Skip to content

Praxis v0.2.0 — The remake

Latest

Choose a tag to compare

@Twarga Twarga released this 14 Jul 20:44
· 1 commit to main since this release

Changelog

All notable changes to Praxis are documented in this file.

The format is based on Keep a Changelog,
and this project adheres to Semantic Versioning.

Unreleased

0.2.0 - 2026-07-14

Added

  • A complete desktop workspace remake with six selectable visual themes.
  • A redesigned first-run flow that lets people activate multiple coaching tracks
    together, including journaling, language practice, interviews, and presentations.
  • Camera and microphone preview before recording, advanced device controls, and
    clearer exit and navigation behavior.
  • Local Faster-Whisper runtime verification and a repaired transcription speed
    benchmark path.
  • A direct "Run fresh analysis" action from every completed session.
  • Single-language and bilingual subtitle export, translation with the selected
    AI provider, and burned-in MP4 output.
  • Provider activation and diagnostics surfaces that use explicit connections
    rather than legacy connection reuse.
  • A new professional repository identity: a continuous-loop logo, release
    banner, refreshed README, product metadata, and GitHub Pages landing site.
  • Reproducible Linux AppImage packaging scripts, release metadata, and a
    one-command installer that downloads and verifies the latest release.

Changed

  • Reports now lead with one main lesson, evidence, and the next drill instead
    of dense scorecards and divider-heavy layouts.
  • Coaching prompts use the selected tracks and actual recording context, so a
    product-testing session is not incorrectly treated as a generic personal
    journal entry.
  • Recording, practice, progress, settings, and onboarding surfaces are calmer,
    more readable, and use top-centre desktop notifications.
  • The landing site is rebuilt around the record → understand → practise loop.

Fixed

  • Provider configuration no longer depends on stale inherited connections.
  • The local Whisper benchmark uses the correct operating-system import.

0.1.0 - 2026-05-07

Added

  • First-run onboarding for journal folder, goal, language, AI provider, and
    Whisper setup.
  • Multi-provider LLM configuration with OpenRouter, OpenAI-compatible APIs,
    and LiteLLM proxy support.
  • Coach-first analysis reports with scorecards, timestamped moments, lessons,
    practice assignments, and language rewrite drills.
  • Session Detail provider/model override for re-analysis without opening
    Settings.
  • Stats page with improvement dimensions: specificity, actionability, clarity,
    and language fluency.
  • Practice tracking for assignment completion and previous-goal follow-up.
  • Session gallery with month grouping, language filter, and thumbnails.
  • LAN phone upload with QR code.
  • Weekly rollups, streaks, and recurring pattern tracking.
  • Subtitle generation, translation, and burned-in MP4 export.
  • Real-time processing status via SSE.
  • Crash recovery and retention compression.
  • Linux AppImage packaging with bundled Python runtime and static
    FFmpeg/FFprobe resources.
  • Production release script with tests, AppImage build, smoke launch, and
    SHA-256 checksum.
  • AppImage installer (scripts/install.sh) that creates a local launcher and
    desktop entry.
  • Optional Whisper cache pre-seeding during AppImage builds via
    PRESEED_WHISPER=1 or PRESEED_WHISPER_FROM=/path.

Known gaps

  • Whisper model files are not pre-seeded inside the AppImage by default, but
    release builds can bundle them explicitly.
  • Prompt calibration still needs 10 real English/French/Spanish recordings.
  • Recurring behavior pattern calibration UI is not implemented yet.
  • Full manual fresh-machine release QA is still pending.

Full Changelog: https://github.com/Twarga/praxies/commits/v0.2.0