A community effort to collect a curated set of analysis pipelines built using Nextflow.

nf-core

GITHUB GHA :fire: :elmofire: :github_on_fire:
&gt; java.io.IOException: Server returned HTTP response code: 504 for URL

ok pulling it, don't want to make people feel uncomfortable

don't think it was that nsfw, just very funny

USING THE #RANTS CHANNEL TO EXPLAIN MY BROWSING HISTORY POST-HOC :wink:

Boring names is such a good policy as it helps out to just focus on a minimal set of things

NEXTFLOW IMPORTS BEING RELATIVE TO THE SCRIPT NOT PROJECT ROOT :shakefist::shakefist::shakefist:

Not with modules from the Nextflow Registry :slightly_smiling_face:

Context: moving a subworkflow from `<http://subworkflow.nf|subworkflow.nf>` to `subworkflow/main.nf` breaks all imports

<https://docs.seqera.io/nextflow/modules/using-modules#including-modules>

can you use `$projectDir` or something? :thinking_face:

Is migrating to registry modules on the nf-core roadmap? I don't recall seeing it mentioned anywhere

We have sort of loosely talked about it (mostly that we could potentially retire a fair bit of nf-core machinery) but not in any concrete manner yet. It'd be a good topic for the <#C069B2Z5P1R> channel!

Does using the nextflow module registry come with some kind of module versioning?

See the "Versions" tab on eg. <https://registry.nextflow.io/modules/cellgeni/anndata/concat@0.0.1>

nf-core ones are a bit weird because they come from git, so they get `0.0.0-[commit hash]`, eg <https://registry.nextflow.io/modules/nf-core/abra2@0.0.0-6c4ed3a>

SEQERA TOOLS STILL NOT WORKING BEHIND REVERSE PROXY :shakefist::shakefist::shakefist:

From what I've checked, Nextflow and Wave, although it might be a flaw inherited from the underlying Java networking. I haven't had issues with any other software in the same setup. It's annoying cause I have to open a Codespace every time I want to run nf-tests.

I'd be curious to see an issue with more details. I don't think that they should be doing anything special 

I am no networking expert, so I can't explain the setup is fine detail, but every request must be passed to the proxy along with a username and password, otherwise the proxy immediately returns 407 and the request goes nowhere. To make the credentials available to programs, they are saved in env variables HTTP_PROXY and HTTPS_PROXY. Python respects these variables, so the tools work as intended. git also works normally. However, Nextflow and Wave rely on some Java machinery that seemingly ignores those variables, and hence all requests originating from Nextflow are rejected. I would love to provide an MWE but I have no idea on how to reproduce the network setup.

Investigated a bit and made an issue. Some PRs incoming.. <https://github.com/nextflow-io/nextflow/issues/7305|https://github.com/nextflow-io/nextflow/issues/7305> 

GITHUB :fire: :elmofire: :github_on_fire: :sarek-on-fire: :aws-on-fire: :aws-combust: :this-is-fine-fire:

Screenshot 2026-07-01 at 13.02.17.png

I haven't been using copilot too much actually

GITHUB :elmofire: :github_on_fire: :this-is-fine-fire:

and deprive me of the pleasure of ranting

STAGING TEST DATA FROM GITHUB :shakefist: :keyboard-slam:

`Can't stage file <https://raw.githubusercontent.com/nf-core/test-datasets/modules/data/genomics/homo_sapiens/genome/genome.fasta> -- reason: Operation timed out`

Screw it, using a local copy of `test-datasets`

Image from iOS.gif

Have fun waiting for the repo clone to finish 

I had one lying around fortunately :smile:

EXAMPLES FROM NEXTFLOW DOCS NOT LINTING CLEANLY:shakefist: HOW IS THIS EVEN POSSIBLE :shakefist::shakefist::shakefist:

