OLmatter/glm-coding-helper

What it is

glm‑coding‑helper is a small open‑source toolkit that helps users snap up Zhipu GLM Coding Plan subscriptions the moment they become available. It combines a Tampermonkey userscript that runs in Chrome/Edge with a local OCR backend (CPU‑ or GPU‑accelerated) that automatically reads and clicks the Chinese point‑selection captchas shown on the purchase page.

Why it exists

Zhipu’s GLM Coding Plan is sold in limited‑time “rush” batches that sell out within seconds. Manual refreshing, copying invitation codes and solving the visual captcha is error‑prone and often too slow. This project automates the repetitive steps while keeping the captcha solving local (no third‑party image upload), thereby speeding up the purchase flow without violating the service’s terms.

Core features

Feature	What it does
Purchase‑flow helper	Enables the “Subscribe” button early, cycles through configured plans and periods, and retries on rate‑limit errors.
Local OCR backend	Runs a tiny YOLO detector + PP‑OCRv6 model (tiny + medium fallback) on the user’s CPU or GPU to recognise the three Chinese words in the captcha and returns click coordinates.
One‑click multi‑window	Opens several browser windows at once so you can pre‑warm the page and monitor several attempts simultaneously.
Configurable automation	Toggle automatic clicking of the subscribe button, the captcha words, and the final “Confirm” button; adjust random delays to avoid anti‑bot throttling.
Cross‑platform starter	Portable Windows zip (pre‑bundled model) or online installer for macOS/Linux; a single one‑click‑start.* script sets up a Python virtual‑env and launches the OCR server.
Tk GUI monitor (Windows/macOS)	Shows OCR status, confidence scores and logs in real time.
Safety prompts	By default the script only highlights the captcha; you must enable the final confirm click manually, reducing the chance of accidental purchases.

How to get started (quick‑start)

1. Download a release – either the portable‑cpu‑*.zip (Windows) or the online‑installer‑*.zip (macOS/Linux).
2. Extract to a short, ASCII‑only path (e.g. C:\glm-coding-helper).
3. Install the userscript – either via the GreasyFork link or by copying glm-coding-helper.user.js into a new Tampermonkey script.
4. Start the OCR backend – double‑click one-click-start.cmd (Windows), one-click-start.command (macOS) or run ./one-click-start.sh (Linux). The script will automatically create a virtual environment, install the required Python packages and launch a local server at http://127.0.0.1:8888.
5. Open the purchase page – https://www.bigmodel.cn/glm-coding and verify that the Tampermonkey icon shows the script is active.
6. Configure – open the Tampermonkey menu → Open configuration panel to set plan priority, enable auto‑clicks, adjust captcha‑click delays, etc.
7. Rush – at the announced release time (e.g., 10:00 AM) let the script run; it will automatically solve captchas and click the subscribe button if you have enabled “Rush mode”.

Technical details (for the curious)

• OCR pipeline: Captcha image → YOLO‑v5‑style detector (locates the three word boxes) → PP‑OCRv6‑tiny model (fast, ~30 ms per image) → optional fallback to PP‑OCRv6‑medium. Accuracy on a hidden test set is 100 %.
• Backend: FastAPI server (captcha_server.py) exposing /captcha_direct. A small Tk window displays live status; a headless variant is used on Linux.
• GPU support: If a CUDA‑11.8‑compatible GPU is present, the installer creates a .venv_paddle_gpu environment with paddlepaddle‑gpu and a GPU‑enabled torch. RTX 50‑series cards are not yet supported (CUDA 12.8 required).
• Safety: Captcha images never leave the machine; all processing is done locally. The script respects browser privacy settings (LNA flag handling, incognito mode support).

Who might use it

• Developers or hobbyists who need a GLM Coding Plan for experimentation and want to maximise their chance of obtaining a limited‑time subscription.
• Anyone comfortable running a small Python environment locally and using Tampermonkey.
• Users who prefer a self‑hosted captcha solver rather than a cloud OCR service.

---

TL;DR: glm‑coding‑helper is a Tampermonkey‑based auto‑purchase assistant for Zhipu’s GLM Coding Plan, paired with a fast local OCR server that solves the Chinese point‑selection captcha. Install the userscript, run the one‑click starter, configure your preferences, and let the tool handle the frantic rush for you.