# Changelog of Answer Drift Planner: Knowledge Base Change Review (`critd/answer-drift-planner`) Actor

- **URL**: https://apify.com/critd/answer-drift-planner/changelog.md
- **Full Actor documentation**: https://apify.com/critd/answer-drift-planner.md

## Answer Drift Planner release notes

Customer-facing changes and compatibility notes. Knowledge sources here mean reference material such as policies and help articles, not software source code. The README owns current preparation, pricing, limits and recovery guidance. Preserve original files before changing a saved workflow.

### Clearer purpose and setup guidance

The listing explains that this is a knowledge-base change review, not source-code review. It uses your prepared inventories, explicit answer/test relationships and supplied retest results. It does not inspect document meaning, assess answer quality or discover missing dependencies.

Inputs, review rules and prices are unchanged. The ordinary and returned-results examples remain available. A later run reviewing returned results is still another paid report at the normal price.

### Retest evidence review in the 0.2 family

**See which retest results belong to this change.**

Bring back the results from the tests you chose to run. Match them to the source changes and test versions under review, and see missing results, older results and disagreements without losing the original review questions.

This compares the records you supply. It does not run a model, verify how a test was performed, prove an answer correct or approve a release.

The optional review uses explicit test-definition revisions and per-test dependency-context fingerprints. It supplies a separate request template for affected declared tests and compares returned receipts without removing original source-change, coverage or mapping findings. Stale, missing and conflicting evidence remains visible. Existing inputs and behavior are preserved when the optional review is omitted.

Up to 1,000 receipts share the existing input and output limits; the request template shares the output budget. A valid unresolved review remains billable, while malformed or oversized preparation does not initiate the completed-report charge. Startup charges can still apply. Unknown effects must be inspected before another run.

**A later run reviewing returned results is another paid report at the normal price**, not a free continuation of the first run. The completed-report and startup prices are unchanged; no per-test or per-receipt charge is added.

### October 8, 2026 — maintenance update in the 0.2 family

The refreshed build retains source-change review while consolidating the underlying file-reading implementation. A published fictional example provides a direct starting point without a model account or private-system connection.

The example uses normal pricing: $0.25 per completed report plus $0.00005 per start. Existing outputs, limits, explicit relationships and prices are retained. Component-level improvements are not a promise of faster hosted runs or measured customer time savings.

### 0.2.1 — prepared-inventory review on Apify

**Know what needs another look when your sources change.** Compare earlier and later inventories with supplied item and dependency records. A readable report identifies revalidation work, missing source coverage and missing links without dropping declared or referenced items.

Prepared CSV and strict JSON routes are available. Completed reports are published before the report charge is requested; a valid report consisting entirely of unresolved questions is still billable. Invalid preparation does not request that completed-report charge, although startup charges may apply.

The Actor does not connect to private systems, invent dependencies, call a model, execute tests or judge answer truth. No unchanged result is permission to skip a test.

### 0.2.0 — guided CSV preparation

Explicit column mapping and separate capture information make supported inventories, item tables and link tables usable without hand-writing nested dependency objects. Missing or invalid preparation receives bounded correction guidance where supported. Strict JSON and the existing uncertainty rules remain available.