Anki Terminator Companion

An elegant, dynamic performance-optimization add-on for Anki.

This companion addon is designed to run seamlessly alongside the original Anki Terminator V2 - ChatGPT DeepSeek Sidebar for Reviewer (1468920185). By dynamically patching memory during Anki startup, it eliminates high-CPU bottlenecks completely without modifying a single line of code in the original addon's directory. <img width="918" height="113" alt="Screenshot_20260614_133846" src="https://github.com/user-attachments/assets/e7b9fa55-1d3b-461a-8646-f1529605be45" />

Key Features & Optimization Highlights

⚡ 1. Smart UI-Freezing with Active Hover Tracking (0% Idle CPU)

• The Problem: Chromium (QtWebEngineProcess) refuses to freeze a webpage if it's visible on the viewport, meaning Gemini/ChatGPT continue to consume high CPU even when you aren't looking at them.
• The Solution:
  • When the sidebar loses focus or you are studying cards, the companion captures a pixel-perfect static screenshot of the webview (webview.grab()) and displays it instantly in a QLabel via a QStackedWidget layout swap.
  • While hidden behind the static view, Chromium completely suspends all paints and JavaScript loops, instantly dropping CPU usage to 0%.
  • Flicker-Free Transitions: By utilizing a QStackedWidget, page swaps are completely imperceptible and flicker-free.
  • Active Hover & Stream Protection: The sidebar stays 100% active and responsive under your cursor. In addition, when Gemini is actively generating a response, freezing is automatically deferred until the response finishes streaming, allowing you to watch the answer populate in real-time.

🚫 2. Optimized O(1) Suffix Ad-Blocker Lookup

• The Problem: If the C-based Rust ad-blocker engine is unavailable, the original addon falls back to scanning a list of ~46,000 domains using slow, linear substring matching (domain in url) for every single network request. This blocks the main Qt UI thread.
• The Solution: The companion parses and splits those rules into pure domain lookups, executing a set-based suffix-matching algorithm in O(1) time. This eliminates frame drops during page loads.

🎨 3. CSS Animation & Transitions Disabler

• The Problem: High-CPU draw on Gemini's website due to continuous background shimmer animations, blur filters, and glow gradients.
• The Solution: Dynamically injects highly optimized CSS when the page loads to disable background animations, transition effects, and intensive CSS blur/backdrop-filters globally.

📝 4. Rich HTML Context Menu Paste Support (Preserves Formatting)

• The Problem: Right-clicking in the sidebar to "Add selection to field" originally uses plain-text copy-pasting, stripping out all markdown, links, lists, bold text, italics, and image elements.
• The Solution: Dynamically intercepts the context menu triggers to extract selection text as raw HTML using a specialized JavaScript range cloner. This allows you to append formatted answers directly into your Anki fields with all formatting and styling intact.

📋 5. Clipboard-Free Text Injection (No Clipboard Pollution)

• The Problem: Original prompt inputs relied on system clipboard copy-paste actions with timers. Lag spikes often caused the webview to paste the user's restored original clipboard data instead of the prompt text, polluting their clipboard history.
• The Solution: Patches the input insertion mechanism to inject queries directly via Chromium's native document.execCommand('insertText') API. It leaves your system clipboard completely untouched and guarantees instant, race-condition-free pasting across all AI platforms.

🧹 6. HTML Stripping for AI Prompts (LaTeX Clean)

• The Problem: Card fields sent to the AI often contain extensive HTML layout tag boilerplate (<br>, <div>, <span style="...">), which wastes token limits and confuses LLMs.
• The Solution: Automatically parses and strips all raw HTML tags from prompts, preserving clean text and raw LaTeX/MathJax expressions (\(...\) and \[...\]) for optimal AI processing.

⚡ 7. Instant Config Access (Header Button)

• The Problem: To change companion settings, users had to navigate through the Add-on Manager, which is slow during active study sessions.
• The Solution: Injects a custom "C" button directly into the sidebar header toolbar (next to the settings cogwheel). Clicking it instantly opens the Terminator Companion settings dialog.

➕ 8. Add to New Card (Context Menu)

• The Problem: Saving AI-generated content into new cards required manual copying, switching to the Add Cards window, and pasting.
• The Solution: Adds a new "Add to new card" option in the sidebar's right-click context menu. It extracts the selected text or images (preserving formatting and math) and pre-populates them into the first field of a new card window instantly.

