Skip to content

CHANGELOG.md

Latest commit

 

History

History
103 lines (72 loc) · 10.5 KB

CHANGELOG.md

File metadata and controls

103 lines (72 loc) · 10.5 KB
Raw

Changelog

v0.7.3 pm1-apply1 — Verified personality Apply and Roll Back (installation pending)

  • PR #17 passed offline CI and merged to main.
  • Adds a guarded managed-section diff/preview, verified active main workspace discovery through openclaw agents list --json, and APPLY BOB confirmation to update only a delimited style block in SOUL.md.
  • Backs up the original file privately with checksums, atomic replacement and a ROLL BACK action that refuses changed files. Existing identity text, other files, publishing permissions and workers remain unchanged.
  • Single bobctl update and browser acceptance. The update adds controls only and does not apply personality automatically.

v0.7.3 pm1-csp1 — Local JavaScript policy fix (operator-verified)

  • PR #16 passed CI and merged to main. Extends the existing System Status script/fetch CSP exception to Overview and Bob's Mind · Personality, so the local JS preview and Overview refresh can run.
  • Keeps inline/third-party scripts blocked and the rest of the dashboard's strict CSP unchanged. One generated Activity Manager file, no identity, database, approval, rate-limit or worker changes.
  • The operator confirmed pm1-foundation1 installed with persistent sliders, and subsequently confirmed the local JavaScript preview fix works as expected.

v0.7.3 pm1-foundation1 — Staged Personality Lab (operator-verified; preview CSP fix released)

  • Adds Bob's Mind → Personality with eight 0–100 sliders and real-time non-LLM text interpretation.
  • Authenticated CSRF-protected profile save and revision/restore history in private SQLite, separate from publishing permissions.
  • Read-only default workspace SOUL.md / IDENTITY.md file metadata inventory. No identity file writes, live agent changes, Ollama calls, public Moltbook activity or extra worker processes.
  • Synthetic regression suite covers validation, revision integrity, safe workspace file checks and original protected action methods. PM1 release notes.
  • One normal bobctl update followed by a browser check; actual Apply to Bob and safe identity diff/backup/rollback are planned for the next PM1 stage.

v0.7.2 redirects1 — Section-preserving action navigation (operator-verified)

  • PR #12 passed offline CI and merged to main. The operator reported Scout Approve/Reject returning to the new Overview.
  • Corrected Scout Approve/Reject/Reset, Queue, Verify, and Auto-enable/disable redirects to the same Scout filter/page and full-history context. Added strict local return-target allowlisting and section-aware error-page links.
  • Audited other actions: Originals, Manage Queues and Activity Limits already return to their sections; password rotation deliberately returns a reauthentication message.
  • Three-file Activity Manager release, no posting worker, authorization, limits or database changes. Operator confirmed installation and correct section navigation.

v0.7.2 overview2 — Compact Overview cleanup (operator-verified)

  • Removed the redundant Mission Control hero, duplicate section-navigation buttons, and long helper paragraphs. Kept daily budget gauges, local rate-slot timestamps, queue and publication data, service/worker cards and live refresh.
  • Removed unused styles and JavaScript references to the deleted header to preserve automatic refresh.
  • PR #11 passed offline CI and merged to main. One-file bob_overview.py update from v0.7.2-overview1; operator confirmed it installed successfully and looks good.

v0.7.2 overview1 — Live overview landing page (operator-verified)

  • PR #10 passed offline CI and merged to main. The operator confirmed the release installed successfully and the new Overview was looking good.
  • Changed / from Scout to Overview, and moved compact review to /scout, retaining /scout/full and all original review/publish controls.
  • Added responsive daily budget gauges, queued/manual/automatic and original-post counts, recorded publication outcomes, observed incoming replies, live service and worker cards, recent events, and rate-slot estimates.
  • Spinners appear only for recently observed active worker stages, not queued/idle/stale workers. Refreshes every 15 seconds through a guarded read-only endpoint and respects reduced motion.
  • No changes to workers, Activity Limits, publishing permission, database schema, or sensitive POST routes.

v0.7.2 activity-limits1 — Settings and worker budgets (operator-verified)

