Release Notes

Added

• ✨ Help Center page — comprehensive documentation for all PeerAI Studio modules (Deployment Copilot, Ask PeerAI, Crew AI, ITSM, Code Insights, etc.) with search and FAQ
• ✨ Roadmap page — interactive roadmap with timeline and kanban views showing planned features across Q1-Q3 2026
• ✨ PeerAI Studio Portal — new Next.js portal site with Cloud Run deployment, downloads, and release notes (#335)
• ✨ In-app announcements — What's New changelog viewer with bell notification in header (#334)
• ✨ AI Cost Calculator — comprehensive AWS Bedrock pricing calculator with compare, quick calc, and scenario tabs (#332)
• ✨ Enhanced feedback widget — 3-tab interface for Feedback, Feature Requests, and Bug Reports (#330)
• ✨ GCP-based distribution — release artifact distribution via Google Cloud Platform (#328)
• ✨ Code Generation and Performance Testing — enabled as project activities (#333)
• ✨ ADR publish action — publish support in ADR detail view with tests (#320)
• ✨ New agents — Medication Listing agent, Meeting Summarizer agent, and PHI guardrail (#324)

Fixed

• 🔒 Fix CVE-2026-28277 (langgraph) and CVE-2026-0540 (dompurify >=3.3.2) (#337, #338)
• 🐛 Fix portal deploy — switch from Cloud Build to direct Docker push (#336)
• 🐛 Fix ADR section metadata — include section metadata in ADR list responses for progress counts (#325)

Fixed

• 🐛 Fix testdatagen build — replace deprecated rand::seq::SliceRandom with rand::seq::IndexedRandom after rand crate upgrade
• 🔒 Fix minimatch CVEs in marketplace — add minimatch override/resolution to marketplace package to eliminate vulnerable 3.1.2 and 9.0.5 transitive versions

Fixed

• 🔒 Remediate minimatch ReDoS vulnerabilities — resolved CVE-2026-27903, CVE-2026-27904, CVE-2026-26996 by forcing all minimatch resolutions to 10.2.4 via overrides and resolutions
• ⬆️ Bump @tauri-apps/api to ~2.10.0 — align NPM package with Rust crate =2.10.2 to fix Tauri version mismatch error

Changed

• 🔧 Dynamic SBOM version — enrich-sbom.py now reads version from package.json instead of hardcoded value
• 🔧 Remove stale package-lock.json — project uses bun exclusively; added to .gitignore to prevent re-introduction
• ⬆️ Dependency updates — merged 6 Dependabot PRs (JavaScript root, marketplace, licensing dashboard, GitHub Actions, Rust testdatagen)

Added

• ✨ Restore Future of Work Careers Discovery — restored full careers page import pipeline, batch analysis, provider-aware discovery, and all associated frontend components/hooks/tests that were inadvertently removed in #306.

Changed

• ♻️ FOW dev-guide and doc references restored — re-added for-developers-future-of-work.md, Sub-Document Index entry, copilot deep-dive link, and alpha release checklist to release management docs.
• ♻️ CHANGELOG historical accuracy — restored missing 0.70.0-alpha.3 release section with full FoW careers import notes.

Added

• ✨ FOW Careers Page Import — bulk-discover and analyze job postings from any company URL with provider-aware ATS extraction (Oracle, Paycom), subdomain detection, and regression test suite (#298)
• ✨ Skills Library integration in Coder — integrated skills library into Coder copilot and pass API keys to OpenCode for seamless LLM access (#306)
• ✨ Release publish script & SBOM report generator — added automated release publishing workflow, SBOM report generation tooling, and vulnerability severity classification updates (#307)
• ✨ Keepalive mechanism for long-running tasks — implemented keepalive heartbeats to prevent timeouts during extended async operations
• ✨ Refactored export functions — improved user stories and test cases retrieval in spec export workflows
• ✨ Decoupled header text mode — separated header text mode from co-branding for independent configuration

Changed

• ♻️ FOW results UX polish — overlapping jobs shown in transformation chart points, automation profile badge in header, positions and status row inside project cards, and list summary metrics
• ♻️ Code Insights streaming resilience — enhanced run recovery and state management for Code Insights streaming sessions
• ⬆️ Migrate web search to ddgs — replaced duckduckgo-search with ddgs package for web search functionality
• ♻️ Artifact prompt refinement — resolved artifact redundancy by updating prompts and fixed cross-artifact update consistency
• 🔧 PAT token support — added support for personal access tokens for remote repository operations

Fixed

• 🐛 FOW careers discovery fixes — fixed subdomain detection, link noise filtering, and Exa API compatibility
• 🐛 FOW company insights — fixed empty data display in FOW detail view
• 🐛 FOW project deletion — fixed detail page project deletion action
• 🐛 FOW batch save crash — fixed typing crash in batch save and added HTTP content fallback
• 🐛 Product management DB field mismatch — fixed field name mismatch in DB for product run
• 🐛 Snake case fallback — fixed snake case conversion fallback handling
• 🐛 Mermaid syntax errors — added error handling for Mermaid syntax errors and suppressed error rendering
• 🐛 Python sidecar resources — fixed resource loading in Python sidecar
• 🔒 Dependabot security fixes — resolved security vulnerabilities in transitive dependencies

Added

• ✨ Future of Work Careers Import (end-to-end) — shipped a guided Import from Careers Page flow that discovers careers portals, previews jobs, and batch-creates analyzed FOW projects from selected listings.
• ✨ Provider-aware careers discovery pipeline — added extraction paths and hardening for Phenom-style portals, Oracle HCM handoffs, and Paycom ATS redirects, plus locale-aware URL scoping and recursive listing fallback.
• ✨ Batch analysis orchestration + resilience — introduced durable batch processing with progress streaming, cancel/resume semantics, and stronger fallback fetch behavior when job pages return thin content.
• ✅ Regression and verification coverage for careers ingestion — added focused backend discovery/batch tests and a verification script for multi-site careers crawling scenarios.

