Skip to content
Log in Register
Donate
GitHub

Docker

AI assistant

Build this with Claude Code, Cursor or Copilot

Copy a Tessero-tuned prompt for Docker, grounded in 16 real API signatures , into your IDE's AI. No chatbot, just exact context.

The repository ships a reference image, docker/Dockerfile: PHP-FPM with ext-tessero, the FFI package, OpenBLAS/LAPACKE and production INI settings (docker/tessero.ini).

docker build -f docker/Dockerfile -t tessero-fpm .
docker build -f docker/Dockerfile --build-arg PHP_VERSION=8.3 -t tessero-fpm:8.3 .
docker run --rm tessero-fpm php -d ffi.enable=1 /opt/tessero/bin/tessero doctor

The build has two stages:

  1. build: compiles libtessero (and runs its C tests), then compiles the extension against the image's PHP and runs the phpt suite. A failing test fails the image build.
  2. runtime: php:*-fpm-bookworm, plus libgomp1 (OpenMP), libopenblas0-pthread and liblapacke (linear algebra), ffi and opcache, the compiled extension, and the FFI package under /opt/tessero. It ends with a smoke check that both backends load.

In an application image you would normally get the FFI package through Composer (vendor/tessero/tessero) and copy only the extension from the build stage. Point opcache.preload at vendor/tessero/tessero/resources/preload.php.

Alpine

Alpine uses musl. Use the linux-musl-x86_64 binary for the FFI package, and build the extension in an Alpine stage:

FROM php:8.4-fpm-alpine AS build
RUN apk add --no-cache $PHPIZE_DEPS linux-headers libgomp
COPY ext/ /src/ext/
RUN cd /src/ext && phpize && ./configure --enable-tessero && make -j"$(nproc)" && make install

FROM php:8.4-fpm-alpine
RUN apk add --no-cache libgomp openblas lapack libffi \
 && apk add --no-cache --virtual .deps libffi-dev $PHPIZE_DEPS \
 && docker-php-ext-install ffi opcache && apk del .deps
COPY --from=build /usr/local/lib/php/extensions/ /usr/local/lib/php/extensions/
RUN docker-php-ext-enable tessero

Alpine's lapack package does not include the LAPACKE C interface on every release. If tessero doctor reports no LAPACKE, use the Debian image for linear algebra workloads. The extension does not need LAPACK.

Resource limits in containers

  • Set tessero.memory_budget (or TESSERO_MEMORY_BUDGET) so that pm.max_children × (memory_limit + memory_budget) fits inside the container's memory limit ([Memory and threads(https://github.com/DMJ-CV-91913/tessero/blob/main/docs/operations/memory-and-threads.md)).
  • OpenMP detects the host's CPUs, not the container's CPU quota. Keep tessero.threads at 1 under FPM, and set it explicitly (not "all cores") for worker containers.
  • OpenBLAS likewise starts one thread per host CPU by default. The FFI package sets it to 1 in non-CLI SAPIs; set TESSERO_BLAS_THREADS for CLI workers.

Status

The Dockerfile is built by CI (docker job). It has not been built in the reference environment, which has no container daemon.

Edit this page on GitHub