  • Split Settings into overview, Users & Access (preserving login/password rotation), and Activity Limits.
  • Added validated configurable UTC comment attempts, minimum comment spacing, independent auto-comment sub-budget, original-post attempts and original-post spacing, enforced directly in each worker.
  • Recommended 6/45/3/2/120 limits require explicit APPLY LIMITS save; installing code preserves conservative previous limits by default. Posting permissions, publisher timers, and no-automatic-retry protections remain unchanged.
  • Reviewed through PR #9 with green offline CI; the owner confirmed installation and saved custom limits (8 total comments/day, 30-minute comment spacing, 4 automatic/day and 4 originals/day).

v0.7.2 compact1 — Compact Scout & Activity (operator-verified)

  • Default Scout & Review and Published Activity views redesigned into eight-item pages, collapsible content, searches, state/visibility filters and category tabs to reduce mobile scrolling.
  • Legacy unabridged pages preserved under /scout/full and /activity/full; Manage Queues full-report links updated.
  • Existing approval, reset, publish queue, verification and autonomous-comment POST handlers preserved unchanged; GitHub offline suite passed and PR #8 merged to main.
  • Release included three application files and the Activity Manager restart. The operator confirmed the updated UI works as expected.

v0.7.2 queues2 — Pending review status fix (operator-verified)

  • Operator screenshots confirm the queue management page is now installed following successful guarded incident reconciliation and retry of v0.7.2-queues1-r1.
  • A review-pending Scout draft with an automatically queued comment was incorrectly labeled only queued, disappearing from the Pending filter.
  • Keep review decision and publication state independent; show both badges, count archived pending separately, allow legitimate review, and refuse duplicate publication queue jobs. Original-post draft records also match Pending.
  • Fix is a single-module replacement (bob_dashboard_manage.py) with Activity Manager restart only; it does not change publication records or archive flags.

v0.7.2 queues1 — Deployed after guarded retry

  • Added an authenticated, mobile-friendly Manage queues screen with Scout, Original Posts and Comment Activity tabs, server-side search/status/visibility filters, ten-item pages, compact previews and batch selection.
  • Implemented bounded bulk approve/reject and explicitly confirmed queue for publication actions, preserving all publishing safeguards and prohibiting direct publication from the UI.
  • Implemented reversible archive/restore flags without deleting source or publication records. Archiving a queued item does not cancel its future publication.
  • Passed GitHub's offline acceptance and merged PR #6. The initial v0.7.2-queues1 deployment failed; the operator's read-only audit confirmed the previous manager and navigation restored and the new module absent. The original incident archive remains preserved. A guarded reconciliation, improved updater readiness check and new v0.7.2-queues1-r1 release are prepared but not yet installed.

v0.7.2 — Deployed (operator confirmed October 29, 2026)

  • Responsive desktop/mobile multi-page Activity Manager, secure Settings password change and trusted-console login reset.
  • Live 15-second System Status with real worker checkpoints, stale/unknown handling, bounded error summaries and authenticated API.
  • Scout, comment publisher, original generation and original publication stages reflected without fake percent progress.
  • Original-thought drafts accept strictly structured variants of title/body output while preserving content validation and human approval.
  • Comment-verification responses without confirmed IDs remain verified_unconfirmed; public visibility is not inferred from an API success response. Added a bounded read-only public comment audit.
  • Manual comment approvals take queue priority over automatically generated comments. Existing two attempts per UTC day, one-hour minimum spacing, and no-automatic-retry safeguards remain unchanged.
  • The operator verified a successful original-thought draft, four Moltbook-visible published comments, and a clean four-test manual-first publisher update.

v0.7.2 — Update infrastructure

  • Implemented bobctl update and a verified one-time bootstrap for the existing host. The operator installed the bootstrap successfully and observed an up-to-date status.
  • Merged the working v0.7.2 development branch to main via PR #5.
  • The operator reports the guarded updater migration from feature/v0.7.1-navigation to canonical main completed successfully. The old checkout remains available for reference.
  • Established the no-op v0.7.2-main-channel release manifest. The first nonempty manifest-driven update will be the forthcoming dashboard-polish release.

v0.7.0–v0.7.1 — Historical foundation

  • Imported and reviewed sanitized host source with private inventory and verified SQLite backup tools.
  • Implemented guided responsive navigation, release rehearsals, version-pinned production upgrades, health probes and code-only rollback workflows.
  • Preserved Moltbook credentials, approval records, publication history and original-post consent boundaries outside GitHub.

Not yet implemented

  • A supported clean-host installer for Windows/WSL/OpenClawLab.
  • v0.7.3+ evidence-linked learning, evolving opinions, reflective memory and multi-turn conversation continuity.

See operator guide, main-channel migration, and installation status.