Changed

• 💄 Future of Work UX polish across list/detail/results — improved discovery/loading feedback, progress visibility, card alignment, link/open behavior, and richer automation profile styling in results headers.
• ♻️ Transformation chart interaction model — overlapping points now expose count badges and aggregate hover details so stacked jobs can be inspected without ambiguity.
• ♻️ Product generation/export stability path — refactored story/test-case retrieval flow and strengthened long-running task keepalive behavior for generation/export operations.
• ⬆️ Web search backend provider swap — migrated Python-side web search integration from duckduckgo-search to ddgs.

Fixed

• 🐛 Careers discovery extraction correctness — fixed Exa compatibility (type=auto), subdomain matching edge cases, and noisy non-job URL acceptance.
• 🐛 Batch analysis save-path crash — fixed dict/model normalization mismatch that caused model_dump errors during analysis persistence.
• 🐛 FOW list query instrumentation compatibility — fixed aggregate cursor handling with instrumented Mongo wrappers (coroutine has no attribute to_list).
• 🐛 Project management UX regressions — fixed non-functional delete action in FOW detail page and ensured project status/positions render consistently inside cards.
• 🐛 Company Insights/FOW integration regression — fixed empty-data rendering path on FOW detail views sourced from company insights.
• 🔒 Dependency vulnerability follow-up — resolved Dependabot-flagged transitive package vulnerabilities.

Added

• ✨ SDLC Pipeline Flow dialog — visual pipeline diagrams on home page showing how copilots compose into Legacy Modernization and Greenfield Development workflows with SVG-connected stage cards (#294)
• ✨ Skills Library expansion — 14 new built-in skills (frontend-design, mcp-builder, doc-coauthoring, etc.), 10 new frontend templates across 6 categories, and 4 seed user skills (#293)
• ✨ Installer vulnerability scanning — DMG and NSIS installer-level SBOM generation (Syft) and CVE scanning (Grype) in all release workflows (#295)
• ✨ SBOM enrichment pipeline — post-processing script for Aikido CycloneDX SBOMs with false-positive enrichment, suppression reports, and quality metrics (#292)
• 📝 SPDX license fields — added license = "LicenseRef-Proprietary" to Rust crates for correct SBOM identification (#290)

Changed

• 💄 Repo Insights UX revamp — hero stats, COCOMO estimate, and top contributors split by human vs AI agents with improved layout
• 📝 Product overview updated — refreshed metrics, added Supply Chain Security section, Skills Library, and Pipeline Flow documentation

Fixed

• 🐛 AgentNodeConfig schema shadowing — fixed config schema collision in agent builder
• 🐛 SBOM artifact naming — fixed scan-artifacts job to correctly find generated SBOMs
• 🔒 Zizmor template injection warnings — fixed GitHub Actions security linting findings in scan workflows
• 🔒 Aikido security fixes — bumped dependencies, Dockerfile integrity checks, CI cache hardening (#289)
• ⬆️ Upgrade cryptography 46.0.4 → 46.0.5 — addresses GHSA-r6ph-v2qm-q3c2 (#296)

Added

• ✨ Full-stack ITSM Copilot launch — delivered end-to-end ITSM project and ticket workflows with new FastAPI routers/services/models, seeded reference data, and broad test coverage for project/ticket lifecycles.
• ✨ ITSM analysis and strategy experiences — introduced dedicated pages and components for executive insights, incident/problem/service-request analysis, AI strategy planning, and benchmark views.
• ✨ ITSM dashboard analytics suite — added KPI and chart modules for MTTR trends, SLA compliance, priority distribution, resolution type, top categories, temporal seasonality, and automation opportunity surfacing.
• ✨ Resolution workflow UX — shipped resolution detail workflows with step cards, script review, related ticket context, SLA timers, and richer ticket info panels.
• ✅ Expanded regression coverage — added focused tests for ITSM routers/services plus crew board/list/timeline and query hook scenarios touched by reliability work.

Changed

• ♻️ Crew run reliability hardening — improved backlog/event orchestration, split activity/board regions for clearer runtime behavior, and strengthened recovery paths around run execution state.
• ♻️ Spec custom-section ergonomics — surfaced custom sections in settings and tightened section handling to align editor behavior with section configuration workflows.
• 🔧 Feature-flag consistency across workers — refreshed stale cache behavior so feature toggles stay coherent across background processes.
• 🔧 Release metadata refresh — bumped desktop shell, backend sidecar, Rust/Tauri metadata, and release badge references for the new alpha line.

Fixed

• 🐛 Crew execution reconnect loops — fixed restart-loop behavior when execution streams disconnect mid-run.
• 🐛 Stale run/execution state mismatches — reconciled stale execution status with run configuration to avoid misleading runtime state.
• 🐛 Review and URL race regressions — resolved crew review conflict false positives, URL race edge cases, and run parallel-config restoration gaps.
• 🐛 Spec section duplication edge case — prevented duplicate creation of standard Spec Copilot sections.
• 🔒 Security hardening follow-up — resolved remaining high and medium-priority Aikido findings (#284, #286).

Added

• ✨ Code Insights unified diagrams — unified architecture diagrams with repo viz workspace for a single cohesive visualization experience (#282).

Changed

• ♻️ Spec Copilot identifier contract — renamed spec payload identity from id to specId, standardized spec persistence around specId while keeping MongoDB default _id behavior, and updated API/types/indexes/migration tooling for single-cutover compatibility (#275).
• 🔧 Licensing metadata alignment — aligned proprietary licensing metadata and removed Nuitka tooling and dependencies (#279).

