Skip to content

Releases: PublicAI01/trajector-cli

v0.2.1

v0.2.1 Pre-release
Pre-release

Choose a tag to compare

@github-actions github-actions released this 31 Aug 10:58
v0.2.1

Changelog

  • 53eef63 Ask the optional setting at enable and undo it on every removal path
  • 57aecc0 Close the status Project section with the optional settings
  • 7db59e4 Lay the optional-settings groundwork: chain, record, public rule
  • 397d4ba One question module and one restore renderer
  • b4cd076 Run CI once per pull-request push
  • b4da4fb True up the README status line: beta, functional end to end
  • 2028c95 Write the 0.2.1 changelog

v0.2.0

v0.2.0 Pre-release
Pre-release

Choose a tag to compare

@github-actions github-actions released this 30 Aug 09:44
6e5a56b

Curated changelog: CHANGELOG.md — 0.2.0

Changelog

  • 086cc31 Collapse selfupdate to one entry point and give semver its own package
  • bd17fda Fail fast on permission errors instead of burning the lock timeout
  • ef02067 Fold proxy health, holder and takeover policy into one Verdict value
  • 784ddb3 Give the test harness named upload-state and refusal capabilities
  • 4c4d07f Give upload stoppage one answer: a persisted Standing per reason
  • fca073d Make install.sh skip drafts and rank releases by semver precedence
  • de7b33d Pause uploads on 451 until the account's data authorization is complete
  • 1faec03 Promote the five-valued settlement verdict to a production type
  • 8fb62ee Report every build refusal at once and put batch transitions on a lease
  • 4ca26bf Split serving and rendering out of lifecycle; one composition root
  • 64cdbf9 Trim unreferenced exports, raise the coverage floor, true up the docs
  • 6e5a56b Write the 0.2.0 changelog and link the version comparisons
  • a6f062e fix: 1 issue(s) found by automated scan
  • 71fb732 fix: 2 issue(s) found by automated scan
  • 668b97f fix: 2 issue(s) found by automated scan
  • f5165fc fix: 2 issue(s) found by automated scan
  • 13198e2 fix: 2 issue(s) found by automated scan
  • f2533e6 fix: 2 issue(s) found by automated scan
  • e6e33fb fix: 3 issue(s) found by automated scan
  • 93be074 fix: 3 issue(s) found by automated scan
  • c21200e fix: 3 issue(s) found by automated scan
  • ab5d04c fix: 6 issue(s) found by automated scan

v0.1.1

v0.1.1 Pre-release
Pre-release

Choose a tag to compare

@github-actions github-actions released this 14 Aug 03:18

Changelog

  • efc314d Define the Homebrew cask ahead of publishing it
  • 357309d Install trajector with one command on macOS and Linux
  • 79e7d77 Keep a shell-configured relay across sessions, and disarm rejection wording
  • d939444 Publish macOS and Linux only
  • 4a40f78 Refuse to install over a path that is not a file
  • ae8770f Relay what the service says about a version refusal
  • ed3db67 Replace this binary with the newest published release
  • 732f7aa Run the install script in the test suite
  • 01bac75 Say what the counts and the retry offer actually mean
  • 585ec6d Stop telling compliant clients to upgrade
  • 239bc39 Unpack the Windows archive into a folder that exists

v0.1.0

v0.1.0 Pre-release
Pre-release

Choose a tag to compare

@github-actions github-actions released this 09 Aug 04:51
8f2e0e7

First release: local capture proxy with lazy lifecycle, consent-gated routing, verbatim capture with streaming reassembly, local redaction, spooling with a bounded quota, acknowledged idempotent batch uploads, enable/disable/logout/uninstall lifecycle, status and doctor (including quarantine requeue and diagnostic bundles), and the release pipeline.

Hardened ahead of the tag:

  • The proxy port's holder is trusted only after it answers an admin-token challenge, with one token published per listen address. Management calls that act on the port wait out a sibling still starting up instead of blaming it, and each rides a connection of its own, so a takeover never leaves a dead pooled connection behind.
  • Only a strictly older release is asked to drain and hand the port over, and the flush a proxy runs on its way out is bounded so the port is released inside its successor's wait; whatever that flush did not upload stays spooled and uploads later under the same batch id.
  • Upload attempts get a time budget sized to the batch, unreadable records are set aside instead of stalling every upload, and trajector doctor discard deletes a quarantined batch for good — the terminal counterpart to trajector doctor requeue.
  • trajector status says why recording stopped — a full or otherwise unwritable spool — in the same words doctor uses, and the dashboard stays whole when a store cannot be read.
  • State files are replaced and read atomically on every platform, Windows rename collisions included.

Verify a download against trajector_checksums.txt and the build provenance attestation as described in the README.