Skip to content

Releases: nageshkr/owl

OWL 1.0.3

Choose a tag to compare

@nageshkr nageshkr released this 15 Oct 11:41
f7308e5

OWL 1.0.3

  • Minor fixes for office days report

OWL 1.0.2

Choose a tag to compare

@nageshkr nageshkr released this 15 Oct 07:19
f7308e5

### OWL 1.0.2

⏱️ Add & edit tasks and meetings right in the hourly view
My Day's hourly timeline is now fully interactive — plan and adjust your day without leaving the grid.

Double-click and drag on an empty slot to create an entry: log time to a task or add a meeting, chosen in the picker.
Click any block to edit — adjust its time, change the meeting title/project or the task note, or delete it, all inline on the grid.
Dragged time ranges snap sensibly and pre-fill the start/end, so creating and tweaking entries is quick.
Longer day view: the timeline now runs to 10 PM by default (and still expands automatically for anything earlier or later), so evening work and meetings are visible.

Everything stays local and private, as always.

OWL 1.0.1

Choose a tag to compare

@nageshkr nageshkr released this 14 Oct 12:10
f7308e5

### OWL 1.0.1 — the first stable release
OWL is a privacy-first, offline desktop app for notes, tasks, time, and a local AI assistant. Everything runs on your device — your notes, documents, tasks, attendance, and AI conversations never leave your machine. This 1.0 brings knowledge-aware AI, faster time logging, hybrid-work office-day tracking, and a layout you can tailor to how you work.

✨ Open Knowledge Format (OKF) — smarter, context-aware AI
Give the local AI better context from your own notes, with zero cloud involvement.

Opt-in (Settings → Notes Q&A) — off by default, so nothing changes until you turn it on.
Add a note's type, description, and tags, or let "Suggest tags" read the note and propose them for you to confirm before anything is written.
A tidy metadata bar above the editor shows a note's metadata, with inline edit / Add metadata — no hand-writing YAML.
When on, metadata both sharpens AI answers and boosts search, surfacing notes by what they're about. Notes without metadata are never penalized.
Your Markdown files remain the source of truth — metadata is only written when you confirm it.
🏢 Office-day & attendance tracking
Stay on top of hybrid-work requirements from the Dashboard.

Mark days as Office / Home / Leave / Sick; set your required office days per week in Settings.
Three views: Week (timeline + progress), Month (calendar heatmap), and Report (compliance gauge, stats, and a trend over week / month / quarter / year).
Fair compliance — leave and sick days reduce that week's requirement, and a full week off is excluded, so time away never counts against you.
CSV reports from the assistant (Chat or Tasks): ask office days report week (also month / quarter / year) to download a report with an overall summary, a per-period breakdown, and a day-by-day log.
⏱️ My Day — faster time logging
The hourly view is now a true time grid: double-click and drag to log time against a ticket, pick an active task, add a note.
The ticket picker now includes tasks that are in progress or in testing.
Pick your default worklog view (List or Hourly) in Settings.
🎛️ Make OWL yours
Hide tabs you don't use (Settings → Appearance). Settings always stays put, so you can bring anything back.
Show or hide Dashboard cards — Office days and Sticky notes.
🔒 Private by design
Fully offline. Local AI via a bundled engine (or your own Ollama). No telemetry, no accounts, no data leaving your device.

Upgrading? Your data migrates automatically on first launch. OKF stays off until you enable it; office tracking starts fresh; your layout preferences are per-device.

OWL 1.0.0

Choose a tag to compare

@nageshkr nageshkr released this 14 Oct 11:58
f7308e5

OWL 1.0.0 — the first stable release

OWL is a privacy-first, offline desktop app for notes, tasks, time, and a local AI assistant. Everything runs on your device — your notes, documents, tasks, attendance, and AI conversations never leave your machine. This 1.0 brings knowledge-aware AI, faster time logging, and hybrid-work office-day tracking.

✨ Open Knowledge Format (OKF) — smarter, context-aware AI
Give the local AI better context from your own notes, with zero cloud involvement.

Opt-in (Settings → Notes Q&A) — off by default, so nothing changes until you turn it on.
Add a note's type, description, and tags, or let "Suggest tags" read the note and propose them for you to confirm before anything is written.
A tidy metadata bar above the editor shows a note's metadata, with inline edit / Add metadata — no hand-writing YAML.
When on, metadata both sharpens AI answers and boosts search, surfacing notes by what they're about. Notes without metadata are never penalized.
Your Markdown files remain the source of truth — metadata is only written when you confirm it.
🏢 Office-day & attendance tracking
Stay on top of hybrid-work requirements from the Dashboard.