Fixed

• 🔒 Critical security fixes — resolved security issues identified by Aikido SAST scan (#283).
• 🐛 Crew run monitoring regressions — fixed stream-first crew run monitoring issues that could cause stale run state.
• 🐛 Cross-platform shortcut handling — fixed shortcut handling and labeling inconsistencies across platforms.
• 🐛 Spec import completion timing — show spec import completion notification only when recent to avoid stale toasts.
• 🐛 Tailwind content scan — excluded test files from tailwind content scan to avoid build noise.

Removed

• 🔥 Nuitka tooling — removed Nuitka build tooling, dependencies, and references (#278).

Added

• ✨ Crew multi-run lineage history — added ancestry-aware run tracking and lineage metadata so teams can follow chained runs, branch from prior outcomes, and keep execution context aligned across iterations.
• ✨ Dedicated Crew run settings surface — introduced a standalone run settings route with reusable form utilities, stack-tag selection, and tech stack suggestion data to make configuration edits safer and faster.
• ✨ Artifact placement hardening service — added crew_artifact_placement_service and related model/index updates to make stage-aware artifact placement deterministic across multi-run and guided flows.
• ✨ Context explorer grounding — introduced backend context exploration support to improve repo/context grounding for Crew spec-to-backlog and coder execution pipelines.
• ✅ Expanded regression coverage — added and extended router, service, hook, and page tests covering lineage, artifact placement, spec import jobs, settings UX, and source-selection behavior.

Changed

• ♻️ Crew run UX architecture — refactored run header/controls/cards and split settings concerns into dedicated components and helpers, reducing coupling in the Crew run experience.
• ♻️ Spec import execution flow — improved import progress ramp behavior and now triggers immediate backlog refresh after import milestones to keep run state current without manual reloads.
• ♻️ Tech stack capture and review flow — improved wizard detail/review steps and run settings synchronization so selected stack context is carried more consistently into execution.
• 💄 New-run/import source selection UX — improved source selector behavior and dialog layout for starting runs from imports with clearer affordances and lower setup friction.

Fixed

• 🐛 Stale backlog after import events — resolved delayed backlog state updates that could leave recently imported items invisible until later polling.
• 🐛 Crew artifact routing edge cases — fixed fragile artifact placement paths that could misclassify or drop outputs in complex multi-run scenarios.
• 🐛 Run configuration drift between UI and backend — fixed mismatches between edited settings payloads and backend run/simulation handling in Crew routes.

Added

• ✨ Git Files tree view — tree view and syntax-colored diff display in Git Files panel
• ✨ CodeInsights performance mode — configuration presets for performance optimization in code analysis
• ✨ CodeInsights token usage detail — detailed token usage display with new table layout in OverviewTab

Changed

• ♻️ Crew artifacts list+detail split view — refactored Crew artifacts to a list and detail split view layout
• 💄 Crew review flow — show artifact before approve/reject in Crew review flow
• 🔧 SQL artifacts and docs generation — updated SQL artifacts configuration and enhanced documentation generation
• 🔧 LLM service improvements — improved token handling and concurrency limits for standalone docs generation
• 🔧 Session knowledge capture — added /done session knowledge capture command
• ⬆️ peerai-tools update — updated peerai-tools to version 2026.2.18.1

Fixed

• 🐛 Crew pause/resume review — fixed crew pause/resume review UX and artifact pending status
• 🐛 Review approval updates — allow review approval updates without active simulation

Added

• ✨ Crew Git Files panel — reusable Git status/diff panel for Crew runs with backend git status and diff APIs, full untracked file listing, and responsive panel sizing
• ✨ Guided crew execution — hardened guided submit UX with clearer activity signals, git repo persistence across crew run sessions

Fixed

• 🐛 Licensing server feature list — added peerai-crew and company-insights to the licensing server feature list so tier gating works correctly for these features

Added

• ✨ Spec Copilot section-level generation — stream section content preview, per-section generate/clear buttons, and CopilotAction integration for targeted AI generation
• ✨ Spec section toggle list — configure visible spec sections from Settings tab
• ✨ AI-driven artifact chat — validation-context issue triage and targeted fixes for Deployment Copilot artifacts
• ✨ Per-SQL-file documentation generation — migration-ready standalone docs for CodeInsights SQL artifacts
• ✨ CodeInsights configuration presets — performance optimization presets for code analysis workflows
• 🔧 Optional dependency check in run.sh — interactive detection and install of PDF export libs, repo visualizer tools, deployment linters, and Neo4j

Changed

• ♻️ Crew Run UX polish — decomposed page into sub-components, hardened import/export, consolidated state management, added run reset and saved board filter views
• 💄 Repo Insights dashboard revamp — improved results dashboard UX and LOC attribution
• ♻️ CodeInsights Ask improvements — optional vectorization, security hardening, dead code removal, and timeout handling

Fixed

• 🐛 Spec editor stability — clipboard paste markdown normalization, turndown marker unescaping, credit limit recovery, add-block reliability, and duplicate content deduplication
• 🐛 Spec clear-all preserves structure — optimistic update keeps section headings, only empties content
• 🐛 Spec apply flow reliability — canvas list rendering, specs list count/delete consistency
• 🐛 CodeInsights doc paths — strip redundant root folder and resolve relative to user home
• 🐛 Zizmor obfuscated expression warnings — fixed in disabled workflow files
• 🐛 Branding blob URLs — CSP defense-in-depth for co-brand logos

Added

• ✨ PeerAI Crew major expansion — delivered Crew Phases A-G with the 7-column SDLC board, async spec-to-backlog jobs, live activity/streaming updates, per-run provider presets, analytics, templates, and stronger end-to-end regression coverage.
• ✨ Homepage journey revamp — replaced the old home experience with a hybrid journey dashboard, visual step progression, and refined modernization flow.
• ✨ Licensing admin operations — added license admin dashboard flows plus a detail endpoint to support license support/debug workflows.

