of everyone -- which reads as a dead bot, and will be the
common case until clients catch up. The default
scope, where the old globalStart the botentry lived, is cleared. Nothing
is offered in a chat where it would do nothing. - One versioning and release document for the whole family.
docs/versioning.md
anddocs/releases.mdare now byte-identical across every Asterfield
repository apart from two clearly marked sections: this repository's own
version line, and the surface where a change here breaks something. They spell
out what each of the three numbers means, what the-alpha.Nsuffix counts,
when alpha becomes beta and when it is legitimate to skip to rc or run a
pre-release in production. - Pre-releases are now tagged on
dev, notmain. Only stable versions are
tagged onmain, on the merge commit fromdev.release.ymlhad no branch
check at all before, so a tag pushed from any branch would publish; it now
refuses a tag that is not on the branch its channel is published from.
Earlier pre-releases were tagged onmainunder the previous rule; they are
left as they are. github.com/FreshLabDev/tgmoves tov0.0.1-alpha.7. It carries one fix:
a preflight probe is marked as a probe, so the404 method not foundit
expects stops being counted and logged as a transport failure. All four bots
on the shared client now run the same version.
Added
-
docs/releases.mdgained a Deploying section, andAGENTS.mdpoints at it.
Releasing was documented; deploying was not, in any repository in the family —
the process stopped at "deploy it" and never said how. That gap mattered more
after the stacks moved from building on the host to pulling a published image,
because the procedure changed on the same day. The section names this stack's
host directory, its env file, the variable that selects the image, the networks
it needs, and what a rollback actually is. -
internal/buildholds the version, commit and build date the linker stamped
in./healthzand the About card read the same value, so the panel cannot
report a build that is not the one answering. -
deploy/ws04/compose.yaml, the production stack, pulling the image the
release workflow publishes to GHCR. The stack on the host built its own image
from a working copy, so what served users was not the artifact CI had tested,
scanned and published, and nothing on the host could say which commit it came
from.MAKEITMD_IMAGEhas no default: an unset one stops the stack instead of
quietly starting something else.