Storage policies confuse agents because the table is storage.objects, not their table. The pattern: policies on storage.objects keyed by bucket_id, with auth.uid() matched against the owner or the…
Bright Data 200 can still be a stub - validate expected fields, not just the status code: Your Bright Data unlocker returning 200 does not mean it returned the data. Use when hitting this exact…
Bright Data CERTIFICATE_VERIFY_FAILED behind the proxy - trust the new CA and move to port 44445: If proxied requests through Bright Data fail with CERTIFICATE_VERIFY_FAILED, the zone is doing TLS…
Bright Data proxies need selenium-wire, not plain selenium - pass proxy auth via seleniumwire_options. Pass the proxy dict via seleniumwire_options at driver creation: http and https keys both set…
Guide Stripe checkout.session.expired: rebuild the session, do not retry it: Check # Stripe checkout.session.expired: rebuild the session, do not retry it ## The symptom Customers report an error on…
Browse.ai robots are trained by pointing in the browser - capture list, text, or screenshot then label fields. Browse.ai's whole model is train-by-pointing: install the Chrome extension, open the…
Guide R2 public access: custom domain vs r2.dev, and the default-deny: Do not build production traffic on it. ## The failure modes - Assuming a bucket is public because the Worker can read it:…
Browse.ai robots self-heal against layout changes and solve text captchas automatically. Two Browse.ai features that matter for long-lived scrapers. Use when hitting this exact issue. Not for…
Browse.ai Bulk Run handles 50,000 URLs at once - and the Monitor robot tracks page changes on schedule. Bulk Run: upload a CSV of URLs and run the robot across up to 50,000 pages at once - that is…
Fireworks rate limits are tokens per minute, adaptive, and 503 can hit you anyway Three things about Fireworks serverless rate limits that trip people: 1.
Guide ToolLoopAgent patterns: runtime context, tools context, and loop control: The default stopWhen is isStepCount(20); set it explicitly so the loop bound is visible. Use this when you hit exactly…
Browse.ai extracts behind login pages - train in your browser, export to Sheets, schedule the run. Browse.ai works behind login pages because training happens in your own browser - you log in, train…
Your Fireworks fine-tuned LoRA cannot run on serverless The most common Fireworks fine-tuning failure is assuming the trained model works like any other serverless model.
Browse.ai missing pages? Retrain the next-page action - pagination and infinite scroll are trainable steps. If your Browse.ai robot misses items on a paginated or infinite-scroll page, retrain the…
Guide Cloudflare rate limiting: start in Log mode, count on the right key - For origin protection, count only requests that reach the origin; for abuse prevention, count everything. Use this when…
Querying a Fireworks deployment uses a different model string, not a different endpoint On-demand deployments on Fireworks do not get their own endpoint.
Fireworks firectl will hang your CI with an interactive shape picker From Fireworks' own cookbook: running firectl deployment create with no shape flag does not fail cleanly.
BrowserCat API key goes in the Api-Key header (or apiKey query param) - not Authorization. If you connect with a raw websocket client that cannot set headers, pass it as the apiKey query parameter…
Guide Stripe Connect deauthorized: detect it and stop sending transfers: # Stripe Connect deauthorized: detect it and stop sending transfers ## The symptom Transfers or destination charges to a…