Frequently Asked Questions

PeelitNow executes a multi-pass DOM cleaning sequence. After retrieving the HTML payload via an encrypted connection, the engine runs an Abstract Syntax Tree (AST) parser that drops non-editorial nodes (<script>, <style>, <nav>, <iframe>, <aside>, tracking pixels, and modal dialogs). It then uses Mozilla Readability heuristics to measure comma frequency, paragraph density, and link-to-text ratios to extract clean editorial prose.

PDF documents can be submitted via direct public URL or uploaded through our pre-signed S3 pipeline. The current system limit is 15MB per file. Documents exceeding standard single-pass context windows are processed via a serverless Map-Reduce engine (unpdf) that chunks content into 8,000-token sections before generating a synthesized executive peel.

Yes. When a valid YouTube URL is submitted, the ingestion engine queries the video transcript via closed-caption tracks without executing headless browsers. The transcript is normalized, stripped of timestamp markers, and passed into our synthesis pipeline. Videos lacking transcripts or with disabled closed captions fall back to video metadata and descriptions.

PeelitNow honors publisher boundaries. If an upstream server returns HTTP 401 Unauthorized, HTTP 403 Forbidden, or hides content behind dynamic subscription authentication challenges that prevent anonymous extraction, PeelitNow returns an explicit extraction error rather than attempting unauthorized credential bypass.

No. All LLM inference executes across enterprise-tier AWS Bedrock foundation model containers (Amazon Nova Micro). Under Amazon Bedrock’s service-level security agreements, zero customer input payloads or generated summaries are retained or utilized for foundation model pre-training or fine-tuning.

Summaries are cached in Amazon DynamoDB using a deterministic SHA-256 hash of the canonical URL or PDF signature. Cached summaries persist for 24 hours to 30 days depending on document type and domain classification. This delivers sub-50ms repeat reads and prevents unnecessary network load on upstream servers.

Before PeelitNow opens an HTTP socket, target hostnames undergo synchronous DNS resolution. Any IP address matching private RFC 1918 allocations, local loopback (127.0.0.1), multicast, or cloud metadata endpoints (169.254.169.254) is blocked immediately with an HTTP 403 response. The scraper monitors all redirect hops with recursive re-validation.

Anonymous users are rate-limited by IP address via a sliding-window DynamoDB throttle. Standard anonymous quotas allow up to 10 peels per hour. This ceiling prevents denial-of-service abuse while allowing frictionless evaluation.

Registering an account via our Cognito identity provider provides higher monthly consumption quotas, access to your persistent personal Peel Archive (/my-peels), custom tag management, and options for ad-free reading.