Repository navigation
Description
Tracking issue for the v0.10.1 milestone. It is a patch release for the things v0.10.0 shipped with, plus the one item v0.10.0 could not measure.
Goal
Close the gap between what a page of documentation promises and what the binary does, and land the CDX filter push once the index server is reachable enough to measure it.
Checklist
Documentation that described something that did not exist:
- content-signals guide documents a content quality command that does not exist #110 content-signals guide documents a content quality command that does not exist, fixed by building the streaming form in Make the content commands do what the guide said they did #118
- The dataset card points readers at open-markdown-v2, but export publishes v3 #88 the dataset card points readers at open-markdown-v2 while export publishes v3, fixed in Name the repo the card is committed to #119
Messages and checks:
- Error messages name flags that do not exist, because the renderer title-cases the first word #121 error messages name flags that do not exist, because the renderer title-cases the first word
- docs-drift has been failing on main since #118 #122 docs-drift has been failing on main since Make the content commands do what the guide said they did #118
- parse wet and wat write Go field names in JSON #125 parse wet and wat write Go field names in JSON, so the keys a script reads are not the names anything else uses
- Flag help prints words nobody wrote, for the same reason the errors did #127 flag help prints words nobody wrote, the same renderer as Error messages name flags that do not exist, because the renderer title-cases the first word #121 on the other surface
- rank host and rank domain demand a URL the program can work out #129 rank host and rank domain demand a --table that host get finds on its own
- A network failure exits 1, and the help says it exits 8 #131 a network failure exits 1, and the help says it exits 8
- crawl seed --max-tier 1 reads the whole rank table and emits nothing #135 crawl seed --max-tier 1 reads the whole rank table and emits nothing, since a seed can never reach tier 1
- A cached crawl list is thrown away the moment the index server is down #137 a cached crawl list is thrown away the moment the index server is down, which took out every command that only needed the word latest
- crawls info ignores -o and prints text whatever you ask for #140 crawls info ignores -o and prints text whatever you ask for, and counted a different set of kinds than stats
- --data-dir does not move the cache, though the reference says it does #138 --data-dir does not move the cache dir, so a run aimed at a fresh tree reads the old tree's cache
- host top and host get report zero for numbers nothing measured #144 host top and host get report zero for numbers nothing measured
- A read that fails partway through a response exits 1, not 8 #146 a read that fails partway through a response exits 1, so a supervisor stops on a failure worth retrying
- An unknown -o value is accepted and silently writes JSONL #133 an unknown -o value is accepted and silently writes JSON Lines, so a typo produces the wrong bytes at exit 0
- ccrawl version ignores -o and prints one prose line #150 ccrawl version ignores -o and prints one prose line, so a CI job needs a regular expression to read the commit
Carried over from v0.10.0, and carried on to v0.10.2:
- E16: push search filters to the CDX server #60 E16, push search filters to the CDX server, moved to v0.10.2. PR Send the URL substring filters to the index server #112 is written and green and stays open; both done-when boxes are byte counts against
index.commoncrawl.org, which has refused connections for 22 days. The reason is written down on E16: push search filters to the CDX server #60.
Ship gate for v0.10.1
- every command the docs describe is a command the binary has
- the dataset card names the dataset it is committed to
- every flag an error message names is a flag that exists
- the JSON a command writes uses one set of field names
- every word the help prints is a word someone wrote
- a rank lookup answers with no flags, the way the host commands do
- a script can tell an unreachable Common Crawl from a command that is wrong
- a command refuses a request it cannot satisfy instead of working for half an hour and returning nothing
- an index server outage does not take out the commands that read from data.commoncrawl.org
- every read command honours -o, and two names for the same question give the same answer
- --data-dir moves the whole tree, and config show names what followed it
- a number a command did not measure is absent from its output rather than zero
- a connection lost mid-response exits 8, and a missing local file still does not
- a format nobody can encode is refused rather than rendered as something else
- every command that has more than one fact to report can report them as data
- ci is green on main
- E16: push search filters to the CDX server #60 measured against the live index, or moved out with the reason written down
Open question for the maintainer
main has no branch protection, so gh pr merge --auto merges as soon as GitHub considers a PR mergeable, whether or not ci passed. That is how #118 merged with docs-drift red. Making the ci checks required is a repository setting rather than a code change, so it is a call for you.