Changed

• ♻️ Crew UX and architecture hardening — split Crew model presets into dedicated settings, improved import/export ergonomics (including native Tauri save dialog), and refactored simulation/event internals for safer partial updates.
• 🔧 AWS Bedrock model/config sync — updated Bedrock model mappings and improved API key synchronization behavior.
• 🔖 Version bump to 0.62.0-alpha.1 across desktop shell, frontend, Python sidecar metadata, and release badges.

Fixed

• 🐛 Crew import and board reliability — fixed streaming progress/count accuracy, explicit sequencing visibility, 60s SSE timeout drop-offs, scrolling/dependency navigation regressions, and select-mode checkbox overlap in run cards.
• 🐛 UI polish regressions — fixed canvas code-block hover visibility and improved co-brand runtime text mode consistency.

Added

• ✨ Company Insights Deep Agent — generates comprehensive business intelligence dossiers from a company URL using 5 specialized subagents (website analyzer, market researcher, company profile researcher, news/funding researcher, report consolidator)
• ✨ web_fetch tool — general-purpose web page content extraction with graceful HTTP error handling, internal link discovery, and structure-preserving text extraction for agent use
• ✨ Company Insights UI — full-page React component with CRUD management, SSE streaming progress indicators per subagent stage, and markdown report rendering

Changed

• ♻️ PeerAI Crew wizard — improved usability of run creation wizard and settings edit flow
• 💄 Splash screen — removed white box around logo on splash/activation screens

Added

• 📝 Tauri CSP licensing lesson learned — documented the Cloud Run/custom-domain connect-src requirement in developer lessons learned to prevent repeat release regressions.

Changed

• 🔖 Version bump to 0.61.2-alpha.2 across desktop shell, frontend, Python sidecar metadata, and release badges.

Fixed

• 🐛 Hosted licensing API blocked by desktop CSP — expanded Tauri connect-src allowlist to include both https://*.a.run.app and the planned routed domain https://license.peerai.peerislands.io, unblocking remote /api/v1/licenses/* calls in alpha builds.

Added

• 📝 Licensing Cloud Run deployment handoff — added an operations runbook with exact deploy commands, failure diagnostics, zsh-specific secret input gotchas, desktop test setup, and recovery checklist for future maintainers.
• 📝 Deployment lessons learned — captured the first Cloud Run licensing rollout pitfalls (image architecture mismatch, production secret guardrails, Mongo URI formatting) in developer lessons learned.

Changed

• 🔧 Alpha DMG licensing defaults — updated alpha release workflow to bake in production licensing flags by default (VITE_LICENSE_SERVER_URL, VITE_FORCE_PROD_MODE=true, VITE_DEV_MODE=false, VITE_LICENSE_STRICT=true) so distributed alpha builds connect to the hosted licensing API out of the box.
• 🔖 Version bump to 0.61.2-alpha.1 across desktop shell, frontend, Python sidecar metadata, and release badges.

Fixed

• 🐛 Cloud Run deployment reliability runbook — documented and standardized mitigations for OCI manifest incompatibility on Apple Silicon and malformed/missing secret values causing startup/runtime failures.

Added

• ✨ Stable release baseline for v0.60.0 alpha cycle — promotes the latest alpha functionality to stable, including Ask PeerAI (RAG Q&A), Deployment Copilot validation/export/publish, and AI-powered brand import from website URLs.

Changed

• ♻️ Sidecar packaging and release workflow controls — standardized PyInstaller packaging behavior with mode-aware CI controls and local tooling for onedir/onefile workflows.
• ♻️ Startup readiness and warmup flow — keeps the critical startup path fast while optional systems initialize in the background with timeout guards.

Fixed

• 🐛 Packaged app environment loading and API key resolution — fixed .env loading in installed builds and ignored empty env values during key resolution.
• 🐛 Onedir sidecar path and preflight reliability — fixed packaged-app sidecar discovery and improved startup preflight checks for onefile and onedir layouts.
• 🐛 Co-branding logo presentation — constrained co-brand logo sizing in the app header for wide wordmarks.

Added

• ✨ AI brand import from URL — enter a company website URL in Settings > Branding to auto-extract logo, brand colors, and company name using AI analysis of HTML and CSS.
• ✨ run.sh prerequisite checks — added checks for Xcode CLT, Homebrew, Node.js, Python 3.11+, and MongoDB with auto-install support.

Fixed

• 🐛 .env file not loaded in packaged app — ~/.peerai/.env was missing from Pydantic Settings env_file list, so core settings like PEERAI_MONGODB_URL were ignored in Homebrew/DMG installs.
• 🐛 Empty .env values blocking key resolution — API key resolver now skips empty values and logs which .env files are loaded for easier debugging.
• 🐛 Co-brand logo oversized in header — constrained logo slots to h-8 max-w-[140px] so wide wordmarks fit naturally alongside the primary logo.

Changed

• ♻️ Sidecar packaging strategy — defaulted release/alpha CI pipelines to PyInstaller --onedir for faster startup after install, while keeping onefile as an explicit workflow override.
• ♻️ Release workflow controls — added sidecar_mode workflow-dispatch input (onedir|onefile) with mode-aware cache keys/artifacts to avoid cross-mode cache pollution.
• ♻️ Sidecar build tooling — unified src-python/scripts/build-sidecar.sh with --mode support and added local helper scripts for onedir staging/installation tests.

Fixed

• 🐛 Packaged app sidecar discovery — release runtime now resolves onedir sidecar payloads from both Contents/MacOS and bundled Resources paths with onefile fallback.
• 🐛 Preflight sidecar checks — startup preflight now validates onefile and onedir layouts correctly to avoid false missing-binary diagnostics.
• 🐛 PyInstaller build noise — reduced misleading missing-module warnings by tightening hidden imports and adding custom hook overrides for LangChain partner package middleware scanning.

