Skip to content

QVeris CLI v0.12.0

Choose a tag to compare

View all tags
@qverisai-release-bot qverisai-release-bot released this 27 Oct 13:57
· 17 commits to main since this release
cli-v0.12.0
f7c0e23
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Install

# One-liner
curl -fsSL https://qveris.ai/cli/install | bash

# Or via npm
npm install -g @qverisai/cli@0.12.0

Highlights

Fixed

  • Classify login, identity, OAuth status, and doctor validation failures without blaming valid credentials for permission, credit, rate-limit, timeout, network, or service failures. Preserve JSON diagnostics, HTTP status, credential source, and established error exit codes.

  • Clarify API key paste login versus OAuth Device Flow and report the active credential source and CLI version in diagnostics.

  • Report successful Calls with pending settlement review as manual review required, preserving execution results without replaying the Call.

Added

  • Add explicit --public-read-only Discover and Inspect without credentials. Reject mixed identities, keep public selections separate from authenticated execution, and refresh inspection, schema and quote after login before an upgraded selection is called.

  • Add Call idempotency keys and read-only recovery with qveris execution lookup <key> and qveris reconcile-call --idempotency-key <key>. Add --require-zero-quote for guarded execution after an exact zero-credit Probe; submitted Calls are never automatically replayed. ([#436])

  • Added qveris call --context <json|@file|-> as a tolerant-input, strict-execution consumer for copied v1 service/task contexts. It rejects sensitive or executable content, refreshes expired snapshots, ignores ordinary future fields with structured warnings, supports capability negotiation, and runs Inspect, Probe, or quote gates only when current execution policy requires them.

Changed

  • Report HTTP 403 responses as actionable permission failures instead of treating every 403 as an invalid API key.
  • Return structured recovery metadata (retryable, action, missing_fields, fallback_available, and current candidates) for context failures; service-only contexts now return candidates without guessing a tool.
  • Use deterministic exact-ID discovery for tool contexts, validate parameters with JSON type semantics (including integers), treat every ambiguous/non-free price signal as paid risk, reject unsupported --max-credits context calls instead of claiming a local quote is a hard cap, and bound context nesting during safe validation.
  • Warn when side-effect or idempotency metadata is absent while retaining strict single-submit Call behavior; only explicit dangerous or non-idempotent signals require confirmation.
  • Added bounded service-identity fallback, unknown-settlement Usage/Ledger reconciliation, and machine-readable next_action guidance.

See README for usage.

What's Changed

  • fix(cli): distinguish credential validation failures (C1–C2) by @linfangw in #459
  • docs: sync API docs from website by @qveris-docs-sync-bot[bot] in #458
  • feat(cli): add explicit trial registration with recoverable credentials by @linfangw in #460
  • chore(docs): retain browser trial guidance in canonical onboarding by @linfangw in #461
  • feat(cli): add explicit public discovery and sign-in upgrade by @linfangw in #463
  • feat(sdk): add explicit public discovery clients by @linfangw in #464
  • docs: distinguish read-only discovery from authenticated execution by @linfangw in #462
  • chore: release CLI 0.12.0 and public discovery SDKs by @linfangw in #466
  • docs: explain explicit public discovery in clients by @linfangw in #467
  • test: normalize temporary paths in runtime checks by @linfangw in #468

Full Changelog: mcp-v0.15.0...cli-v0.12.0

Contributors

linfangw
Assets 2
Loading