v1.5.3

v1.5.3

2026-04-15

✨ Features

• Per-agent thinking level control: Review agents now support a thinkingLevel setting in their config. When set, it takes precedence over the pipeline-level default — giving you fine-grained control over reasoning depth per agent. The agent creation flow guides you through setting it, and existing agent configs are automatically migrated on first use.

v1.5.2

v1.5.2

2026-04-15

🐛 Fixes

• Installer now correctly copies the skills/ directory and merges peer dependencies into the runtime package.json, ensuring all required files and dependencies are present after install or update
• Running install or update now restores node_modules by executing bun/npm install automatically, so the extension is immediately usable without a manual step
• Installer preserves the actual runtime dependency set rather than a hardcoded subset, preventing missing-module errors at runtime

🏗️ Maintenance

• README documents release flags (--raw, --dry-run), commit scoping behaviour, and the multi-agent installer workflow

v1.5.1

v1.5.1

2026-04-15

✨ Features

• Release notes now only include changes relevant to the published package paths, keeping changelogs focused and noise-free.

🐛 Fixes

• Review agent loading is now resilient to missing or malformed agent definition files — a bad agent no longer breaks the entire review pipeline.
• Chunk and tool response sizes are now hard-capped to prevent hitting API size limits during long context-mode sessions.

🔧 Improvements

• Planning steps now consistently use plain prose over pseudocode, producing clearer, more actionable task breakdowns.

🏗️ Maintenance

• README and AGENTS.md updated to reflect context-mode behavior and installer changes.
• Package metadata expanded with engines, author, homepage, bugs, and additional keywords.

v1.4.0

v1.4.0

2026-04-13

✨ Features

• /supi:review — AI multi-agent review pipeline. A full programmatic review pipeline that runs headless AI agents in parallel, collects structured findings, validates them against actual code, and persists sessions to .omp/supipowers/reviews/. Includes an integrated LSP diagnostics gate, config recovery, and progress tracking.

• /supi:agents — review agent management command. New command for listing, creating, and configuring review agents. Agent prompts expanded with comprehensive review criteria.

• /supi:generate docs — AI-driven documentation drift detection. Detects drift between code and documentation using a headless sub-agent pipeline.

• Deterministic quality gate system. Replaced profile-based quality configuration with a composable, schema-validated gate registry. Gates are scope-aware (project/global), cover LSP diagnostics, linting, type-checking, test suites, and builds, and emit a structured ReviewRunEvent stream. Setup flow is interactive and guided.

• Context-mode tools integration. Wired web fetcher, HTML-to-markdown converter, knowledge chunker/store, sandbox executor, and shell runners into commands and bootstrap. Session lifecycle hooks and a per-project event store added.

• Planning improvements. Added an ask tool to the planning command, extracted the system prompt, and revamped the prompt builder for clearer AI steering.

• Platform utilities. Added a debug logger, progress indicators, TUI color utilities, and a generic createWorkflowProgress abstraction used across commit, review, and release workflows.

• Release executor: rebase before push. The release command now runs git pull --rebase before pushing, reducing merge conflicts in CI-heavy workflows.

• Path and text helpers. Added moduleDir and toBashPath path utilities, and a normalizeLineEndings utility for consistent cross-platform text handling.

• Self-update: cleans up legacy MCP registrations. The update flow now copies the bin/ directory and removes stale MCP server configurations left by older installs.

🐛 Fixes

• Fixed CRLF line endings causing parse failures in git output, release notes, and storage files.
• Fixed path resolution for Next.js root routes (app/page.tsx, pages/index.tsx) and broadened routing rule detection.
• Improved Next.js detection to use the presence of app/ or pages/ directories rather than relying solely on next.config.
• MCP configuration is now written to agent/mcp.json instead of the incorrect settings/mcp.json.
• Fixed path resolution and Bash argument encoding on Windows.
• An EPERM error when killing a process now surfaces as a 'failed' status instead of being silently swallowed.
• Release command now correctly uses buildFixPrompt and defers drift state updates to the fix step.

🔧 Improvements

• Quality system overhaul. Migrated from profile-based defaults to a gate-based configuration schema throughout config, commands, storage, and the review pipeline. Legacy profile modules removed.
• LSP bridge rewrite. The LSP bridge now collects structured JSON diagnostics instead of raw text, improving reliability and downstream processing.
• Release command overhaul. Prompt generation, commit utilities, and the overall release flow significantly revised for clarity and correctness.
• Context-mode renamed. The module is consistently named supi-context-mode throughout the codebase; the detector was slimmed and a hooks module added.
• Installer simplified. Removed the context-mode wrapper logic from the installer; install flow is leaner.
• Shared text utilities. Extracted stripMarkdownCodeFence as a reusable utility. fetch-pr-comments.sh and several QA/visual shell scripts replaced with TypeScript equivalents.
• Config and schema simplification. Model resolver, shared types, and the config schema simplified as part of the gate-registry migration.

