v1.1.6

1.1.6 (2026-07-17)

Summary

This release hardens the release automation pipeline by fixing tag identity configuration in CI and moving tag and GitHub Release finalization to the authenticated local process after successful npm publication. Previously, the CI workflow attempted to create release tags using the GITHUB_TOKEN, which cannot tag commits containing workflow changes; now the local release wrapper waits for CI to publish the package, retries npm metadata propagation to verify the published gitHead matches the exact release commit, and only then creates the annotated tag and GitHub Release using the maintainer's authenticated git and gh clients. This ensures verification-before-tagging while avoiding GitHub App token restrictions, and uses bounded retries with exponential backoff to handle delayed npm metadata propagation.

• fix(release): configure CI tag identity (cae54af)
• fix(release): finalize tags after CI publication (707170e)

v1.1.5

1.1.5 (2026-07-16)

Summary

Release automation was hardened to require explicit GitHub Actions dispatch with an exact version and commit SHA, replacing the previous tag-push trigger so that all CI gates—typecheck, lint, format, tests, build, pack validation, and release consistency checks—complete successfully before npm publication, remote tag creation, and GitHub Release creation, preventing accidental or premature publishing and enabling safe reruns for tag recovery.

• fix(release): gate publishing and tags behind CI (55908c5)

v1.1.3

1.1.3 (2026-07-16)

Summary

Release 1.1.3 fixes a layout issue in the /context dialog where a duplicate dialog wrapper caused incorrect rendering. The fix moves dialog container ownership to the host and applies "xlarge" sizing after the content is replaced, ensuring the context breakdown displays with the intended dimensions and no nested modals. This improves the reliability of the context inspection interface for all users.

• fix(tui): correct context dialog layout (0e22c00)

v1.1.2

1.1.2 (2026-07-16)

Summary

This release fixes the TUI plugin registration to use the pinned base package specifier instead of the legacy /tui subpath, ensuring OpenCode correctly selects the managed ./tui export. Existing installations with the old subpath are automatically migrated during sync. Additionally, vvoc status and vvoc doctor now report the installed OpenCode version and check whether it meets the minimum 1.18.2 requirement for TUI compatibility, making it easier to diagnose host version issues.

• fix(tui): register pinned package entry (55920a2)

v1.1.1

1.1.1 (2026-07-16)

Summary

This release introduces a new /context TUI command that gives users real-time visibility into their OpenCode session's context usage directly within the editor interface. The feature displays the latest provider-reported token consumption alongside a clearly labeled, approximate breakdown of observable context contributors—such as system instructions, loaded skills, tool schemas, conversation messages, files, and compaction summaries—without claiming exact provider attribution. It also surfaces MCP server statuses and reports estimation drift when visible estimates exceed measured usage. The plugin is enabled by default and can be toggled via vvoc plugin. The installation, sync, launch, status, and doctor commands now also manage a dedicated TUI configuration file (tui.json / tui.jsonc) conservatively, preserving unrelated settings and existing plugin entries. Package dependencies have been updated to OpenCode 1.18.2 and OpenTUI 0.4.3 to support the modern TUI plugin API, and a new @osovv/vv-opencode/tui package subpath is published for consumers.

• feat(tui): add context usage inspector (328ed9f)

v1.1.0

1.1.0 (2026-07-11)

Summary

This release introduces orchestration profiles—single-session, balanced, and orchestrated—that let users control how vv-controller delegates work at runtime, with built-in presets automatically selecting a sensible default, a new vvoc orchestration show|set CLI command for explicit profile management, and status diagnostics that report the effective profile from the selected config source; alongside this, the managed OpenAI preset has been renamed to vv-codex with explicit Codex subscription-safe token limits to prevent compaction failures, the managed controller prompt and skill templates have been made profile-neutral to eliminate context pollution, and the repository agent guide has been rewritten for clarity.

• chore: archive applied orchestration profiles change (09228ac)
• chore(grace): archive applied Codex preset change (5d7157c)
• docs: document orchestration profiles (01b5786)
• docs(grace): improve repository agent guide (f5e9e3b)
• docs(grace): project orchestration modules and flows (e66875c)
• feat(cli): add scoped orchestration command (c5bb612)
• feat(cli): register orchestration completions (ec26cfc)
• feat(config): add orchestration profiles to schema v3 (6c920c3)
• feat(context): inject concrete controller policy (1941c34)
• feat(grace): plan preset orchestration profiles (997a1ea)
• feat(orchestration): add profile policy domain (6de8219)
• feat(preset): apply orchestration profiles atomically (161d7ff)
• feat(status): report effective orchestration profile (473d92e)
• feat(workflow): select profile-compatible guidance (b664d4b)
• test(skills): protect workflow discovery metadata (be76244)
• refactor(controller): keep managed prompt profile neutral (092b995)
• fix(codex): rename managed presets and set safe limits (ae77c7c)

v1.0.2

1.0.2 (2026-07-10)

Summary

This release fixes the fast role alias in the built-in osovv presets by replacing the unavailable GPT-5.6 Luna Low model with GPT-5.4 Mini Low, restoring functionality for users relying on that role assignment without affecting the Terra and Sol models.

• fix(models): use GPT-5.4 mini for fast role (5068545)

v1.0.1

1.0.1 (2026-07-10)

Summary

The osovv presets in vvoc patch-provider now replace the deprecated StepFun model with GPT-5.6 aliases (Luna Low, Terra High, Sol XHigh), delivering consistently higher throughput for the explore subagent after StepFun performance degraded from over 150 TPS to approximately 25 TPS. The vv-osovv and vv-osovv-cheap preset role assignments are updated to use these new fast and smart models, and the patch-provider openai preset now includes all three GPT-5.6 variants alongside the existing GPT-5.4 and GPT-5.5 aliases, ensuring users get reliable performance without manual reconfiguration.

• feat(models): adopt GPT-5.6 in osovv presets (fd36444)

v1.0.0

1.0.0 (2026-07-03)

Summary

vv-opencode reaches version 1.0, establishing a daily-driver baseline for the curated OpenCode workflow. The release formalizes a stability posture where setup commands, managed skill names, public package exports, schema v3, and the spec artifact layout are treated as compatibility surfaces, meaning breaking changes to these areas will be explicitly documented in future release notes. User-owned config is never silently clobbered, and invalid config continues to fail loudly. This marks the transition from iterative development to a practical, documented baseline for real project use.

• docs: declare 1.0 stability posture (8a3e590)

v0.35.33

0.35.33 (2026-06-26)

Summary

This release removes the harmful RTK (rtk-ai/rtk) recommendation from the interactive vvoc init outro and the README, because RTK proxies developer commands and distorts their output shape, causing automated sessions to receive unexpected responses and work around RTK instead of completing the intended work. Users will now see a clean init success message and no longer be directed to install a proxy that interferes with standard command output.

• fix(init): drop harmful RTK recommendation that distorted command output (cd5c12a)