Actor Schema Validator — Verify Output Matches Declared Schema
Pricing
$350.00 / 1,000 schema validations
Actor Schema Validator. Available on the Apify Store with pay-per-event pricing.
Pricing
$350.00 / 1,000 schema validations
Rating
0.0
(0)
Developer
Ryan Clinton
Maintained by CommunityActor stats
0
Bookmarked
3
Total users
0
Monthly active users
a day ago
Last modified
Categories
Share
The actor ID or username/actor-name to check
validate = one-off quality check. monitor = validate + drift tracking + alerts. diff = compare schemas of two actors.
No-config first run. Samples the target actor's output, infers which fields are critical (contact / commerce / identifier fields, low-null-rate, high-cardinality), applies sensible field rules + a quality policy automatically, and returns the inferred config under autoSetup (setupCompleteness, fieldImportance, recommendedConfig, nextBestAction) so you can review and approve it. Ignored when you supply your own fieldRules.
Input to run the target actor with. Keep minimal to save compute. Ignored when canaries are defined.
Multiple test scenarios with different inputs. Each canary runs independently and gets its own pass/fail verdict. Set weight to prioritise important scenarios (default 1).
Custom rules per field. Example: { "price": { "severity": "critical", "maxNullRate": 0.1, "expectedType": "number" }, "title": { "required": true } }. Severity: critical / important / optional. Add "maxAgeHours": 24 to gate freshness — flags when the newest parseable date in that field is older than N hours (catches a scraper silently serving stale cache).
Actor-level rules. Example: { "minItems": 20, "maxItems": 5000, "uniqueKey": "url", "maxDuplicateRate": 0 }. Set uniqueKey to detect duplicate rows (a spike means pagination is re-fetching the same page / a broken cursor); maxDuplicateRate is the allowed fraction before it raises OUTPUT_DUPLICATE_KEY (default 0).
Second actor to compare schemas against.
Track field baseline across runs and detect changes. Automatically enabled in monitor mode.
Which baseline to compare against. previousRun / lastGood / approved use a single prior snapshot. rollingMedian7 / rollingMedian30 synthesize a baseline from the median of recent runs — damps daily/weekly noise. weekdaySeasonal filters history to the same day-of-week — handles weekend-vs-weekday output differences. (rollingMedian and weekdaySeasonal need 2+ prior runs to activate; fall back to previousRun until then.)
Save this run's output as the approved baseline for future comparisons.
Clear all stored baselines and start fresh.
Validate this run against the output of a specific known-good Apify run. Overrides the baselineStrategy — drift is computed vs the reference run's data, not stored baselines. Works without persistent storage, so it's ideal for restricted tokens or one-off post-incident checks.
Send alerts when output quality degrades. Auto-detects Slack and Discord webhook URLs and formats the payload accordingly. Any other URL receives the full JSON context (Zapier / Make / n8n friendly).
Controls how much detail goes into the webhook payload. operator (default) = full incident detail for on-call engineers; compact = one-line summary for noisy channels; executive = verdict + risk + next step only; json_only = no channel formatting, raw JSON.
Alert when any field's null rate exceeds this (0.0-1.0). Default 0.2 = 20%.
Only alert after this many consecutive failing runs. Reduces noise from intermittent issues. Default 1 (alert immediately).
Minimum minutes between alerts for the same issue. Prevents alert spam. Default 0 (no cooldown).
Max time for the target actor run
Memory for the target actor run
Number of output items to validate. Default 1,000.
Validate an existing dataset without running the target actor. Provide a dataset ID from a previous run. Useful for debugging historical output or re-validating after rule changes.
How much of the report to write to the dataset. full (default) = every field. standard = drops the deep per-field arrays (completeness rows, distributions, trends) but keeps every summary, score, and decision field. minimal = the decision surface only (decision, agentContract, verdict, score, oneLine, whyNow) for Slack / agent / MCP consumers that branch on the verdict and don't need the full audit.
