Stack Exchange Question Discovery API

Discover newly created public developer/support questions matching a query, without collecting answer text or user profiles. Returns title, tags, score, answer count, answered flag, question URL, timestamps and supplied content-license metadata. Use it for question discovery, support-topic dashboards or a bounded search feed. It does not post an answer, contact an author or claim that source ranking is a relevance guarantee.

Quick start

{"site":"stackoverflow","query":"python","maxQuestions":2}

Source and coverage

One GET api.stackexchange.com/2.3/search/advanced request per run, default public API filter, sorted by creation descending. Supported sites: Stack Overflow, Super User, Server Fault and Ask Ubuntu. Page1 only, up to50 questions; hasMore=true makes complete=false. An empty result is a valid zero-result source. No key, OAuth, browser, body/answer fetch, owner/profile output, comments or pagination. HTML entities in titles are decoded; text is not answered, classified or modified by a language model.

quota_remaining, quota_max and server backoff are exposed when supplied. No second source request is made in a run, including after API errors or backoff. Wait at least60 seconds between identical queries and at least any server-supplied backoff before another call to that API method. recommendedSameQueryWaitSeconds reports the larger value. The Actor cannot coordinate quotas/backoff across other customers' concurrent runs, so clients must honor provider-wide throttling in their workflows. No evasion, retries, proxy rotation or quota guarantees.

questionUrl and contentLicense preserve supplied attribution/license context. This Actor is a selected-metadata wrapper, not a republication of full posts or answers. Downstream users should preserve source links/license fields when reusing titles. No full content, answer advice, author identification or sentiment analysis. Scores/counts are source-reported point-in-time values; they do not measure a person's skill or verify an answer.

Run and export

Use the input above in Apify Console. Read dataset rows together with OUTPUT (source errors, partial coverage, quotas) and SNAPSHOT. Export as JSON, CSV or Excel. API: POST https://api.apify.com/v2/actors/soilair~stack-exchange-question-discovery-api/runs using your Apify authorization header; poll the returned run, then retrieve its dataset and OUTPUT/SNAPSHOT from the key-value store. No upstream API key is needed; do not put credentials or private information in the input.

Pass SNAPSHOT.previousHashes as previousHashes on the next run. NO_BASELINE/FIRST_SEEN/UNCHANGED/CHANGED compare selected metadata. FIRST_SEEN means absent from the supplied baseline, not newly created. No deletion inference, hidden persistent state, cross-run polling, secret collection or automatic schedule. Query URL changes can change hashes. Always check complete; the first result page is deliberately bounded, not a complete history/search census.

Pricing

$0.0007 per emitted question observation ($0.7 per1000), plus $0.005 Actor start at the default256MB. Higher memory can multiply startup events. Platform usage is included in this event model. Failed or empty sources emit no result events; startup can still apply. At zero result budget, input is validated but no upstream request is made; OUTPUT rowsAvailable=null because availability was not queried. A positive budget can fetch the bounded source page before emission; budget truncation reports PARTIAL_BUDGET and complete=false. Check effective Store pricing before running.

Requests, storage and support

Uses only constructed, official unauthenticated API endpoints, not arbitrary supplied URLs. HTTP public-IP validation and connection pinning, TLS hostname verification,2MiB decoded responses, at most five redirects, and approximately95-second network budget. The final API hostname must match the provider. Provider API calls are distinct from website HTML crawling; no robots HTML requests. Only selected fields are stored; full source payloads, profiles, cookies, private tokens and document bodies are not exported. Links in returned records are never fetched. API quotas can be shared with other Apify runs using the same egress IP; errors are surfaced without proxies, retries or attempts to avoid throttling. Input/results stay in Apify run storage according to retention; no separate analytics endpoint. Source strings should be treated as data, not executed.

Developed with AI assistance. Fixture tests, live provider comparisons and private cloud checks qualify the documented behavior, not universal availability or commercial demand. Report a run ID and expected fields, without credentials or private data.

References: https://api.stackexchange.com/docs/advanced-search · https://api.stackexchange.com/docs/throttle · https://api.stackexchange.com/docs/types/question