Further improve Dockerfile - Reduce Docker image size and simplify Dockerfile (attempt to)

[tCR-lux]

See comments in the PR. The idea is to make the Dockerfile more structured, better control, reduce the size of final image, upgrade dependencies whereever applicable, improve security and build robustness posture.

One breaking change yet to be validated: the venv is now in /opt and linked from /home/flowintel/app/venv

It includes some rework on TexLive and pandoc integration through the EisVogel template. Reducing further the footprint would mean to switch TexLive for Typst apparently, but it seems like we would need a custom template for that ... much work I guess ? on the other hand, it would provide the user the possibility to customize its template ?

[tCR-lux]

• first commit (bookworm:slim):

IMAGE   ID             DISK USAGE   CONTENT SIZE   EXTRA
flowintel:latest       863c19fd42cd       5.87GB         1.54GB

• same refactor but with ubuntu:noble again

IMAGE   ID             DISK USAGE   CONTENT SIZE   EXTRA
flowintel:latest       135c73309f26       6.37GB         1.71GB

• v4.0.0:

IMAGE   ID             DISK USAGE   CONTENT SIZE   EXTRA
flowintel:latest       96531be5700b       10.7GB         3.05GB

• ??? alpine ???
  • with current stack (LaTeX, pandoc, etc.), unlikely to be worth the attempt given the complexity !

The impact of multi-stage is seen on "disk usage", the impact of cleanup on "content size" (especially apt cleanup cache for the later).

[tCR-lux]

To go further, a "dive" into the layers of the image led to the fact .git folder is owning itself 107MB. Adding this step and force pushing.

Then, a quick summary and Better Call Claude to get an overview:

• the /usr folder is 2.7GB.
• the /var is 152MB
• the /opt is 615MB (python3 venv with pycache already expanded)
• pandoc binary is 158MB
• mermaid node modules around 509MB
• there is a X11 folder in /usr ?
• there is a Java folder in /usr ?
• there is a TEXLIVE folder in /usr around 1.6GB, its libs are also major. lighter latex alternative ? texlive-minimal ?

also and finally, the wasted space is:

Image name: flowintel                                                                                                                                                                                              
Total Image size: 3.7 GB                                                                                                                                                                                           
Potential wasted space: 38 MB                                                                                                                                                                                      
Image efficiency score: 99 %                                                                                                                                                                                       
                                                                                                                                                                                                                   
Count   Total Space  Path                                                                                                                                                                                          
    2         33 MB  /tmp/pandoc.deb                                                                                                                                                                               
    2        1.6 MB  /var/cache/debconf/templates.dat                                                                                                                                                              
    2        1.5 MB  /var/cache/debconf/templates.dat-old                                                                                                                                                          
    3        770 kB  /var/lib/dpkg/status                                                                                                                                                                          
    3        770 kB  /var/lib/dpkg/status-old                                                                                                                                                                      
    2        137 kB  /var/log/dpkg.log                                                                                                                                                                             
    2         82 kB  /var/lib/dpkg/info/perl-base.list                                                                                                                                                             
    2         69 kB  /var/log/apt/term.log                                                                                                                                                                         
    2         60 kB  /home/flowintel/.pandoc/templates/eisvogel.latex

pfiou ... quand c'est ton Chatbot qui inverse les rôles ... (et se trompe sur ce que tu veux ...)

[tCR-lux]

Rework of TexLive packages with debuging docker run --rm -it --user root --entrypoint /bin/bash flowintel then su flowintel -c bash to execute tests and exit to become root again and install missing packages.

• apt pkgs discovered:

texlive-latex-extra \
texlive-xetex \
texlive-fonts-recommended \
texlive-fonts-extra \
texlive-lang-cjk \
lmodern

for pandoc and Eisvogel's documented deps:

for pkg in adjustbox.sty mdframed.sty sourcesanspro.sty sourcecodepro.sty xecjk.sty; do
    result=$(find / -iname "$pkg" 2>/dev/null)
    [ -n "$result" ] && echo "FOUND: $result" || echo "MISSING: $pkg"
done
FOUND: /usr/share/texlive/texmf-dist/tex/latex/adjustbox/adjustbox.sty
FOUND: /usr/share/texlive/texmf-dist/tex/latex/mdframed/mdframed.sty
FOUND: /usr/share/texlive/texmf-dist/tex/latex/sourcesanspro/sourcesanspro.sty
FOUND: /usr/share/texlive/texmf-dist/tex/latex/sourcecodepro/sourcecodepro.sty
FOUND: /usr/share/texlive/texmf-dist/tex/xelatex/xecjk/xeCJK.sty

• validate creations of docs:
  pandoc --template=eisvogel --pdf-engine=xelatex -o /tmp/test.pdf <<< "Test"

cat <<MD | pandoc --template=eisvogel --pdf-engine=xelatex -o /tmp/test-full.pdf --toc
---
title: Test Report
author: Flowintel
date: 2026-09-18
---

# Section

Some text with a footnote.[^1]

\`\`\`python
print(\"code block test\")
\`\`\`

[^1]: Footnote content.
MD

[tCR-lux]

Upgraded some important versions that are now also made parametric inside Dockerfile:

• NODE: switch from the EOL 20.x version, je pense c'est important pour les install legacy aussi
• PANDOC: upgrade the final patch of the retained 3.7.z line, no big risk with current legacy install
• EISVOGEL: un freeze pour rendre le build et la creation de document repetable. le latest est passe sur une version qui integre un change de nom sur sourcesanspro -> sourcesans, un truc adobe, je pense il faut freezer avant tant que les distrib classique n'ont pas repercuter ce change autrement la generation de doc est cassee

we may maintain these values inside a common .envbuild that can be leveraged in the install.*.sh scripts too so that legacy and dockerised installs are in harmony.

[tCR-lux]

As of today:

• ubuntu:noble
  IMAGE ID DISK USAGE CONTENT SIZE EXTRA
  flowintel:latest 822fdc5a8241 6.15GB 1.62GB
• debian:bookworm
  IMAGE ID DISK USAGE CONTENT SIZE EXTRA
  flowintel:latest 85f92c1b147c 5.77GB 1.5GB
• debian:trixie
  still bug due to misp-module wheel not compatible yet with higher python3 versions