Changelog
299releases tagged feature
Every PAPI release carrying the feature tag. What shipped, when.
Get started freeMarch 202618 releases
- Research: PAPI as Claude Code plugin — single-command installation
- Add gotchas section to alpha tester guide — 95 cycles of learned failure modes
- Harden .claude/settings.json for external users — pre-approve PAPI workflow permissions
- Split CLAUDE.md under 200 lines — monorepo-aware structure
- Audit all PAPI skill descriptions — rewrite as trigger conditions, not summaries
- MCP Registry listing — register PAPI on modelcontextprotocol.io
- 5-minute quickstart guide for external users
- MCP tool reference page — what each tool does, inputs, outputs
- Add "Run setup first" guard to orient and plan when project doesn't exist in DB
- Update init.ts to generate npx-based .mcp.json config instead of hardcoded node path
- Fresh alpha readiness audit (Cycle 29 audit is stale)
- Example project brief template for new users
- Public README rewrite with Taskmaster-benchmark structure
- Create llms-install.md — AI-readable install guide
- One-command Claude Code install — verify and document npx path
- License decision and LICENSE files for all packages
- Ship .env.example with placeholder values for all packages
- Validate DATABASE_URL on first tool call — surface clear error instead of silent degraded mode
- Auto-add .mcp.json to .gitignore during init — contains credentials
- New project smoke test — friction audit and checklist
- Dogfood pipeline — strategy reviews write observations to DB
- Deprioritise preserves BUILD HANDOFF, only defer clears it
- Fix 3 pre-existing server test failures
- Add existing project adoption flow to setup
- Add one-line pitch, elevator description, and tagline
- Strategy review action items — structured output with follow-through gate before plan
- Create PAPI positioning doc — consolidate scattered messaging into single artifact for S3 completion
- Remove dead code — CycleHeader.tsx, ActiveCycleTasks component, brain.json file fallback handler
- Idea tool auto-routing silently swallows ideas into Discovery Canvas instead of backlog — revert to always create task
- Build @papi-ai/shared before next build for Vercel
- AD cards — collapsible with markdown rendering, strip HTML comments
- Idea tool smart routing — classify discovery context vs backlog tasks
- Standard doc management system generated by papi setup
- Review feedback loop — wire v_recommendation_effectiveness to track if strategy recs were followed
- Real-time dashboard updates via Supabase Realtime
- Dead ends field in build reports — track failed approaches to prevent repeat blind alleys
- Dashboard design audit — comprehensive UX/visual pass using frontend-design skill
- Cockpit re-fetches on papi-refresh events
- Phase/horizon timeline band — clean up phase data and add visual roadmap band above project timeline bars
- Cycle assignment + In Cycle status for planned tasks
- Add optimistic UI for status changes, saves, and comment posts
- Setup smoothness audit — test and polish the install-to-first-orient flow end-to-end
- Taxonomy migration Batch D2: Dashboard roadmap — restructure for Horizon→Stage→Phase hierarchy with management UI
- Setup/init flow update — support new Horizon→Stage→Phase hierarchy for new PAPI projects
- Taxonomy migration: DB schema + code rename — sprint to cycle across tables, columns, and TypeScript
- Graceful fallback if getPlanContextSummary() fails — dead code path in plan service
- Cockpit dashboard polish — compact board, strategy layout, typography
- Dashboard consolidation — cockpit as central hub with embedded kanban and project timeline
- NorthStar supersede indicator — show previous/replaced NorthStar on dashboard
- Remove file fallback readers from data-provider.ts — dead code when Supabase active
- Populate AD-14 hierarchy and wire read/write paths
- Time-window strategy_review board read — filter to last N cycles
- Estimation calibration by task type — wire v_estimation_accuracy into planner context
- Connect orphaned DB views to dashboard and health tool — 7 pre-built analytics views unused
- North Star drift detection — flag when North Star hasn't been validated in 10+ cycles
- Guard briefImplications with Array.isArray before .map()
- Fix strategy review rendering on dashboard
- Taxonomy migration: Coordinated 'In Sprint'→'In Cycle' status rename across packages/shared, MCP server, and adapter-pg
- Taxonomy migration: Rename brain data types (SprintIntelligence, BrainData.sprints, ReviewSummary.bySprint) to cycle terminology
- Taxonomy migration Batch E: Documentation — rename Sprint→Cycle in all docs, templates, and domain model
- Review fixes — missed GroupMode enum, URL params, CSS class, DataPoint renames
- Taxonomy migration Batch D: Dashboard — rename Sprint→Cycle in all components, labels, API routes, and chart titles
- Taxonomy migration Batch B2: MCP services — update plan, strategy, orient, health, build, review, release services for cycle/hierarchy
- Rewrite prompts.ts with cycle/hierarchy taxonomy (AD-14)
- Deprecate Feature entity (AD-23) — subsumed by Phase in AD-14 taxonomy
- Orient shows orphaned sprint tasks after release
- Pg adapter — cycle/horizon/stage support
- Taxonomy types — Horizon/Stage/Cycle interfaces and aliases
- Taxonomy migration Batch A — DB schema foundation
- Document Builder/Owner/Client user journeys for discovery canvas
- Reframe velocity metrics as sprint sizing input
- Show sprint-assigned tasks separately from backlog handoffs
- Research: Project hierarchy taxonomy — define clean model for phases/horizons/epics/milestones
- Separate lifecycle from orient — show in health only
- Wire AD lifecycle tracking — outcome, revision count, lifecycle summary
- Product brief update triggers — decision-driven, not age-based
- Rewrite research doc with verified findings
- Research v2 retrospective learnings applicable to v1
- Add minimum effort-point floor to prevent deflationary sprint sizing