All notable changes to this actor are recorded here.
-
Reordered the README around the start path, four workflow files, human review
boundary, inspect-first recovery, and migration so buyers can reach the
decision path before implementation detail.
-
Added a dated point-in-time Store price/unit statement, explicit
no-maximum-invoice boundary, and a privacy-safe support diagnostic checklist
without claiming response time, delivery, or an SLA.
-
Aligned Cargo 2.0.0, Actor version 2.0, remote Actor-version request
identity 2.0, and contact-cleanup-output-v3.
-
Made the v3 whole-file workflow the sole production entry in this source
package. The retained v2 behavior is hidden behind an explicit historical
compatibility test/fixture harness and is not a buyer runtime selector.
-
Reframed the package around four discoverable buyer files:
CLEANED_CONTACT_REVIEW.csv|json, RECORDS_TO_REVIEW.json,
POSSIBLE_MATCH_GROUPS.json, and RUN_SUMMARY.json.
-
Added a step-by-step route, field-mapping, preview, export-mode,
export-format, review, reimport, and recovery guide without implying
verification, deliverability, automatic merge, or survivor selection.
-
Froze documentation examples from the committed 12-row smoke input through
the maintained whole-file runtime rather than maintaining independent
hand-authored output claims.
-
Corrected the supported buyer ceiling to the selected machine-qualified
separate profiles of 1,000 accepted rows at 6 source fields and 261 source
fields at 12 rows; their cross-product was not qualified. The retained 10,000-row
parser/runtime guard is not marketed as supported: that qualification
candidate failed before Dataset publication at the single-record KVS
boundary.
-
Documented exact same-format preservation and replay boundaries, including
compact JSON-text encoding for JSON-origin values in CSV cells and native
source value types in JSON exports.
-
Added GET-first recovery guidance for deterministic rejection, unknown
Dataset commit, final OUTPUT uncertainty, and hard process death. No
caught-failure state authorizes blind replay.
-
Recorded the selected local configured estimate of $0.100000 for 1,000
synthetic Dataset-item event units and kept it explicitly separate from
current live pricing, a bill, or settled charged-event proof. The separately
reserved $0.125000 later-test cap candidate is not a price or run
authorization.
-
Kept deployment, Store availability, pricing, and rendered Console state out
of this source-package history; those remain point-in-time Apify
observations.