🏗️ Maintenance

• Added CHANGELOG.md covering v0.6.1 through v1.3.0.
• Removed obsolete MCP server configurations and documentation.
• Excluded omp_source/ from the published npm package.
• Overhauled all project skills for clarity and completeness; added the omp-extension-dev skill with a full reference suite and a creating-supi-agents interactive guide.
• Added tests for the agents command, agent-loader APIs, quality gate scope saving, Next.js detection edge cases, EPERM kill handling, and context-mode tool name detection.
• Updated architecture documentation with new modules and commands.
• Removed stale plans, specs, and the legacy install log.

v1.3.0

v1.3.0

2026-04-10

✨ Features

• add /supi:optimize-context token analysis command (commands) 6369a9d
• add system prompt analyzer and tech stack context report builder (context) 662fe55
• persist resume snapshots to DB; track compaction metadata in hooks (context-mode) 2be9b99
• reference-based resume snapshot with ctx_search query hints (context-mode) be61a0b

🔧 Improvements

• remove full-file read block; expand HTTP routing patterns (context-mode) a653850
• head+tail read compression preserving hashline anchors (context-mode) 74ef5c6
• use PRIORITY constants; extract rule and skill event categories (context-mode) db3cbb6
• numeric EventPriority, session meta, resume persistence in event-store (context-mode) 832b3db

🏗️ Maintenance

• remove read-block policy; document automatic head+tail compression (context-mode) 94b43ac

v1.2.6

v1.2.6

2026-04-05

✨ Features

• display commit model override in status bar during commit flow (git) 1131ae8
• show active model override in footer status bar (model-resolver) 39480d7
• add thinking column to model dashboard (model) f7fbbc7
• register model actions and apply override in all command handlers (commands) 897d331
• re-apply model override on plan execution handoff (planning) a27a878
• resolve model for commit sub-agent and validate staged files (git/commit) 5b0c63d
• add applyModelOverride with auto-restore on agent_end (model-resolver) 61ad6ca

🐛 Fixes

• pass actionId to applyModelOverride in all command handlers (commands) 18910bf

🔧 Improvements

• remove inline ModelPref config; route model via model-resolver (fix-pr) 514b22a
• make setModel async and add setThinkingLevel to Platform (platform) d590d1f

🏗️ Maintenance

• cover status bar behavior and updated applyModelOverride signature (model-resolver) 10d1d82
• add supipowers install log (install) 2091957
• add @oh-my-pi/pi-ai as dev and optional peer dependency (deps) 0c723e1
• add applyModelOverride suite and fix bun:test imports (model-resolver) b4ef291

v1.2.5

v1.2.5

2026-04-04

✨ Features

• add --debug flag with structured install log (installer) 7ad41c5
• add Windows local-install script (PowerShell) (installer) b41f8b6

🐛 Fixes

• always pass --force when invoked via bunx/npx (installer) f503f10
• forced local installs to reinstall missing deps (installer) 8d4bf50
• add natives-test to ignored files (.gitignore) b3bb132
• strip devDeps from installed package.json and drop --frozen-lockfile (installer) bc08397
• update registry and tests for revised dependency model (deps) 9c6515e

🏗️ Maintenance

• run installer with --debug in local-install scripts (installer) 07b6383
• remove leaked natives-test fixture directories (test) 21022d2

v1.2.4

v1.2.4

2026-04-04

🐛 Fixes

• run bun install in ext dir after copy to resolve runtime imports (installer) 9d90ed3

🏗️ Maintenance

• declare pi-ai and typebox as optional peer deps (deps) cb2b420

v1.2.3

v1.2.3

2026-04-04

✨ Features

• add release as valid conventional commit type (release) 1f8ac9a
• add skipTag option to executeRelease (release) fef7aeb
• add isTagOnRemote to distinguish local-only tags (release) 3f40019

🐛 Fixes

• add Windows compatibility to installer binary lookup and spawn (install) 215dfff
• use 'where' instead of 'which' on Windows in checkBinary (deps) cec12d8
• handle local-only tag (push interrupted) as resumable release (release) 3006140

🏗️ Maintenance

• bump version to 1.2.3 (release) a0b16be
• update invalid-type rejection test to use unknown type (git) 9065302
• verify checkBinary uses platform-correct lookup command (deps) 7f6f33a
• update isInProgressRelease test for local-only tag case (release) 8fb6140
• add tests for isTagOnRemote and skipTag executor behaviour (release) 9f09d66

v1.2.2

Fixes

• fix(release): use conventional chore(release) commit type in release flow

Features

• feat(release): skip confirmation when resuming a staged unreleased release

Refactor

• refactor(git): extract commitStaged shared primitive from commit.ts

Tests

• test(release): update executor snapshot to chore(release) commit message
• test(git): add commitStaged unit tests