Changelog
388releases across 7 months
Every PAPI release, cycle by cycle. Each one pushes back against project drift.
Get started freeMarch 202618 releases
- Planner prompt — explicit discovered issues to task pipeline
- Data mutator layer + action API routes for dashboard writes
- Shared service layer — task transitions + review validation
- Plan context optimisation — section hashing + reviews 20→5
- Truncate notes field to 2000 chars in idea/bug/ad-hoc tools
- Increase pg adapter connection pool from 1 to 5
- Reduce plan apply DB queries — remove metrics snapshot, context util, add 60s timeout
- Reduce MCP server DB queries — remove metrics write-back, slim health
- Surface data fetch errors instead of silently swallowing them
- Task type system with context-tier filtering for plan
- Replace Sprint Timeline pill scroller with Recharts area chart
- Papi-reader tests — create fixture .papi/ directory instead of wrong path
- Revert to max:1 connection pool — Supabase ignores custom GUC startup params
- NextAction expandable tasks, timeline cleanup, cycle ring text clipping
- Setup as progressive advisory — 3 required fields, defaults for rest
- Add decision lineage tab to Strategy page
- Add continuity surface to Cockpit — next action, active decisions
- Plan performance — parallelize writeBack, guard handoff overwrites, exclude Deferred from context
- Add entity-references API endpoint for dashboard consumption
- Add bug command for reporting bugs as backlog tasks
- Add maturity column to tasks table and pg adapter
- Add task maturity pipeline — raw ideas gated from scheduling
- Wire up roadmap features from phase data in plan and strategy writeBack
- Surface decision scores in health output and strategy page
- Add project timeline narrative to Analytics
- Enrich strategy_change with board and velocity context
- Expand recommendation classification from 4 to 7 types
- Reports interactivity — expandable cards, fix Sprint 0, newest-first ordering
- Increase minimum text sizes across all dashboard pages
- Add task type metadata to Task schema with board filtering
- Kill costs, merge accuracy into performance, fix jargon, improve reviews
- Proper kanban with In Review column, fixed sprint filter, deferred toggle
- Fix sprint timeline, backlog age, and build insights
- Strategy page fixes — sort decisions, last-updated, inline roadmap expansion
- Ship default PAPI skills for external users
- Dashboard Phase 1b — Analytics + Strategy pages absorb Brain/Reports/Brief/Decisions
- Validate phase on task creation — resolve from adapter instead of hardcoding "Unscoped"
- Dashboard rethink Phase 1a — Cockpit + Cycle Ring + nav consolidation
- Soften review language for non-technical users
- Prepare npm packages for publishing
- Add maturity gating to plan prompt
- Fix methodology trends showing undefined/NaN
- Log createSprint failures instead of silently swallowing
- Suppress compression directives in strategy review for pg adapter
- Add build/review pattern detection to strategy review
- Add human reviews to strategy review context
- Auto-add mcp__papi__* permission during setup
- Inject dogfood log into strategy review context
- Fix health metrics NaN by normalizing JSONB and adding build report fallback
- Fix sync script sprint statuses and handoff parsing
- Add getSprintLogSince(n) for scoped sprint log fetching
- Extract strategy reviews into dedicated strategy_reviews table
- Fix boardHealth/strategicDirection reads and strategyReviewDue format on pg adapter
- Address PR review — card navigation, health status, duplicate fetch
- Add multi-project portfolio page with health metrics
- Wire sprint_metrics_snapshots, cost_snapshots, sprints into data-provider
- Release marks sprint complete, build_execute backfills sprint on tasks
- Dashboard data integrity audit — fix all stale/broken page data
- Add architecture health check to strategy review prompt
- Onboarding CLAUDE.md should lead with pg adapter setup, not .papi/ files
- Papi init CLI command — one-command MCP setup for new users
- Add infrastructure connection health checks
- Context utilisation metric — measure wasted tokens per tool call
- Entity reference tracking — which decisions actually get used
- Dead input audit — find all adapter read calls that silently accept empty results
- Add product_briefs table to Supabase, remove sync footgun from release
- Add sprint cycle stress test, fix SPRINT_LOG_TEMPLATE heading
- Brain page defaults to papi-ui instead of papi-framework
- Brain reviews query uses correct column name, add debug logging
- Brain handler picks source by max sprint number, not count
- Compute brain sprint intelligence from build_reports table
- Wire reports and handoffs to DB-backed data on Vercel
- Stringify jsonb build_handoff from DB before UI consumption
- Add Supabase connectivity check to debug endpoint
- Add Supabase URL format validation to catch .com vs .co typos
- Correct Supabase URL on Vercel (.co not .com), remove debug logging
- Debug: add temporary logging to getHealth DB path
- Dashboard data integrity — env validation, sprint statuses, remove War Room
- Make getHealth() DB-first to show current sprint on Vercel
- Exclude packages/ from root tsconfig to unblock Vercel builds
- DB-first architecture — remove dual-write, update CLAUDE.md for pg-only reads
- Auto-update task status in build_execute and review_submit — fix chronic board drift
- Disable review_submit Slack notification to stop per-review spam
- Update 83 failing tests to match two-phase prepare/apply refactor
- Make adapter-factory pg path resilient to unreachable DB
- Fix sync script tsx → npx tsx, remove stale known issues