Feature Highlights

Privacy-first clipboard mastery with on-device intelligence

Core Capture

Real-time clipboard watching, screenshots, per-app ignore, smart dedupe
Menubar popover with hotkeys, favorites, pinning, selection mode
SwiftUI polish: hover states, springy transitions, dark mode ready

Local AI (MLX)

On-device titles, tags, summaries; zero cloud calls
Model picker & loader for Llama 3, Mistral, Qwen 2.5 (4-bit)
Status-aware controls, resource toggles, manual model paths

Visual Intelligence

OCR via Vision; searchable extracted text with correction
Color detection, conversion, palette extraction (k-means)
Image metadata and timestamp labeling for screenshots

Organization & Safety

Advanced search across titles, summaries, tags, OCR text
Tag system for text/code/url/color/screenshot/file and more
Local-only storage, retention rules, protected pinned clips

Effort & Throughput

High-intensity 5-day sprint with 144 commits and sustained velocity

Commits

144

Duration

5d 12h

Net Lines Added

26,388

Active Coding

~132 hrs

Daily Commit Pulse

Nov 28

5

Nov 29

49

Nov 30

15

Dec 1

11

Dec 2

34

Dec 3

25

Dec 4

7

39,248 lines inserted

12,860 lines deleted

Peak: Nov 29 (49 commits)

Latest: Dec 4 09:10 (+0545)

Next Steps

Sharpen the on-device AI experience and deepen reliability

Ship semantic search (embeddings) to pair with current heuristics and tags.
Add resumable model downloads + checksum validation for MLX weights.
Privacy UX: redaction presets on export/share and clearer local-only indicators.
Smart summaries for screenshot OCR blocks; auto-title tuning per content type.
Performance: lazy image decoding + cache metrics for large histories.
Quick Actions panel for frequent transforms (case, encode/decode, clean up).

Tools & Stack

Everything touched to build and automate ClipKunda

Antigravity

Codex

Claude

Trae

Grok

SwiftUI + SwiftData

Vision & MLX

Local MCP server

Project Effort & Snapshot