🗃️ 9. Multiple Fields Support

• The Problem: Original sidebar only reads the single designated priority field.
• The Solution: Allows concatenating and sending all non-empty card fields to the AI, formatted clearly as FieldName: FieldValue. It automatically filters out empty fields and fields containing only boilerplate spaces or tags.

🌐 10. Integrated Browser Navigation & Search

• The Problem: The sidebar was locked to the AI chatbot URL, making it difficult to quickly look up external references or navigate between pages.
• The Solution:
  • Minimal Address Bar: Direct URL entry and keyword search support.
  • Navigation Controls: Back, Forward, and Reload buttons integrated into the header.
  • Custom Search Engines: Configurable search providers (Google, DuckDuckGo, Bing, or Custom).
  • Sleek Progress Bar: A browser-style progress indicator at the top of the sidebar replaces intrusive loading overlays.
  • Persistent View: The current page remains visible while the next one loads for a smoother browsing experience.

🪟 13. Popup Window Navigation Controls

• The Problem: Links opened in new popup windows had no navigation controls — no back, forward, reload, or address bar.
• The Solution: Injects the same navigation toolbar (back, forward, reload, home, address bar) into every popup window, matching the sidebar's style and search engine configuration.

⚙️ 14. Non-Modal Config Dialog & Lazy Logs

• The Problem: Opening the config dialog blocked all Anki interaction. The logs tab could freeze with large log files.
• The Solution:
  • Config dialog is now non-modal — use Anki freely while it's open.
  • Logs tab lazy-loads only the last 200 lines on open.
  • Refresh button for manual log reload.
  • Auto-scroll only when you're already at the bottom.
  • Monospace font and fixed-width for better readability.

🤖 11. Header AI Dropdown Selector (Fast Switching & Dynamic Favicon)

• The Problem: Switching between different AI providers (ChatGPT, DeepSeek, Claude, Gemini, etc.) was slow and required clicking the button repeatedly to rotate through them one by one.
• The Solution:
  • Replaces the static "AI" header button with a dropdown selector.
  • Displays full AI names and logos in the dropdown menu for fast, direct selection.
  • Automatically updates the sidebar's window title bar and tab icon (favicon) to the active AI's logo.
  • Cache-Bypass Hard Refresh: Selecting the already active AI triggers a cache-bypassing hard reload and clears webview history to start a fresh chat session.

🧩 12. Censored "Explain Cloze" Context Menu Action

• The Problem: Selecting card text containing active Cloze deletions (e.g. {{c1::answer}}) and sending it to the AI often reveals the correct answer directly within the prompt, prompting the AI to simply repeat it rather than explaining the concept.
• The Solution: Introduces an "Explain cloze with AnkiTerminator" action to the right-click selection context menu. It automatically censors all cloze deletion segments, replacing them with [...] before passing the text to the AI so the LLM explains or solves the blank without being biased or influenced by the answer.

Configuration & Settings

The companion add-on introduces a dedicated settings dialog inside Anki's Add-on Manager (accessible via Tools > Add-ons > Select Anki Terminator Companion > Config).

Tabs Overview:

1. General Settings:
   • Enable Smart CPU Freezing (Lifecycle State): Toggles the dynamic freezing/thawing system for Gemini/ChatGPT.
   • Enable Ad-Blocker O(1) Suffix Match Optimization: Toggles the fast suffix domain set lookup algorithm.
   • Inject CSS Gemini Animation Disabler: Toggles the custom animation disabler styles.
   • Enable HTML Stripping for AI Prompts: Toggles stripping of HTML tags from prompts (preserves math).
   • Enable AI-Hints O(n) Regex Bypass Optimization: Toggles fast-path rendering when no AI hints are present.
   • Enable AI-Hints Context Menu Preservation: Toggles safety check when appending right-click selections.
   • Send Multiple Fields to AI: Toggles sending all non-empty fields instead of just the priority field (configurable in both companion config and original Terminator "Fields" tab).
   • Thaw Duration after Query: Configures the duration in seconds that Gemini stays Active before freezing back after clicking any prompt buttons.
2. Performance Logs:
   • Lazy-loads the last 200 lines on open for fast startup.
   • Displays thread-safe, real-time diagnostic and performance events from the companion.
   • Includes Refresh, Copy Logs, and Clear Logs buttons.
   • Auto-scrolls only when you're already at the bottom.
3. Support:
   • Offers direct links and QR codes to support the creator (UPI, BTC, ETH, and Ko-fi links).
   • Includes a checkbox option: "I have supported this addon (Hide automatic update welcome)" to disable the welcome screen popping up after future updates.

Documentation