feat(cli): add batch URL input support

[DyNooob]

Summary

Adds support for batch URL input in the PixelRAG CLI.

Users can now provide either a single URL or a .txt file containing multiple URLs (one per line). Each URL is processed using the existing rendering pipeline.

Changes

• Added support for .txt file input in CLI
• Each non-empty line is treated as a URL
• Single URL usage remains unchanged
• Batch URLs are processed sequentially using existing render_url function
• Failed URLs are skipped without stopping execution

Usage

Single URL

pixelshot https://example.com --output ./tiles

Batch input

pixelshot urls.txt --output ./tiles

Example urls.txt

https://example.com/a
https://example.com/b
https://example.com/c

Notes

• CLI-only change
• No changes to rendering / embedding / indexing logic
• No new dependencies introduced
• Fully backward compatible

Error handling

• Invalid or failing URLs are skipped
• Batch processing continues normally

Impact

This improves usability for processing multiple pages without changing the core system design.

Risk

Low. Isolated CLI enhancement with no core pipeline modifications.

[vercel]

@DyNooob is attempting to deploy a commit to the andylizf's projects Team on Vercel.

A member of the Team first needs to authorize it.

[andylizf]

Thanks for adding .txt batch input support — the file-reading and error-handling per URL are good additions.

However, this PR has a silent performance regression: the old code called render_urls(urls, ...) which dispatches all URLs across --workers parallel browser processes. The new code loops for url in urls: render_url(url, ...) — making rendering serial across URLs. The --workers flag still exists but is now effectively a no-op on the CLI path, so a user with a 100-URL text file would see ~4x slower throughput without knowing why.

Could you keep the batched render_urls call (which parallelizes across workers) and add the .txt expansion + per-URL error handling around it? Something like:

# expand .txt inputs into URLs
urls = []
for inp in args.inputs:
    if inp.lower().endswith(".txt"):
        urls.extend(line.strip() for line in open(inp) if line.strip())
    else:
        urls.append(inp)

# still use the batched parallel path
render_urls(urls, args.output, workers=args.workers, ...)

Also, ruff format is needed — CI's ruff format --check fails on the new test.

Minor: the bare open(inp) for .txt files should be guarded (FileNotFoundError → clean error message).

[DyNooob]

Good catch.
I restored the batched render_urls(...) path so--workerscontinues to parallelize URL rendering, and kept.txtinput as a URL expansion step before the batch call.
I also added a clean FileNotFoundError parser error for missing .txt files and ran ruff format; the new tests now assert that .txt input uses render_urls with the requested workers