Repository navigation
Releases: SacitDev/Fluentfolio
Release list
Fluentfolio 0.1.1
A security update for Fluentfolio on Windows and Ubuntu.
- Updates filelock, OpenTelemetry and setuptools to address five published dependency advisories. The security scan now passes without ignoring or suppressing findings.
- Updates the same affected packages in the optional MFA aligner environment. An older environment is shown as needing an update and cannot run until repaired.
- Keeps recordings, analyses, settings and downloaded speech models. The app's privacy settings and default tools are unchanged.
Install or update
Download Source code (zip) below and extract it. On Windows, run Install.cmd; on Ubuntu, run ./install.sh. If Fluentfolio is already installed, close it first and use the installer to update it.
If you use the optional MFA aligner, install that component again from Setup, or run fluentfolio components install mfa after updating. This rebuild downloads its environment packages again and needs an internet connection. MFA remains experimental and off by default.
The existing limitations and known risks in SECURITY.md still apply, including the separately documented older pillow/certifi packages in the optional MFA environment. This update addresses the five reported advisories; it does not claim that every third-party component is free of vulnerabilities.
The v0.1.0 release and its history are preserved.
Fluentfolio 0.1.0
The first release.
- Analyses a recorded IELTS Speaking practice test on your own computer, on Windows 10 22H2 or later, Windows 11, and Ubuntu 22.04 or 24.04.
- Three modes (solo, interview, role swap) and three recording layouts (one recording of everyone, one track per person, two files).
- Speech recognition with Parakeet (the default) or Whisper; speaker separation with pyannote (with your own Hugging Face token) or sherpa-onnx (no account).
- Finds the questions, the three Parts and who asks and who answers, with a review form to correct them.
- Measures fluency, hesitation, linking words, vocabulary, grammar pointers, pitch and loudness. The MFA aligner is experimental and off by default, because it did not pass its accuracy check. The pronunciation check did not pass its accuracy check either: it is on by default, marked experimental, and its words to check go to the AI only as hints.
- Copy for AI: one text with a grading prompt for any AI chat, and an optional short audio file for chats that can listen.
- A web app in your browser and a
fluentfoliocommand for scripts and AI agents, with JSON output, an answer checker and a guide for agents. - An installer that needs no administrator rights, with update and uninstall.