Changed

• ♻️ LLM model naming consistency — standardized model naming/version references across providers to improve settings alignment and reduce mismatch across runtimes.

Fixed

• 🐛 Installed Atlas DB reliability — normalized Atlas URLs to ensure explicit peerai_studio_ts DB resolution in packaged runs.
• 🐛 Ask PeerAI Atlas connection tests — reused shared Mongo TLS client options in one-off connection checks to prevent CERTIFICATE_VERIFY_FAILED in installed app flows.
• 🐛 Startup DB readiness UX — improved /ready re-check behavior so DB status recovers without requiring a hard reload.

Added

• ✨ Ask PeerAI — RAG-powered document Q&A with tri-database architecture (metadata, vector, chat). Upload documents (PDF, DOCX, TXT, URLs), chunk and embed them, then chat with an AI assistant over your knowledge base. Priority-based job queue (2 async workers), SSE streaming for chat and job progress, collection management, suggested questions, and Exa search integration. 9 backend services, 6 frontend hooks, 8 UI components
• ✨ Deployment Copilot: Validation, Export, and Publish — artifact validation with 5 external linters (hadolint, yamllint, shellcheck, actionlint, kubeconform) plus heuristic fallbacks for 30+ artifact types. Linter setup popover with install commands. Auto-validation on content change. Secure ZIP export with file-save dialog before download. GitHub PR publishing with safe branch workflow and cleanup on failure. Path traversal protection on publish (#214)
• ✨ API Tester catalog generation — generate API Tester endpoint catalog from FastAPI OpenAPI spec with curated overrides and drift checking. Scripts: bun run api-catalog:generate and bun run api-catalog:check
• ✨ PyInstaller resource bundling — centralized resource_path() utility (core/resource_path.py) handles sys._MEIPASS detection for bundled YAML templates, agent configs, and skill files across 6 consumers
• ✨ Engineering notes — research snippets for ongoing architecture decisions

Changed

• ⚡ Ask PeerAI performance — optimized session list responsiveness with lightweight projections, progressive job loading, camelCase schema writes throughout
• ⚡ CI/CD optimization — merged lint+tests into single job, added path filters for selective checks, sidecar build caching (#219)
• ♻️ Startup improvements — 14-step startup timeline with deferred optional services (Ask DB, analytics, job queue, agent hub, indexes, plugins, default provider). Background warmup with per-step timeouts. Critical path completes first for fast /ready response
• ♻️ camelCase field normalization — 3-phase pipeline for MongoDB writes (Phase 1: remove snake_case fallbacks, Phase 2: field-map normalization, Phase 3: canonicalize Python writes). Covers product runs, versions, playground runs, deployment artifacts
• ♻️ Domain service extraction — delete dead code, fix stats + API consistency, extract domain services from fat routers
• ♻️ Fix sidecar terminology — corrected all misleading "TypeScript backend" references to "Python backend" in Tauri config, Rust code, CI/CD, and docs
• ♻️ Restore CI/CD workflows — all 3 workflow files restored to match origin/alpha byte-for-byte
• ♻️ Markview performance — chunked rendering, lazy Mermaid/Markmap via IntersectionObserver, partial preview for large files, serial Mermaid queue
• ♻️ Spec Copilot performance — chunked section rendering, debounced section edits, partial MongoDB updates ($set, $push, $pull)
• ♻️ MongoDB list projections tightened — heavy list endpoints exclude large fields across Python backend

Fixed

• 🐛 Ask PeerAI — chat DB routing, camelCase schema writes, session loading reliability, query read performance, data/schema compatibility, progressive job loading, Docs KB status errors, streaming/session sync
• 🐛 PyInstaller resource bundling — prompt YAML templates, agent configs, and skill files now included via --add-data flags
• 🐛 Playground runs — use camelCase MongoDB fields consistently
• 🐛 Self-heal prompts text index conflicts during startup
• 🐛 Repo visualizer preview flicker during history selection
• 🐛 Mermaid rendering — hardened cleanup and loader recovery across frontend and backends
• 🐛 SkillResponse crash on empty body from MongoDB
• 🐛 Data model and migration runs — list response shape handling
• 🐛 Testdatagen build — missing lifetime specifier in ensure_workload (Rust E0106)

Fixed

• 🐛 Fix PyInstaller resource bundling — prompt YAML templates, agent configs, and skill files now included in release binary via --add-data flags and centralized resource_path() utility
• ♻️ Centralized resource path resolution — new app/core/resource_path.py handles sys._MEIPASS detection for all 6 consumers (CodeInsights prompts, agent configs, skill loader, test data router)

Fixed

• 🐛 Fix testdatagen build — add missing lifetime specifier in ensure_workload function (Rust E0106)
• 🔧 Update uv.lock for 0.53.0-alpha.1 version bump

Added

• ✨ TS Backend Experiment — full Next.js/Bun TypeScript backend (apps/studio/) with 42+ routers matching Python sidecar CRUD parity, LLM abstraction, plugin system, agent runtime (ON HOLD)
• ✨ API Tester catalog — generate API Tester catalog from OpenAPI with drift checks
• ✨ Spec Copilot enhancements — structure-first workflow, project linkage, spec-level generate/regenerate, section apply flow improvements
• ✨ Ask PeerAI improvements — source processing pipeline, docs KB reindex, tri-database chat architecture, startup reconciliation
• ✨ Performance metrics dashboard — request/DB performance metrics with settings dashboard

