Ghost Kitchen & Virtual Brand Intelligence
Pricing
$0.10 / investigation bundle completed
Turn an authorized business-listing cohort into a traceable investigation queue with entity review, evidence coverage, missing-evidence actions, and durable outputs.
All notable changes to this actor are recorded here.
Use a resolved current-run price for bundle spending checks and result prices; reject unresolved tier maps. Preserve durable-output-before-charge ordering, one bundle event, and uncertain-settlement no-replay behavior.
Explain how to replace the fictional Console prefill or remove it before selecting a storage source. Keep retired-input migration separate from analysis.
Remove nested storage field-map prefills that created an incomplete second input route in Console. Omitted mappings retain their existing runtime defaults.
Hide retired boolean controls and accept their inactive false values
alongside current input: Console inserts these values even for hidden editors.
Other mixed legacy values still fail before effects; legacy-only input keeps
its no-charge migration behavior.
Bound shared input reads and reject malformed UTF-8 or types without echoing input values. Allow effective prices in the output schemas.
Patch Rustls to 0.23.45 for RUSTSEC-2026-0285 (TLS record-boundary validation).
Source candidate; hosted publication, health reevaluation and discount activation are recorded separately.
Supplied-intelligence production and commercial release contract. This version turns the accepted private output shape into the bounded production and settlement package. Hosted build, publication, recurrence, monitoring, and support remain independently verifiable operating surfaces.
Added the single investigation-bundle-completed PPE event, titled
Investigation bundle completed, at USD 0.10 per completed durable cohort
bundle with no platform-usage pass-through.
Made Dataset, report, compact OUTPUT, and charge-ready BUNDLE_STATE
acknowledgement precede the only charge attempt. Final BUNDLE_STATE now
distinguishes charged, already_charged, spending-limit, invalid-limit,
and uncertain_do_not_retry paths.
Added a stable opaque idempotency key, persisted only by SHA-256, and exact
micro-USD enforcement of ACTOR_MAX_TOTAL_CHARGE_USD. Incomplete,
ambiguous, invalid, budget-blocked, and legacy routes remain nonchargeable.
Sanitized charge HTTP failures to selected status/error-type metadata so raw provider response bodies cannot escape through runtime errors or proof.
Promoted the accepted source-neutral intelligence core into the normal Actor entry while retaining the prototype driver as a development aid.
Added ghost-intelligence-input-v1 with exact analyze_supplied mode,
declared market context, exact production rights values, 1-3,650 day
freshness horizon, and one-of inline or bounded storage input.
Added selected-field CSV, JSONL, JSON-array, Dataset, KVS, and Apify file-upload/KVS resolution. The candidate caps accepted rows at 250, KVS/file bodies at 8 MB, Dataset pages at 100 rows and 2 MB, and fails over-limit input without truncation.
Derived analysis identity, retrieval/analysis clocks, rule/source-plan
revisions, selected-input SHA-256, and body-free source-reference SHA-256 in
the Actor. Buyer observedAt remains nullable source evidence.
Added ghost-investigation-action-v1 Dataset items,
ghost-intelligence-report-v1 KVS report,
ghost-intelligence-output-v1 compact OUTPUT, and
ghost-useful-bundle-state-v1 acknowledgement/recovery state.
Made the complete durable action/report/OUTPUT/state bundle the only
provisional completed_useful_investigation_bundle_v1 unit. Sparse, stale,
conflicting, collect-more-data, and no-signal conclusions remain eligible
only when they retain an honest next action and every required artifact is
acknowledged.
Reused the shared no-blind-retry Dataset transport. Unknown commit is never replayed or made eligible and carries exact GET-first recovery guidance; later required-write failures retain partial state when possible.
Initially fixed production billing at pricing_not_effective, left event
slug and price absent, stored only the future idempotency-key hash,
classified the maximum-total-charge setting without raw-value retention,
and made zero custom charge calls. The commercial additions above supersede
that candidate-only posture while preserving its no-charge failure routes.
Quarantined every retired input before old provider/dependency construction.
Legacy-only requests now receive a structured no-charge migration result;
mixed legacy/new input fails before any effect, and demoMode can no longer
activate the fictional demo from the normal entry.
Replaced retired Actor input/output/storage definitions, smoke input, and buyer documentation with the supplied-intelligence production contract.
Kept optional storage-source field-map hints UI-only with prefill while the
runtime owns their exact defaults. This prevents an omitted source object
from being materialized during live validation and blocking inline records.
Deprecated-product retirement package.
ghost-kitchen-availability-v1 result: empty Dataset, JSON OUTPUT, and
bounded OUTPUT.html, published through the existing durable artifact
coordinator.liveCollectionAvailable=false, unavailable run mode, zero results, zero
custom charge attempts, no charged event, and an optional fictional-demo
next action.demoMode=false as the honest default, and retained the
explicit deterministic fictional demo unchanged and nonchargeable.Unreleased local candidate. This source package has not yet completed the governed GitHub/Apify build, run, storage, Store, or rendered release loop.
demoMode and three deterministic demo-v1 scenarios:
shared_kitchen_cluster, quiet_market, and partial_coverage. Demo work is
fictional, source-free, enrichment-free, and charge-free while still
publishing dataset rows, JSON OUTPUT, and OUTPUT.html to platform
storage.ghost-kitchen-batch-output-v2 identity with top-level and
per-location runMode, reachability-based sourceCoverage, bounded
addressClusters, and coverage-gated qualitative marketScorecard products.
Existing live input keys and legacy dataset/OUTPUT field names, JSON types,
and meanings remain compatible.complete, partial, limited,
unavailable, and not_started. Reached-empty sources remain distinct from
failed or unstarted sources, and result count is not used as a coverage
proxy.concentrated_candidates,
candidate_presence, quiet_or_inconclusive, and
insufficient_coverage. Limited, unavailable, and not-started coverage
overrides candidate and cluster counts.OUTPUT.html with
text/html; charset=utf-8.BATCH_STATE and LOCATION-#### so buyers and agents can discover
the report, market evidence, and operational retry/charge checkpoints without
confusing checkpoint state with settled platform counters.OUTPUT -> raw
OUTPUT.html coordinator. Live charge eligibility begins only after the
complete bundle is durable; selected write failures preserve truthful
partial-artifact state and make no charge attempt.locations[], deterministic
per-request rows, duplicate aliases, spend-limit skips, source-health
diagnostics, output-safe checkpoints, serialized charges, and
batchConcurrency from 1 to 3.$0.10 per eligible
location-analyzed event. Demo scenarios, duplicate aliases, spend-limit
skips, all-source blackouts, pre-output failures, and incomplete artifact
bundles are not separate charged units.groups-auto,session-... form.OUTPUT failure summary with scraper diagnostics before
exit when storage is available.radiusMiles and maxRestaurants materially affect candidate collection and analysis selection.OUTPUT summary.google* to external*.ActorRuntime for graceful failure reporting.LazyLock RegexSet matching for virtual brand patterns.