<https://docs.seqera.io/nextflow/workflow#takes-and-emits|Example in question>:
```workflow my_workflow {
    take:
    data

    main:
    ch_bye = bye(hello(data))

    emit:
    ch_bye
}

workflow {
    result = my_workflow()
    result.view()
}```
Note: you need to add dumy definitions of `hello()` and `bye()` , and pass some string to `my_workflow` to make it valid. Even then, it gives 2 lint warnings (unused variable and single named workflow, neither of which is true).

Feel free to make a PR :slightly_smiling_face:

Stale docs are very possible and very easy. Especially whilst transitioning between two totally different language parsers that have different rules, whilst trying to keep the docs valid (and useful) for both. We try our best :innocent:

Turns out the parser was wrong in this case, I believe a PR was merged today

I'd like us to look into using an LLM or some kind of automation to check all code snippets in the docs too, longer term 

We have some nice stuff for the training that <@U03SLE7RBAS> developed (where it runs all the examples and checks / updates the outputs)

I wonder if we could do something similar for the docs. Though it'll be more difficult 

PULL REQUESTS TARGETING MASTER/MAIN BY DEFAULT :shakefist::shakefist::shakefist:

I know that's how GitHub is supposed to work
BUT STILL

SEQERA "AI" MAKING UP COOL BUT FAKE NEXTFLOW FEATURES :shakefist::shakefist::shakefist:

Can you report in the seqera-ai channel on the Nextflow slack? then the team can take a look

The agents implementing those features have already been kicked off by the very same model :wink:

COWBOY MERGE :rolling_on_the_floor_laughing:

<mailto:datarama@hachyderm.io|datarama@hachyderm.io> - @eniko The leatherback sea turtle almost never rests.  In fact, its activity level is so high that despite having a typical reptilian metabolism, its muscle activity is high enough that it maintains a constant body temperature. It's estimated that it rests for less than 0.1% of the day.

This means that an actual turtle has more nines of uptime than Github

WAITING PATIENTLY FOR THE UNPAID BUT MUCH APPRECIATED WORK OF BIOCONDA MAINTAINERS :shakefist::shakefist::shakefist::shakefist:

Try pinging folks on <#CM46YC6BZ> if you just need an approval :slightly_smiling_face:

ERROR: THE OPERATION WAS CANCELED :shakefist: :shakefist: :shakefist:

MODULES BEING INCONSISTENT AS TO WHETHER THEY ARE TUMOUR THEN NORMAL OR NORMAL THEN TUMOUR FOR THEIR INPUT STRUCTURE

At least tumor/tumour doesn't impact the channel order

&gt; Normal then Tumour: 11 modules
&gt; • ascat
&gt; • biscuit/pileup
&gt; • controlfreec/freec
&gt; • deepsomatic
&gt; • manta/somatic
&gt; • msisensor/msi
&gt; • msisensorpro/msisomatic
&gt; • oncocnv
&gt; • sequenzautils/bam2seqz
&gt; • strelka/somatic
&gt; • varscan/somatic
&gt; Tumour then Normal: 9 modules
&gt; • cnvkit/batch
&gt; • lofreq/somatic
&gt; • muse/call
&gt; • parabricks/mutectcaller
&gt; • platypus (uses "control")
&gt; • savana/cna
&gt; • savana/run
&gt; • svaba
&gt; • telomerehunter (uses "control")
:neutral_face:

:shakefist: hackathon group to come up with yet another standard?

`val(meta), path(index_normal), path(input_tumour), path(input_normal), path(index_tumour)`

Screenshot 2026-08-27 at 18.02.52.png

claude just asking me the same thing :smile:

uhh, are you working on parabricks deepsomatic?

but I think tumour should come first :wink:

barely started (unless there is a PR somehwere already) , been doing a sweep on some of the parabricks stuff to clean up some of the implementation bits in sarek.

<https://github.com/nf-core/modules/pull/12797|github.com/nf-core/modules/pull/12797> I made a module :wink:

but couldn't convince the bot to make the snapshots

:woohoo: nice, should have made claude search PRs before thinking about input order

EVERYTHING TO DO WITH ACADEMIC PUBLISHING SUBMISSION, REVIEWING, PROOFING INTERFACES (OR JUST SCIENTIFIC PUBLISHING IN GENERAL) :shakefist: :shakefist: :shakefist: :shakefist: :shakefist: :shakefist:

&gt; Error connecting to GitHub
&gt; 
&gt; We are experiencing high volume and rate limits. Please try again in the next hour.

&gt; Unable to read response from the server. Please try again later.

What's happening, you seems to be the one ranting for the last 5 rants :laughing:

yeah, I have no patience for this kind of things

WEBIN-CLI :shakefist::shakefist::shakefist:

I have no idea what this is about, but I :shakefist: with you

I'm uploading 15Tb of fastq to ENA and their uploader sucks!

I join the rant! We have quite a lot of experience with the tool though, shout out if you are struggling....

Main problem is an intermittent "username/password" failure that goes away on a retry... <@U08AHEKTQQ4> says it's to do with overloading the server though I can't think I'm pestering it that much!

&gt; 15TB is impressive!
900 Illumina Oak "metagenomes" (i.e. 95% oak sequence...)

Feels like you are covered, Sonya and I are in the same team :slightly_smiling_face: I would have suggested the same!

Singularity CI fails! :shakefist:

&gt; 503  Service Unavailable [IP: 185.125.189.187 443]


STRINGS BEING CHANGED TO DATAFLOWVALUES WHEN EMITTED FROM A SUBWORKFLOW :shakefist::shakefist::shakefist::shakefist::shakefist::shakefist::shakefist::shakefist:

I thought this was always the case? When you emit a bare value, it gets converted into a channel object

It probably was but it leads to some awkward behaviour if you want to e.g. process a raw datafile in a subworkflow with Groovy

You didn't want to use a native process for it?

Needed to be included in an `if()` statement

And edge release around lunchtime I think

yep, I couldn't push at all, it has now resumed

"ERROR ~ Invalid method invocation `call` with arguments:" NOT RETURNING A LINE NUMBER :shakefist::shakefist::shakefist::shakefist:

I'm guessing you changed a tuple's entries. Catches me often.

It's happening insidiously inside a function, I think

I found it, that was very insidious. I read a bunch of stuff into `meta` with samplesheetToList, and I added a new meta entry but didn't correctly add the `meta` field to the schema file, so it created a new field in the input tuple.