Changed

• ♻️ Backport TS backend improvements to Python sidecar — camelCase field normalization (3-phase pipeline), MongoDB list projection tightening, field-map normalization
• ♻️ Fix sidecar terminology — corrected all misleading "TypeScript backend" references to "Python backend" in Tauri config, Rust code, CI/CD, and docs
• ♻️ Restore CI/CD workflows — all 3 workflow files restored to match origin/alpha byte-for-byte, Python backend-tests job restored
• ♻️ Startup improvements — split critical startup from optional warmup, instrument startup timeline, streamline API-key sync
• ♻️ Performance optimizations — fire-and-forget metrics, indexes, TTL, O(N) SSE accumulators, $regex fallback search, marketplace polling flag
• ♻️ Markview and Spec Copilot performance — optimized for large markdown viewers and large specs

Fixed

• 🐛 Self-heal prompts text index conflicts during startup
• 🐛 Repo visualizer preview flicker during history selection
• 🐛 Mermaid rendering — shared hook, render queue, stale state, timeout fixes
• 🐛 Ask PeerAI — streaming/session sync, source citations NaN%, docs KB path resolution
• 🐛 Agent Playground — NaN display, guardrail history, stream status
• 🐛 Docs KB — progress percentage display, reindex timeout, setup visibility
• 🐛 Data model and migration runs — list response shape handling
• 🐛 SkillResponse crash on empty body from MongoDB

Added

• ✨ CodeInsights clustering enhancements — improved module clustering with first-level clustering prompts and upgraded peerai-tools package (#210)

Changed

• ♻️ Refactor mermaid rendering — serialized render queue in useMermaidRenderer hook prevents concurrent render corruption; consolidated MermaidDiagram to shared hook; added bar chart → xychart-beta normalization for mermaid v11; Vite alias pointed at mermaid.esm.mjs
• ♻️ Migrate product mermaid components — JourneyMapViewer, RoadmapTab, and StructureTab migrated from independent mermaid instances to shared useMermaidRenderer hook (-220 lines)
• 🗑️ Remove unused LLM providers — deleted Azure OpenAI, Cohere, and Hugging Face options from LLMFactory
• 🗑️ Clean up temp_repos — removed deprecated subprojects and added temp_repos/ to .gitignore

Fixed

• 🐛 Fix repo visualizer mermaid — reset streaming state when selecting historical jobs to prevent stale mermaidCode from previous sessions; extracted cleanMermaidCode to shared utility with 25+ diagram type headers; added 15s render timeout to prevent infinite loading on cold start

Fixed

• 🐛 Fix CodeInsights Documentation tab scroll — content and tree view were clipped instead of scrolling. Root cause: AppLayout's p-6 wrapper padding caused h-full children to overflow by 48px, clipped by overflow-hidden. Added CodeInsights detail route as a full-bleed layout (p-0)
• 🐛 Fix flex column height chain — added min-h-0 throughout the Tabs → TabsContent flex chain to allow proper shrinking below content size
• 🐛 Remove -m-6 negative margin hacks from DocsTab now that outer padding is removed

Changed

• 💄 Group header action buttons — primary actions (Start/Save) separated from secondary actions (Pipeline/Activity/Publish/Delete) with a vertical separator
• 💄 Add subtle tab content background — bg-muted/30 on tab content area and bg-background/80 on tab bar for visual separation
• 💄 Add documentation module count badge on the Documentation tab trigger

Added

• ✨ All LLM creation respects AI Models settings — every feature now uses the user's configured default provider/model from Settings > AI Models (#208)
• ✨ Code Insights Ask tab — conversational Q&A over analyzed codebases with RAG-powered responses (#181)
• ✨ Skills node in Agent Builder — drag-and-drop Skills node for deep agents with left-side input handle (#204, #205)
• ✨ Skills Management UI — browse, filter, and manage agent skills with grid/list views and database-backed storage
• ✨ User skills and Development Assistant agent — custom user-defined skills with a built-in Code Assistant deep agent
• ✨ Hub (Marketplace) — unified marketplace service with UI revamp, seed data, favicon, and production Vercel deployment (Phase 10)
• ✨ External Plugins — install/uninstall with permissions model and Settings UI (Phase 8)
• ✨ Internal Plugins — integration glue with SRE Copilot plugin (incident CRUD, AI analysis, runbook generation) (Phase 6)
• ✨ Agent SDK — standalone plugin development package (Phase 5)
• ✨ Plugin foundation — manifest, loader, registry, and slots system (Phase 4)
• ✨ Markdown export for GenericArtifactEditor
• ✨ Enhanced content normalization in markdown node generation

Changed

• 💄 Rebrand Marketplace to Hub in desktop Studio UI with improved logo, header spacing, and styling
• ♻️ Refactor skills to database storage with CrudService pattern
• ♻️ Phase 2: Extract service layer + agent type registry
• 🔧 CI/CD hardening — coverage gates, security scanning, bundle tracking (Phase 7)

Fixed

• 🐛 Fix ImportError — remove non-existent PluginDependencies import (#202)
• 🐛 Fix sidecar build — port config, tkinter exclusion, output path detection
• 🐛 Fix Vercel data loading — use direct DB queries instead of self-fetch
• 🐛 Fix deployment copilot infinite PUT/GET loop
• 🐛 Fix await on sync unload_plugin + auto-grant manifest permissions

Added

• ✨ Architecture diagrams — LLM-powered C4 diagrams from codebase analysis with streaming generation. Supports multi-diagram output (Context, Container, Component)
• ✨ Bubble maps — D3 circle-packing visualization of file tree (sized by bytes, colored by extension). Click directories to zoom, scroll wheel to pan/zoom
• ✨ Unified zoom/pan controls — react-zoom-pan-pinch for all diagram types with zoom in/out buttons, reset, fit-to-view
• ✨ Configurable default LLM — Settings > AI Models now lets you select default provider & model for all features. Stored in backend settings collection
• ✨ LLM settings resolver — llm_settings_resolver.py centralizes provider/model resolution across all services with proper fallback chain