Process
Every PR that lands against this milestone gets a comment on this issue with the PR number and one line on what changed, and the matching checklist box is ticked when the linked issue closes. Automated by .github/workflows/milestone-tracker.yml. #110 and #88 merged before this issue existed, so their boxes are ticked by hand here.
Activity
tamnd commentedon Aug 12, 2026
Merged #118, Make the content commands do what the guide said they did.
Closes:
The guide promised
content quality -reading URLs from stdin, six output fields, a--warcflag and a--langflag, none of which existed. All four content commands now take-and read a URL per line, a JSON line with aurlfield works too sosearchandcrawl fetchoutput pipes straight in, a URL that fails is named and skipped, and the guide was rewritten from what the commands actually emit. Two more real doc errors turned up on the way: the CLI reference claimedcontent outlinkscarries anchor text, and the requirements page claimedcrawl fetchreads robots.txt when it only does with--robots.tamnd commentedon Aug 12, 2026
Merged #119, Name the repo the card is committed to.
Closes:
The card is now built from the repo the run is pushing to, so a run publishing to somebody else's dataset gets snippets pointing at theirs instead of ours, and
local_dirfollows the repo name. A card rendered outside a run falls back toccrawl.DefaultMarkdownRepo, which is also the export flag default now, so the two cannot drift apart again.github-actions commentedon Aug 12, 2026
Merged #123, Check the requirements page against the crawl commands too.
Closes:
github-actions commentedon Aug 12, 2026
Merged #124, Stop printing flags that do not exist.
Closes:
github-actions commentedon Aug 12, 2026
Merged #126, Give the WAT and WET records their JSON names.
Closes:
github-actions commentedon Aug 12, 2026
Merged #128, Stop printing help nobody wrote.
Closes:
github-actions commentedon Aug 12, 2026
Merged #130, rank: find the table instead of demanding one.
Closes:
github-actions commentedon Aug 12, 2026
Merged #132, exit 8 when Common Crawl cannot be reached.
Closes:
github-actions commentedon Aug 12, 2026
Merged #136, Refuse crawl seed --max-tier 1 instead of scanning for nothing.
Closes:
github-actions commentedon Aug 12, 2026
Merged #139, Keep working from a cached crawl list when the index server is down.
Closes:
github-actions commentedon Aug 12, 2026
Merged #142, Make crawls info an operation so -o works.
Closes:
github-actions commentedon Aug 12, 2026
Merged #143, Make --data-dir move the cache with it.
Closes:
github-actions commentedon Aug 12, 2026
Merged #145, Leave out host counts nothing measured.
Closes:
github-actions commentedon Aug 12, 2026
Merged #147, Give a read that fails partway through a response exit 8.
Closes:
github-actions commentedon Aug 12, 2026
Merged #148, Refuse an output format nothing can encode.
Closes:
github-actions commentedon Aug 13, 2026
Merged #151, Let ccrawl version render as data.
Closes:
github-actions commentedon Aug 13, 2026
Merged #134, docs: release notes for v0.10.1.
tamnd commentedon Aug 13, 2026
v0.10.1 is out: https://github.com/tamnd/ccrawl-cli/releases/tag/v0.10.1
Sixteen issues closed and seventeen pull requests merged. Every box above is ticked. The release notes are at docs/content/release-notes/v0-10-1.md.
One thing this milestone set out to do is not done. #60 wanted a byte count for the pushed CDX filters, and index.commoncrawl.org has been refusing connections for twenty two days, so the measurement cannot be taken. The change itself is written and green in PR #112, but the whole claim it makes is a number, and there is no honest way to publish a number nobody took. #60 and #112 have moved to v0.10.2 and the release notes say plainly that the promise made in v0.10.0 is not kept yet and why.
Checked against the published darwin arm64 binary rather than a local build:
Closing this one. v0.10.2 is open and has #60 in it.