Summaries to please show links to open the source document at the correct page

[cedeling]

Dear wdoc author

We must make summaries of transcripts of evidence given in court cases.

We have a PRIVATE website that restrics access to lawyers that are part of the team for the case.

We can provide url links to any page in any transcript

We request a way to load the text from the ocr'd pdfs (or even easier to just load the PDF's).

The wdoc summary must then create citations, that must have a link to the correct page of the relevant PDF that when clicked, will open the pdf at the correct page.

Perhaps the link for each PDF can be inserted as part of the metadata?

Can we be shown where in the wdoc code we shall be able to specify the format of the citation links?

Chris Edeling
cedeling@gmail.com

[cedeling]

The wdoc summaries will be based on PRIVATE documents that we upload, although any summaries exported in .docx or .pdf must have links to the citations, and a click on those links will open the pdf at the correct page - based on the format of the citation that we shall specify if we can edit the code. Only in that sense will the citation links expose public url's to pdf's under our control. It all remains private, with no internet access.

I have not yet seen an example of a wdoc summary with citations, and shall be grateful if you will provide a link to a sample document summary (in your response to this issue) so that users can see what the citations look like from wdoc

[thiswillbeyourgithub]

Hi,

Thank you for the interest in wdoc. I just pushed the branch enh-citations which should cover your use case. Can you please take a look at the diff (especially the new documentation), then give it a go and report back?

[cedeling]

Hi Olivier

Thanks for your quick response

We have not yet installed w working version of wdoc, it was a massive install and it MAY just be up and running tomorrow. My engineer loaded in the last three hours. Must he do any updates?

Is a branch a separate application? So we must reinstall?

Regards

Chris

[thiswillbeyourgithub]

Yes you have to install the new branch.

It depends on how you installed it but you could try for example with uv pip install git+https://github.com/thiswillbeyourgithub/wdoc@enh-citations[full]. Or to simply use uvx directly.

In any case it might be simpler to communicate directly with your engineer, if possible of course.

[thiswillbeyourgithub]

Btw all the tests were passing so I merged it to the dev branch. But I'm not confortable doing a new release until you tested it.

[cedeling]

Evening Oliver We did an installation and were able to reach the web interface *wdoc Web Interface (Experimental)Process documents with AI-powered query, summarization, and parsing.* We then tried twice to summarise, and Upload a pdf file 380.pdf (as now attached hereto) We waited a while and clicked Output where we saw an error message ❌ Error An error occurred while processing: Cli_kwargs 'doccheck_min_lang_prob' is of type '<class 'int'>' instead of '<class 'float'>' It also showed this error in the Logs, explaining that the error came from File "/tmp/wdoc-source/wdoc/wdoc.py", line 201, in __init__ Hope you will fix, and send me the file/s to be replaces Kind regards Chris

…

On Mon, 13 Apr 2026 at 16:37, Olivier Cornelis ***@***.***> wrote: *thiswillbeyourgithub* left a comment (thiswillbeyourgithub/wdoc#22) <#22 (comment)> Btw all the tests were passing so I merged it to the dev branch. But I'm not confortable doing a new release until you tested it. — Reply to this email directly, view it on GitHub <#22 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA46NU25IJO4GBJ5EOGPB3D4VT3UJAVCNFSM6AAAAACXWTPDJGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DEMZXGIYTQMBXG4> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[thiswillbeyourgithub]

The Web Interface is currently experimental. I suggest getting familiar with wdoc's many features first using the CLI and only then switching to the WebUI.

The idea here might be that your engineer finds the proper arguments appropriate for your setup (many tweaks here and there to make it work with local models appropriately as it depends on exactly the kind of model you have and the kind of documents you have etc). Then, the engineer would set those values as defaults valuefor the WebUI so that non technical user can use it without having to tweak things.

That being said, I might have found a bug thanks to your message and just pushed a fix on the dev branch. So hopefully this bug is no more. Thank you :)

[cedeling]

Oliver, my Engineer's name is Trent. I have asked him to sign up so that he can communicate directly with you - as you suggested in a prior message. He had trouble doing a pip install, and then did a docker installation (I think - may be wrong). Once Trent has done a re install with the enh-citations I shall happily test and give you more feedback.

Once we are more familiar with you code and arguments, we shall make our own web interface. You asked for my comments also on the new documentation. Where will I find that?

[trentmu]

Hi Olivier
I am the Trent mentioned above.
I have re-installed the system using the enh_citations branch (via docker)

Chris will test and revert.

Let me know if there's anything I should do to assist.
Thanks

[thiswillbeyourgithub]

Hi,

Sorry for maybe not being clear enough: I have since merged the enh_citations branch into the dev, and the last fix I added was on dev, not enh_citation.

I don't plan on using other branches than the dev branch so please, use the dev branch instead from now on.

To do so, assuming you have a local clone, I think the simplest would be to simply checkout to dev, then be sure to pull.

@cedeling I believe you might want to find a specific court document you find representative of what you want to give to wdoc.

Then @trentmu I believe you should experiment a bit using the CLI (not the webui, it's experimental so if something crashes we won't know if it's wdoc or just the webui) until you find the exact arguments you want to give to wdoc, including the env variable. Only then should you move up to the webui.

To use the cli, the easiest way is to use uvx --from git+https://github.com/thiswillbeyourgithub/wdoc.git@dev[full] wdoc as it should, hopefully, pull from the dev branch directly without having to deal with your local clone.

Does that seem alright to you?

[trentmu]

Thanks Olivier - I will test it again tomorrow

[cedeling]

Olivier we are still trying to come to grips with running your examples in a python3 environment. Your documentation and examples are comprehensive and appear to be top class, but they assume a user that knows how to run them.

We shall get there, and then revert with our further feedback.

[thiswillbeyourgithub]

(Sorry, I missclicked on the close issue button)

My 2ct: I highly recommend using LLMs to get this working. It really speeds things up regarding "setting up a new project I'm not familiar with". I personally am happy with Anthropic's claude code and the $20 or so dollars a month are easily worth it in saved engineer time.