Fixed

• 🐛 Fix bubble map blank screen — unified to single zoom system (was dual D3 + react-zoom-pan-pinch transforms conflicting)
• 🐛 Fix C4 diagram rendering — robust fence stripping, handle multi-diagram output, better prompts
• 🐛 Fix TooltipProvider crashes — added missing providers in several components
• 🐛 Fix architecture diagram API key resolution — proper cache → env var fallback chain

Changed

• ♻️ Migrate repo_analysis_service to get_configured_llm — uses centralized LLM settings instead of hardcoded defaults

Added

• ✨ Backend reconnect without restart — when the Python sidecar health check times out, the process is kept alive instead of killed. The app auto-detects when the backend comes up via 2-second polling — no page reload needed. If the user clicks "Continue Without Backend", the SafeModeBanner auto-reconnects in the background and shows a toast when the backend is ready
• ✨ "Check Connection" button — new manual health check available on the splash screen (timeout state), SafeModeBanner, and Settings > System Status Backend card. Invokes get_sidecar_status directly for instant feedback
• ✨ Process-alive detection — get_sidecar_status now verifies the child process via try_wait() before checking health. If the sidecar process exited, port/PID state is cleared immediately — prevents ghost state where the UI thinks the backend is "starting" forever
• ✨ notify_sidecar_connected Tauri command — frontend notifies Rust when the backend comes up after a timeout, updating the system tray from "Starting (slow)..." to healthy green
• ✨ Richer SafeModeBanner — two visual states: blue spinner with "Backend starting — will reconnect automatically" when a port is assigned, amber warning with "Backend unavailable" when no port. Both use context methods instead of window.location.reload()
• ✨ Backend card actions in System Status — "Check Connection" button when backend is not ready, refresh button when healthy. Consistent with OpenCode/TestDataGen action patterns

Changed

• ♻️ retryConnection() no longer reloads the page — clears safe mode + errors and lets existing health polling handle reconnection naturally
• ♻️ sidecar-timeout event replaces sidecar-failed for health check timeouts — actual failures (binary missing, spawn crash) still emit sidecar-failed with full diagnostics
• 💄 Settings sidebar grouped into 6 logical sections — General, AI & Models, Development, Data, Integrations, System — with collapsible section headers for better navigation at scale

Added

• ✨ Startup debug panel — press Cmd+D during splash to see preflight timeline with timestamps and backend startup steps
• ✨ Structured sidecar failure reporting — SidecarFailure includes stderr tail (last 30 lines), log path, and copy-friendly recovery commands
• ✨ /startup-status endpoint — tracks backend startup steps with timing (sentry, templates, mongodb, etc.)
• ♻️ ResourceListPage<T> generic component — shared search, filters, view toggle, and virtualization for list pages (~665 lines removed from Prompts, Tools, Guardrails, Ontologies)
• ♻️ Shared process.rs and sidecar_manager.rs — extract ~700 lines of duplicated port/kill/health logic from 3 sidecar modules into shared Rust utilities
• ♻️ Shared core/json_parser.py — 4-tier LLM JSON extraction strategy replacing 3 duplicate implementations
• ♻️ Shared core/filters.py — text search + tag filtering utilities used by 7+ routers
• ♻️ Shared models/common.py — Visibility, EntityStatus enums + BaseEntityResponse + TimestampMixin replacing 17 duplicate enum definitions
• ♻️ routers/integrations.py — Neo4j/Sonar endpoints extracted from main.py (~390 lines)
• ♻️ Component splitting — CodeInsightsDetail (1984→340 lines), AgentEditorPage (1002→570), DeploymentEditorPage (749→491) split into tab/toolbar sub-components
• ♻️ lib/config/status-colors.ts — centralized entity status color mapping
• ♻️ SSE type guards — isToolResultData, isErrorData, isDoneData with property checks

Changed

• ♻️ useState → useReducer for complex forms: ArtifactStoragePage (10→1), DatabasePage (14→1), use-agent-streaming (7→1) — batched multi-field updates
• ♻️ Stable keys everywhere — replaced key={index} with stable keys across 28 components
• ♻️ Stale closure fixes — useRef+useEffect pattern applied across 19 pages
• ♻️ Remove 22 backward-compat enum aliases — all imports use canonical Visibility/EntityStatus from models.common
• ♻️ Tray quit stops all 3 sidecars — was only stopping Python, now also stops OpenCode and TestDataGen
• 📝 Updated all CLAUDE.md files, FOR-DEVELOPERS.md, and README.md to reflect new architecture

Fixed

• 🐛 Fix Python sidecar not killed on Cmd+Q / window close — sidecar cleanup was fire-and-forget (spawn() exited before kill completed). Now intercepts window close, runs cleanup to completion, then exit(0). Port-based kill in RunEvent::Exit as safety net
• 🐛 Fix greedy JSON regex — [\s\S]* → [\s\S]*? in json_parser.py (was merging multiple JSON objects into one malformed blob)
• 🐛 Fix dangling timer in useAutoSave — track statusTimeout ref to prevent setState on unmount
• 🐛 Fix componentStack not passed to Sentry in ErrorBoundary

Fixed

• 🐛 Fix splash screen stuck when TermsGate blocks AppLoader on fresh installs — move splash dismissal to App-level useEffect so it fires regardless of gating components

Changed

• ⚡ Cache testdatagen sidecar binary to skip redundant builds during development
• 🔧 Disable TermsGate until EULA is finalized — removes the terms acceptance gate from the app startup flow

Fixed

• 🐛 Fix app stuck at initializing when initial /ready call fails — add retry logic every 5s to re-fetch database status, fix spurious status change callbacks from Rust/frontend field mismatch
• 💄 Show "Continue anyway" button after 30s stuck at any loading stage (database, license, api-keys), not just on error

Changed

• 📝 Document PyInstaller freeze_support requirement for sidecar builds

Fixed

• 🐛 Fix duplicate sidecar spawn in PyInstaller builds — add multiprocessing.freeze_support() and Rust-side guard to prevent second server start

Changed

• ⚡ CI/CD caching optimization — add Bun and uv dependency caching across all workflows
• 🔧 Disable branch gate checks — allow PRs to beta/main from any branch

Changed

• 🔧 Release workflow updates alpha cask for pre-release tags
• 🔧 Alpha cask installs as "PeerAI Studio Alpha.app" to coexist with stable

Added

• ✨ Project Code Insights integration (#154)
• ✨ MCP execution steps with timing info in tool responses
• ✨ Configurable MCP endpoint setting for stable IDE configs
• ✨ Expose PeerAI agents as MCP servers for external IDEs (Cursor, etc.)

