Skip to content

potential space fix - #583

Open
dadukhankevin wants to merge 5 commits into
mainfrom
daniel/fix_db_inflation
Open

dadukhankevin wants to merge 5 commits into
mainfrom
daniel/fix_db_inflation

Conversation

@dadukhankevin

Copy link
Copy Markdown
Contributor

No description provided.

@dadukhankevin dadukhankevin linked an issue Feb 4, 2026 that may be closed by this pull request
@dadukhankevin
dadukhankevin force-pushed the daniel/fix_db_inflation branch 2 times, most recently from 0c52f29 to c301a99 Compare February 21, 2026 05:07
@dadukhankevin
dadukhankevin changed the base branch from main to dev March 22, 2026 20:40
dadukhankevin and others added 5 commits March 22, 2026 15:45
- Pass language code (lang param) to ASR endpoint for better transcription quality
- Include language in getAsrConfig response from extension to webview
- Add "Transcribe All" button alongside per-cell "Transcribe" to encourage batching
- Reuse single WhisperTranscriptionClient across batch to leverage warmed Modal container
- Default batch command to "all untranscribed cells" instead of requiring count
- Show cold-start warning (~45s) in transcription info messages
- Save actual language code instead of "unknown" in transcription metadata

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ignment (#429)

When importing a translation, the cell aligner correctly matches cells by
cellLabel (e.g. "GEN 1:1"), but handleWriteTranslation was using the
imported content's newly-generated UUID instead of the matched target
cell's existing UUID. This broke source-to-target linkage because the
target cells ended up with different IDs than their source counterparts.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Tests confirm the aligner matches by cellLabel and returns existing
target cell IDs, and that the handleWriteTranslation ID resolution
logic correctly prefers the matched notebook cell ID over the imported
content's newly-generated UUID.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
)

FTS5 virtual tables don't support INSERT OR REPLACE (no implicit rowid
matching), so every "upsert" was creating duplicate rows causing unbounded
database growth.

Changes:
- Update triggers to DELETE+INSERT instead of INSERT OR REPLACE
- Fix manual FTS sync in upsertCellWithFTSSync to use DELETE+INSERT
- Add bulk mode: FileSyncManager disables FTS triggers during sync and
  rebuilds the FTS index in a single pass from the cells table afterward
- Add disableFTSTriggers/enableFTSTriggers/rebuildFTSFromCells methods
- Bump schema version to 16 to force recreation with correct triggers
- Add FTS integrity tests verifying no bloat and correct search after
  updates, upserts, and bulk rebuilds

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The test's locateFile approach fails in webpack bundles because the
bundled sql.js WASM loader can't resolve filesystem paths correctly.
Instead, use eval("require")("fs") to bypass webpack's memfs polyfill,
read the WASM binary directly, and pass it via the wasmBinary option.
Tries multiple candidate paths (__dirname-relative + cwd-relative) to
work in both local dev and CI environments.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@dadukhankevin
dadukhankevin force-pushed the daniel/fix_db_inflation branch from 5c1de3d to 90b5949 Compare March 22, 2026 20:47
@BenjaminScholtens
BenjaminScholtens changed the base branch from dev-old to dev March 27, 2026 16:42
@TimRl
TimRl changed the base branch from dev to main April 16, 2026 20:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Investigate why sqlite db in project is becoming bloated

1 participant