Changelog

v1.23.0

Creator acquisition starts with your business

Turn your website into a focused creator brief. Console now helps you define which creators fit your business before Caivex prepares your workspace.

  • Explore four distinct creator profiles, each explaining how they could work with your business and which audiences they reach.
  • Add your brand’s Instagram - or a competitor’s - alongside a creator you’d love to work with. These references give us concrete examples to guide targeting.
  • Refine your brief and submit it directly to Caivex. We’ll email you when your workspace is ready.
  • Improved mobile layouts make reviewing creator profiles and refining your brief easier on smaller screens.
Start your creator brief

v1.22.0

Sharper images for creator review

Higher-resolution profile images are now available as evidence for creator qualification. Saved copies keep the visual evidence consistent across subsequent reviews.

  • Added higher-resolution profile images where the source provides them.
  • Verify actual image dimensions before accepting an image as higher resolution.
  • Preserve captured images so reviews can refer back to the same evidence without depending on a changing source link.

v1.21.0

  • Added checkpointed profile enrichment and post-collection runs with request budgets.
  • Interrupted batches now recover from durable journals without repeating completed writes.
  • Post collection checks that the selected creator scope and qualification rules have not changed before resuming.
02 / Evidence
Checkpointed collection
  1. 01
    Bound requests

    Respect the run budget

  2. 02
    Save progress

    Write a durable journal

  3. 03
    Recover batch

    Validate scope and rules

  4. 04
    Continue

    Skip completed writes

Interrupted work resumes from its saved state.

v1.20.0

  • Expanded creator-profile enrichment with response identity checks and shared configurable qualification filters.
  • Added frozen-input qualification experiments with explicit abstentions and recent-activity evidence.
  • Evaluation runs preserve human labels and stop new work when a configured cost limit is reached.

v1.19.0

  • Creator discovery now stops at the configured target and resumes from recorded progress.
  • Added client-and-batch locking to prevent duplicate discovery runs.
  • Run results now distinguish exhausted sources, capped batches, resolution failures, and partial errors.

v1.18.0

  • Qualification model comparisons now use fixed sampling parameters and record the provider serving each response.
  • Added confidence intervals around model agreement so overlapping results are not presented as a clear winner.

v1.17.0

  • Reply ingestion now uses the source’s confirmed completion signal instead of treating a short response as complete.
  • Responses without a continuation signal are handled as complete.

v1.16.0

  • Outreach records now retain a snapshot of the reviewed targeting evidence and cohort assignment.
  • Added target-resolution fields for reporting and duplicate checks.

v1.15.0

  • Added per-client manual-reply mode that continues ingestion and classification without generating draft responses.
  • Already queued drafts also respect the client’s drafting setting.

v1.14.0

  • Added referral attribution through the creator identity-verification journey.
  • Verification returns are signed and checked before an application status is updated.
  • Inbox transport errors now preserve the original provider failure after retries are exhausted.
06 / Outcomes
Referral attribution through verification
  1. 01
    Application

    Creator enters the flow

  2. 02
    Referral code

    Keep the source attached

  3. 03
    Verification

    Complete identity checks

  4. 04
    Signed return

    Validate the attribution