Skip to content

fix: restore fast human-readable Console pricing #364

Description

@douglasmonsky

Problem

Installed R7 Console inspection on merged 827be57 exposed three connected product regressions:

  • /explore routes correctly but its default broad cost-aware share request took 33,006 ms on the maintained generation, despite the named top_threads MCP template completing its query in 22.827 ms.
  • configured costs and estimated credits have 0% coverage because clean kernel installs have no rate-card.json; the product reset retired the old OpenAI documentation updater and bundled pricing assets while retaining only the rate-card calculator.
  • major Console tables compress uncached input, cached input, reasoning, and output into one Token Mix cell instead of four sortable decision columns.

Required correction

  • make Explore initial views use bounded persisted/named fast paths and prove warm installed browser latency;
  • restore a maintained, source-stamped pricing bootstrap/update path using official OpenAI pricing provenance without hidden network work during reports;
  • render the four token classes as separate sortable columns everywhere they are available, retaining compact responsive behavior without losing values;
  • preserve explicit estimate/coverage labels, snapshot consistency, privacy, and exact selectors;
  • cover clean install, absent/invalid/partial pricing, installed browser, deterministic assets, package data, and performance budgets.

This is an R7 blocking correction discovered through installed product qualification.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions