Skip to content

Install and connect

Use the reviewed wheel or source distribution from each package's GitHub release. Private repositories require authenticated access. Keep credentials in your normal GitHub tooling; never put them in documentation, URLs or Python scripts.

Install inside FontLab

Install Instamix 1.0.9 or newer first. Download its standalone install.vfpy, then open that file in FontLab. It bootstraps the stdlib-only installer from a validated wheel; a pre-existing pip or venv is not required. The current generated project installer checks migration and rollback capabilities before reusing its provider. Do not use Instamix 1.0.8 to upgrade an old install.vfpy installation: a failed replacement could remove other files recorded by that legacy installer.

In FontLab's Scripting panel, install ordinary distribution requirements with:

from fontlab_instamix import install_packages

install_packages("uharfbuzz>=0.57")

Instamix finds or downloads uv, verifies a standalone Python matching the target, and creates or checks the shared venv. Packages go to <data>/python/<major.minor>/site-packages, which FontLab already reads. The venv reads that same directory through its own managed .pth; FontLab does not need to process that file. An incompatible existing venv is refused rather than replaced. Installation refreshes import caches, without reloading imported modules.

For a source project with [tool.fontlab-install], open its current install.vfpy. For a ready .fontlabext, .fontlabextz or root manifest, use the Instamix extension API. Its records and uninstall commands are under Scripts > Instamix.

Default data-folder discovery uses an explicit argument, then FONTLAB_DATA_DIR, then native preferences' dataPathDefault, then the platform default. A custom current application dataPath is a different value: pass the intended folder explicitly when those differ. Never assume a terminal's Python version or architecture matches the running FontLab.

Install outside FontLab

Install external clients such as Remopy into an ordinary environment with uv pip install PATH_TO_WHEEL. That environment does not make packages available to embedded FontLab. Instamix explains explicit targets and the external CLI. FontLab supplies its native bindings and existing Qt application; install neither PySide binaries nor a second Qt application into that host.

Temporary release inputs

App 1.2.31, Plugmix 1.1.19 and Remopy 2.0.7 have checked GitHub release artifacts, but their PyPI uploads failed because trusted publishers were not configured. Do not lower a dependency floor to work around a missing release. Use the checked release-wheel download URLs with uv pip install in an ordinary Python environment. The App and Plugmix versions above resolved together in a fresh ordinary arm64 Python 3.11 environment.

That is an ordinary package-install check, not a FontLab plugin deployment test. Use the reviewed source installer or extension pack for its Scripts and bundles. Recheck publication status when choosing later releases; keep version floors and artifact identity aligned. Instamix 1.0.9 or newer is still needed for legacy project upgrades.

The captured API uses FontLab9, Python3.11 and Qt5. The source-only documentation environment uses Python3.12 or newer. A PySide2 binary wheel is a reference input; it is not loaded into FontLab's PythonQt runtime.

Connect from another process

FontLab 9.1.9813 bridge defect

The native MCP bridge can crash in BridgeConnection::updateTimeout on this build (fontlab/fl#8657). Remopy setup/tools/run and Instamix --remote can time out. A submitted operation's outcome is unknown until its receipt is checked. Do not retry automatically. Until a fixed bridge ships, run an owned .vfpy directly through FontLab's File > Open and have it write a result file; submission alone proves no work.

On a build with a working MCP bridge, enable FontLab's MCP server, then connect:

from fontlab_remopy import FontLab

with FontLab() as app:
    print(app.fonts())
    print(app.output())

eval() returns a JSON-native value. run() captures stdout/stderr and a result; bind variables as keyword arguments rather than interpolating them into code. The native execution deadline still applies. A timeout can leave an operation's outcome unknown, so mutating calls are never retried automatically.

with FontLab() as app:
    answer = app.eval("value + 2", value=40)
    execution = app.run("print(message)\nresult = 42", message="Hello, FontLab")

For asynchronous programs, use AsyncFontLab with async with and await calls.