Hacker News Story & Comment Tree Export

Export exact Hacker News discussions or ranked Top/Show/Ask/New/Best/Job feed threads with their bounded authored reply trees. Use parent IDs, reply depth, sibling order and source timestamps to analyze developer feedback without matching a story by a loose search phrase.

Input

Default: Top and Show feeds, up to 5 valid threads, 50 candidate IDs per feed, 500 comment nodes per thread, depth20 and 800 total requests. Empty feeds plus storyIds/storyUrls select exact discussions. Global caps and interleaving preserve source coverage. This Actor does not provide Algolia keyword history, user profiles, external article crawling or alert delivery.

Output

One story_thread row contains story identity/title/date, observed score and the flattened ordered comment tree. Each comment identifies its parent, depth and siblingIndex. Deleted/dead nodes can remain as hierarchy anchors without text. Unknown optional facts are omitted and reported in availability. coverage states complete/incomplete, node limits, pending children and source observation. An evolving discussion is not an atomic snapshot.

Price

$0.004 per unique delivered thread includes its bounded reply tree. There is no start or per-comment add-on. At least one nonempty authored readable comment, source story identity/title/date and valid parent relationships are required. Empty, dead, metadata-only, duplicate, excluded and failed source results are not charged. Default cap$0.02; maximum10thread cap$0.04. Source user-selected limits can deliver a clearly marked partial tree, while an upstream error fails before delivery.

Failure and empty results

All source/auth/robots failures fail the run;429 is rate-limit, invalid item/tree structure is structure-change, timeout remains timeout. No automatic retries: transient responses and timeouts fail before output; no follow-up request after429. Missing/deleted roots are skipped, but a successful run with no valid thread is an explicit no-matching-results failure. No empty success or billed metadata summary.

Privacy and source boundaries

Only official public Firebase JSON endpoints are accessed; robots is checked first and redirects are restricted to that exact HTTPS host. No API key, login, cookies or private endpoint input. Author handles, avatars, emails, phones and profile data are excluded. Contact strings embedded in text are redacted; incidental self-disclosure may remain, so output is not claimed perfectly anonymous. Fixtures use synthetic text. No external article/user links are followed.

Source and reliability

Uses the official Hacker News API. Source terms include restrictions on commercial copying and automated extraction, recorded in the current scout underDR-0007; native API availability is not a resale license. Data may change during collection. Cloud cost, row/event PPE parity and final independent review are release gates, not claims based on local fixtures.

Limits

Hard maximum1000requests,10threads,500comment nodes per thread anddepth20. Acquisition is bounded to240s within a300s Actor timeout at256MB. At most5 requests run concurrently with at least0.1s between starts, applicable robots delays and30s HTTP timeout. Stream limits are2MiB wire/decoded per response,32MiB cumulative wire/decoded and16MiB staged output. Size breaches fail before delivery. A declared acquisition deadline returns honest pending-node coverage; upstream timeout remains failure. All selected feeds/root candidates are discovered before round-robin reply expansion. The entire positive output is staged and checked against the event allowance before SDK charged delivery. Default5threads/800requests can be incomplete for large discussions; coverage never claims full history when a bound is reached.