Skip to content

CHANGELOG.md

Latest commit

 

History

History
89 lines (72 loc) · 8.33 KB

CHANGELOG.md

File metadata and controls

89 lines (72 loc) · 8.33 KB

Changelog

All notable changes to this project are documented in this file.

The format is based on Keep a Changelog, and this project adheres loosely to Semantic Versioning. Detailed feature breakdowns by version live in MVP_FEATURES.md.

Unreleased

Added

  • Documentation polish for public release: SECURITY.md, CONTRIBUTING.md, CODE_OF_CONDUCT.md, CHANGELOG.md, .env.example, docs/ARCHITECTURE.md, docs/TESTING.md, docs/IMPROVEMENTS.md, GitHub issue/PR templates.
  • README rewritten for an external audience: tightened hero section, Mermaid architecture diagram, privacy-and-security table, expanded documentation map.

Changed

  • LICENSE copyright year bumped to 2025-2026.

0.2.1 — Analytics & Focus Mode

Added

  • Tab Habits Score — composite score with 4 weighted components (closable %, cleanup frequency, average tab age, count trend). Trend detection vs. 7-day baseline. Surfaced in the AI panel Tab Insights card.
  • Recommendation Tracking — every Accept / Skip / Change-Action in a cleanup session is logged to the recommendation_actions SQLite table, with acceptance rate and per-action breakdown shown in Tab Insights.
  • Activity Heatmap — 7×24 grid (days × hours) built from tab_history_events, with a per-domain filter.
  • Persistent Topic Clusters — topic_clusters table; AI clusters are auto-merged across analyses by case-insensitive name match → 2+ shared tags → otherwise new entry. Saved Clusters section can rename, delete, or focus a cluster.
  • Focus Mode — focuses Chrome on tabs that match a saved cluster: creates a tab group from matching tabs and collapses everything else; one click to exit.
  • Resumable Analysis — analysis_runs SQLite table stores full run state (pending tabs, per-tab statuses, partial result, metadata). The user can stop a long analysis and resume later with the same fingerprint, surviving extension reloads.
  • Per-tab analysis status — TabStatusCoverageCard shows which open tabs are cached, freshly analyzed, pending, or failed, including pending-tab domains.
  • Search dialog (/chat) — dedicated Search side-panel tab; conversational retrieval over url_analysis, tab_history_events, and topic_clusters via a single UNION ALL query (up to 6 keyword patterns, 20-second LLM timeout). Falls back to SQLite-only ranking if the model is unavailable. Input stays active while a query runs.
  • Refresh buttons — amber ↻ on Analysis (forces full LLM re-analysis ignoring SQLite cache, with confirmation) and ↻ on Analytics (sends aggregated SQLite analytics to the LLM via POST /analytics/refresh and renders an AI Insight card).
  • Suggested next steps — cards for "finish analysis", "review close candidates", "group cluster", "process reading queue".
  • Theme comparison — current AI clusters matched against persistent ones by shared tags; loading state prevents false "new theme" flashes before persistent clusters load.
  • 4 new request types (OPEN_URL, FOCUS_TAB, CHAT_SEARCH, REFRESH_ANALYTICS) on top of the analytics/focus-mode requests added earlier.

Changed

  • AI panel split into Analysis / Analytics sections via activeSection state.
  • AIProgressPhase extended with stopping and stopped phases; AIProgress.tabsRemaining added for resume tracking.
  • Recommendation list collapsed by action with per-tab status badges; only close recommendations get a destructive button.
  • Theme tokenization is now title/path/query/group-aware instead of relying on raw hostnames, so theme-first clustering works across multiple-domain topics.

