What is copy-assets.mjs for

For WASM, instead of letting angular.json handle the copying to the public directory, it is done via scripts/copy-assets.mjs. This is due to a limitation of ng serve, which ignores angular.json asset copying.

When to bump service-worker.js version

• Updating vercel.json headers
• Updating/replacing any stable-named files: WASM binaries, worker scripts (e.g. tessworker.js, mupdf.js), or any asset not processed by Angular's build pipeline
• Update SSG pages

Note: Angular-compiled JS and CSS bundles use output hashing (main.abc123.js), so their filenames change automatically on every build — the SW will treat them as new files and fetch them fresh without needing a bump. The bump is needed for files whose names never change between deployments.

Why this matters

Everything requested over the network gets captured by the SW's runtime caching — on first request, the response is stored in cache and served from there on every subsequent request. If a stable-named file changes but the SW version isn't bumped, users will keep receiving the old cached version instead of fetching the updated one from the network.

Why updating vercel.json header requires bumping

• on first load all the static files are cached, including wasm, which comes from the network which includes CSP headers and so on. On next load, all will be served from cached which uses the same old header. To apply vercel.json changes, a SW bump is required

DEVELOPMENT

add NG_APP_ENV to your local env, value = development

PRODUCTION

add NG_APP_ENV to your env, value = production

SCRIPT TO CONVERT PADDLEPADDLE into ONNX

Cell 1 — Install

!pip install paddlepaddle paddlex -q

Cell 2 — Download PP-OCRv5 mobile models (paddle3.0 format)

!wget -q https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/PP-OCRv5_mobile_det_infer.tar !tar xf PP-OCRv5_mobile_det_infer.tar

!wget -q https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/PP-OCRv5_mobile_rec_infer.tar !tar xf PP-OCRv5_mobile_rec_infer.tar

Cell 3 — Install paddle2onnx plugin via PaddleX, then convert

!paddlex --install paddle2onnx

!paddlex --paddle2onnx --paddle_model_dir ./PP-OCRv5_mobile_det_infer --onnx_model_dir ./det_onnx !paddlex --paddle2onnx --paddle_model_dir ./PP-OCRv5_mobile_rec_infer --onnx_model_dir ./rec_onnx

Cell 4 — Verify outputs

import os print("det:", os.listdir("./det_onnx")) print("rec:", os.listdir("./rec_onnx"))