Speakr is a personal, self-hosted web application designed for transcribing audio recordings
Python
Commits on master
tests: force external-service settings empty so a dev container or .env never reaches a real LLM, embeddings API, ASR or mail server; hide sentence-transformers in the local embeddings fixture so the image does not decide the result (fixes the 4 local-only test_cov_embeddings failures)
872b7cd Murtaza Nasir committed 1 month ago
Voice training commits separately from the save it follows and rolls back as a whole on failure (the label map is still kept); a savepoint was tried and dropped because it holds the SQLite write lock through training
7308b9a Murtaza Nasir committed 1 month ago
Merge chore/release-prep: coverage tests, ffmpeg pin, CI floor 60
49dbb05 Murtaza Nasir committed 1 month ago
Fix pre-release review findings: per-user voice samples, label map on rename, one complete delete for bulk/batch/merge, per-user API v1 archive with strict parsing, vectorised calibration, audio-removed endpoint alias
18ffcd0 Murtaza Nasir committed 1 month ago
ci: raise the coverage floor from 54% to 60%; CI measured 62.6% at v0.10.6 and this release adds well-covered code
16ff181 Murtaza Nasir committed 1 month ago
test: cover the untested branches of the release changes found by diff coverage (variant cap, singleton penalty, ambiguity, calibrated thresholds, refusal paths, deletion failure handling)
90e9f40 Murtaza Nasir committed 1 month ago
build: refresh the pinned ffmpeg build to autobuild-2026-09-28 (n8.1.3), the previous pin is ten days old and BtbN prunes after about two weeks
fff5998 Murtaza Nasir committed 1 month ago
docs: README release highlights include the voice matching rework
5352aa6 Murtaza Nasir committed 1 month ago
Merge branch feat/voice-matching: sample-based voice profiles with variants, voice spaces per embedding model, calibrated thresholds, profile management UI
8e5d16a Murtaza Nasir committed 1 month ago
test: voice sample list and removal endpoints, pre-sample profile entry, speaker voice counts, admin space counts
cbc6309 Murtaza Nasir committed 1 month ago
feat: voice profile panel per speaker with samples, variants and removal, voice spaces and calibrated thresholds on the admin card, and docs for sample-based voice matching
2cc738c Murtaza Nasir committed 1 month ago
Merge branch fix/deletion: shared recording deletion, sidebar toggle labels
f14f31d Murtaza Nasir committed 1 month ago
fix: drop the on/off icons from the Archived and Shared sidebar toggles so their labels fit; the highlighted background shows the state
81e5de8 Murtaza Nasir committed 1 month ago
fix: one deletion path for the web app, API v1 and retention; API v1 left the media file behind because it joined the upload folder with the storage locator, and retention sent no recording.deleted webhook
c494bea Murtaza Nasir committed 1 month ago
docs: README highlights for the next release: Archive and Audio removed, header actions, speaker dialog, title prompts
ab00de2 Murtaza Nasir committed 1 month ago
Merge branch feat/archive: per-user archive, Delete audio (including retained video), Audio removed state, grouped header actions (#394)
883115a Murtaza Nasir committed 1 month ago
feat: Archive and Delete audio in the recording header, grouped header actions with Reprocess and More menus, an Archived view and an Audio removed filter in the sidebar, bulk archive, and video-aware labels (#394)
8882b21 Murtaza Nasir committed 1 month ago
feat: per-user archive that hides a recording from the main list without deleting anything, and a Delete audio action that keeps the transcript; the audio-deleted state is now filtered as audio_removed (#394)
64157f5 Murtaza Nasir committed 1 month ago
Merge branch feat/title-prompt: configurable AI title instructions with the standard prompt precedence (#400)
9bdbe91 Murtaza Nasir committed 1 month ago
feat: title prompt fields in Prompt Options, the tag and folder dialogs, group tag and folder forms and the admin Default Prompts tab, with docs and translations (#400)
4092b36 Murtaza Nasir committed 1 month ago
fix: keep the page loader hidden once the page has shown itself, a late overlay timer could re-add it after /account finished loading and leave the page covered
59097f0 Murtaza Nasir committed 1 month ago
feat: configurable AI title instructions at tag, folder, user and admin level with the same precedence as the summary prompt, used for regular and incognito titles and exposed through the tag, folder, account and API v1 endpoints (#400)
05126f1 Murtaza Nasir committed 1 month ago
Merge branch feat/speaker-modal: rebuilt Identify Speakers modal and shared speaker save path (#395)
173b9a7 Murtaza Nasir committed 1 month ago
feat: rebuild the Identify Speakers modal with speaking-time sorting, a filter to one speaker, grouped short speakers that can be merged, name suggestions on focus with keyboard support, per-speaker samples and a playing-segment highlight (#395)
da38b9c Murtaza Nasir committed 1 month ago
feat: pure speaker-modal helpers for speaking-time stats, sorting, minor-speaker grouping, name options and merging (#395)
a1f45c3 Murtaza Nasir committed 1 month ago
fix: apply speaker names through one helper in both save routes, so saving with a staged line edit also updates voice profiles and snippets, and match saved speakers ignoring case instead of creating a second profile per spelling (#395)
31adc3b Murtaza Nasir committed 1 month ago
test: blank the logout form CSRF token too when comparing shell renders, the second timed token kept the label shell test flaky
a048d32 Murtaza Nasir committed 1 month ago