0.2 — AI analysis, history, cleanup, advanced exports

Added

  • FastAPI AI server (agent.py) with CLI provider layer (claude_code, codex_cli), CORS for the extension, and structured prompt + JSON-schema output.
  • Endpoints — /health, /analyze, /analyze/cancel, /stats, /history, /cache-stats, /settings (GET/POST), /db-status, /runtime-logs, plus tab-history, snapshot, and analysis-run endpoints.
  • SQLite-backed cache — per-URL AI analysis cached in tab_analysis.db with a 180-day TTL, namespaced by provider/model so settings changes bust the cache correctly.
  • SQLite-backed settings — extension/server config persisted in app_settings.
  • Automatic provider failover — primary CLI → fallback CLI → heuristic recommendations; per-provider timeouts so a stuck CLI is abandoned.
  • Incremental batch analysis — service worker sends tabs in batches of 30 and updates the UI after each completed batch; fire-and-forget so 1000+ tabs don't hit the message-channel timeout.
  • Stop/resume with cancel — POST /analyze/cancel kills the CLI subprocess server-side, the extension persists partial state, the user can resume later. 8-second force-stop safety net in the UI.
  • Topic clusters — AI generates named clusters with descriptions and tags; collapsible UI; export to Obsidian; one-click create Chrome Tab Groups via chrome.tabs.group().
  • Tab Insights dashboard — top domains, average stats, snapshot trend.
  • LLM call logging — llm_call_logs SQLite table with provider, model, tokens, duration, status; viewer in Settings.
  • AI analytics record management — URL cache browser, analysis-sessions list, with delete actions.
  • Tab History panel — events tracked (opened / closed / activated), persisted to SQLite (with local fallback buffer), timeframe filters, search, sort, per-URL stats.
  • Cleanup Session — pre-cleanup auto-snapshot, step-by-step guided flow, summary screen, Export Cleanup Report to Obsidian.
  • Auto-snapshots — Chrome Alarms API, configurable 1-24h interval.
  • Content script (page-extractor.ts) — on-demand meta description / H1 / 500-char text excerpt; injected via chrome.scripting.executeScript.
  • Advanced Obsidian exports — TopicCluster → Topics/{slug}.md, TabSessionSnapshot → Sessions/{date}-{slug}.md, CleanupReview → Cleanups/{date}-cleanup.md.
  • Settings additions — local_server provider, server URL + Test Connection, primary/fallback CLI selection, Codex model + optional CLI path overrides, SQLite tools (refresh DB info, sync local data, clear DB, recent runtime logs), auto-snapshot toggle/interval, history retention input.

Changed

  • Manifest bumped to 0.2.0 with scripting permission and http://localhost/* host permission.
  • Vite build now has 4 entry points (added the content script).
  • 29 request types and 8 broadcast events; Zustand store gained 3 new slices (history, AI, cleanup).
  • 7 views in the side panel (added History, AI Recommendations, Cleanup Session).

0.1 — MVP

Added

  • Side panel as the primary UI; minimal popup that opens the side panel.
  • Tab list across all windows, grouped by window, with title / URL / domain / favicon / pinned / active / Chrome group name; search by title or domain.
  • Tab actions — close single, close selected (bulk), pin/unpin, mark important, restore recently closed via chrome.sessions.
  • Rule-based analysis — exact duplicate URL detection, near-duplicate (same domain + path similarity > 0.8), stale tabs (configurable threshold), domain grouping; badges on each tab.
  • Overview stats bar — total / duplicate / stale / important counts and top-5 domains.
  • Manual snapshots — create with optional name, snapshot history list, snapshot detail with restore-all/restore-selected/delete; max-snapshot cap with auto-pruning.
  • Obsidian export — LinkNote to TabOptimizer/Links/{domain}/{slug}.md with YAML frontmatter; vault path configurable; File System Access API; dedup guard against re-exporting the same URL.
  • Settings — Obsidian vault path, AI provider selection, AI mode (none / local_server), protected domains, stale-day threshold, max-snapshot count.
  • Service worker with message routing (11 request types in v0.1), tab event listeners (onCreated, onRemoved, onUpdated).
  • i18n — English + Russian translations.
  • Manifest V3 compliance.