Mark days as Office / Home / Leave / Sick; set your required office days per week in Settings.
Three views: Week (timeline + progress), Month (calendar heatmap), and Report (compliance gauge, stats, and a trend over week / month / quarter / year).
Fair compliance — leave and sick days reduce that week's requirement, and a full week off is excluded, so time away never counts against you.
CSV reports from the assistant (Chat or Tasks): ask office days report week (also month / quarter / year) to download a report with an overall summary, per-period breakdown, and a day-by-day log.
⏱️ My Day — faster time logging
The hourly view is now a true time grid: double-click and drag to log time against a ticket, pick an active task, add a note.
The ticket picker now includes tasks that are in progress or in testing.
Pick your default worklog view (List or Hourly) in Settings.
🔒 Private by design
Fully offline. Local AI via a bundled engine (or your own Ollama). No telemetry, no accounts, no data leaving your device.

Upgrading? Your data migrates automatically on first launch. OKF stays off until you enable it; office tracking starts fresh.

OWL 0.7.0

Choose a tag to compare

@nageshkr nageshkr released this 10 Oct 04:17
f7308e5

OWL 0.7.0 — Images in Notes

Add images to your notes without giving up privacy. Everything stays local — nothing is uploaded, and each note remains a single self-contained Markdown file.

✨ New
Images in the editor

Add images three ways: paste a screenshot, drag & drop a file, or the new 🖼 Image toolbar button.
Supports PNG, JPEG, GIF, WebP, and BMP.
Large images are automatically downscaled (max 1600px) and re-encoded to keep notes light — you'll get a toast when it happens. Transparent PNGs keep their transparency, and animated GIFs stay animated.
Images embed directly into the Markdown, so notes still encrypt, sync, export, and copy as one portable file.
Images in PDF export

Exported PDFs now render embedded images as real pictures, scaled to fit the page and split across pages when needed.
Fully offline: images we can't reach without a network (like remote http links) fall back to their alt text instead of failing.

🔒 Privacy & Search
Embedded image data is never indexed — your notes stay searchable by their text and image captions, without image bytes bloating or polluting search results. Your Markdown files on disk are never modified.
Notes that had images from before this release are cleaned up automatically on first launch.

Full upgrade is seamless — no action needed. Images and search indexes update themselves on launch.

OWL 0.6.6

Choose a tag to compare

@nageshkr nageshkr released this 09 Oct 11:53
54d9356

Search quality — optional reranking

  • Cross-encoder reranking for Notes Q&A. Hybrid retrieval (keyword +
    semantic + RRF fusion) can now run an optional second pass that re-scores the
    retrieved snippets with a small on-device cross-encoder and keeps the best
    ones, improving answer relevance. Fully offline via a third bundled
    llama-server sidecar (--reranking --pooling rank, port 11437).
    • Off by default, user-toggleable in the new Settings → Search tab.
    • Download-on-demand model (bge-reranker-v2-m3, ~600MB) — not bundled in
      the installer, so the download stays small. The toggle stays disabled until
      the model is present.
    • Configurable candidate pool — over-retrieve N snippets before reranking
      down to top-K. Defaults to a smart max(top-K × 4, 30); override 10–100 for
      power users (0 = default).
    • Best-effort and never fatal — if the model is missing, the sidecar can't
      start, or the provider (Ollama/External) has no rerank endpoint, OWL falls
      back to the fused order so a question is never failed by reranking.
    • The reranker loads on demand and is released right after each question to
      keep idle memory low (first question after an idle period warms up for a few
      extra seconds).

AI master switch

  • New Settings → AI tab with a master "Enable AI features" toggle. Turn OWL
    into a pure notes, tasks and search app — the assistant panel, its toolbar
    button, the ⌘/ shortcut, Notes/Document Q&A, the Jira "Refine with AI" action,
    and LLM query expansion all disappear when it's off.
    • Backend-enforced, not just hidden. When off, OWL never starts the chat
      model; turning it off also stops a running bundled chat server (including an
      adopted/orphaned one on its port) so no chat model stays resident.
    • Search keeps working. Keyword and embedding-based semantic search are
      independent of this switch. The Search tab stays available with its own
      on/off switches for Semantic search and Reranking.
    • On by default and persisted in settings.json (existing installs are
      unaffected; older settings files load as enabled).

Settings screen

  • Toggle settings now render as switches instead of checkboxes (Appearance,
    Notes Q&A, Search, Data & Sync, Reminders), for a clearer on/off affordance.
    Multi-select lists (e.g. DevTools clean options, worklog pickers) stay as
    checkboxes.

Fixes

  • Turning AI features off now reliably frees the bundled chat port (11435),
    including a leftover/orphaned llama-server from a previous session, so "AI
    off" is a hard guarantee on the bundled provider. User-managed Ollama/External
    servers are never force-stopped.

OWL 0.6.5

Choose a tag to compare

@nageshkr nageshkr released this 08 Oct 16:19
54d9356

OWL 0.6.5

OWL 0.6.4

Choose a tag to compare

@nageshkr nageshkr released this 07 Oct 14:19
f768c42

OWL 0.6.4

OWL 0.6.3

Choose a tag to compare

@nageshkr nageshkr released this 07 Oct 13:50
f768c42

OWL 0.6.3

OWL 0.6.2

Choose a tag to compare

@nageshkr nageshkr released this 07 Oct 13:38
f768c42

OWL 0.6.2