Fixed

• 🐛 Fix startup crash when MongoDB requires authentication (#157)
• 🐛 Fix MCP tool response duplication
• 🐛 Fix subagent loading for MongoDB-stored deep agents
• 🐛 Fix MCP protocol responses for Cursor compatibility
• 🐛 Fix dialog content overflow with long API keys
• 🐛 Fix Cursor deeplink and dialog overlap issues
• 🐛 Fix CORS preflight 500 errors from missing bcrypt dependency

Changed

• 🔧 Sidecar prefers port 8765 for stable MCP configs
• 💄 Improve MCP connection setup dialog UX
• ♻️ Revert multiple sequential subagent delegations

Fixed

• 🐛 Splash screen blocked by strict CSP on managed machines - Moved inline <style> and style= attributes to external CSS file. GlobalProtect and similar enterprise tools inject CSP that removes unsafe-inline, blocking the splash screen. Requires rebuild to take effect.

Fixed

• 🐛 CSP blocking Tauri plugin IPC on locked-down machines
  • Added blob:, data:, asset: to style-src for dynamic stylesheets
  • Added blob: to img-src for blob URL images
  • Added explicit ipc://localhost to connect-src for strict CSP enforcement
  • Fixes notification plugin and stylesheet loading on enterprise machines

✨ Highlights

The "Connect Everything" Release - MCP (Model Context Protocol) integration lets you connect external tools and services to your agents. Code Insights now speaks PHP and SQL, exports to Neo4j for graph analysis, and vectorizes code for semantic search. The Deployment Copilot gains five new specialized subagents. Plus, we swapped CodeMirror for Monaco in the artifact editor.

Added

• ✨ MCP Integration (#137)

  • Connect external tools via Model Context Protocol servers
  • New Settings → MCP Servers configuration page
  • MCP Hub with curated server catalog
  • Agents can invoke MCP tools alongside built-in tools

• ✨ Code Insights: PHP & SQL Support (#141)

  • Parse and analyze PHP codebases
  • SQL file analysis and documentation
  • Neo4j graph export for code relationships
  • Atlas Vector Search integration for semantic code search

• ✨ Deployment Copilot Subagents (#128)

  • GitOps Generator - ArgoCD, FluxCD configurations
  • Monitoring Generator - Prometheus, Grafana, alerting
  • Security Generator - RBAC, network policies, secrets
  • Scripts Generator - deployment automation scripts
  • Pipeline View drawer for orchestration visibility

• ✨ Test Data Generator Build Pipeline

  • Rust sidecar now built in CI/CD workflow
  • Bundled with app distribution

Changed

• ♻️ ArtifactEditor: CodeMirror → Monaco
  • Better syntax highlighting and theming
  • Improved performance for large files
  • Consistent with VS Code experience

Fixed

• 🐛 CSP Blocking Tauri Plugins

  • Added ipc: and tauri: protocols to Content Security Policy
  • Fixes notification plugin and other Tauri 2 features on locked-down machines

• 🔧 Rust Compiler Warnings

  • Fixed warnings in testdatagen sidecar

✨ Highlights

The "Now You See It" Release - This release fixes the invisible artifact content bug that had users squinting at blank CodeMirror editors. We also silenced the chatty Tauri updater that was spamming logs like an overenthusiastic notification system.

Fixed

• 🐛 Deployment Copilot Artifact Visibility

  • Fixed invisible text in artifact editor - line numbers showed but content was blank
  • Added explicit EditorView.theme() extension with proper text colors
  • Works correctly in both light and dark modes
  • Syntax highlighting now visible for Dockerfile, YAML, HCL, and other file types

• 🔧 Tauri Updater Log Spam

  • Disabled Tauri updater plugin that was flooding logs
  • Cleaner console output for developers and users alike

✨ Highlights

CI Pipeline Polishing - A quick maintenance release that ensures the testdatagen-api sidecar is properly integrated into the CI workflow. This keeps the build pipeline running smoothly and sets the stage for more exciting features ahead.

Fixed

• 🐛 CI Workflow Enhancement
  • Added testdatagen-api placeholder to CI workflow
  • Ensures Rust sidecar builds correctly in the release pipeline

✨ Highlights

Bundled Sidecars Release - The Rust test data generator sidecar is now bundled with the app, eliminating the need to build it separately. Both Python and Rust sidecars are now included in the release pipeline for a complete out-of-the-box experience.

Added

• ✨ Testdatagen Sidecar in Release Pipeline
  • Rust test data generator now bundled with app distribution
  • Parallel sidecar builds in CI (Python + Rust)
  • High-performance local data generation works out of the box

Fixed

• 🐛 TypeScript Build Error
  • Added missing scripts phase to GENERATION_PHASE_CONFIG

✨ Highlights

Agent Powerhouse Release - This release supercharges the agent system with three major features: guardrails that protect your agents from going off-rails, custom tools that let you extend agent capabilities with your own code, and a completely revamped Deployment Copilot with intelligent subagent orchestration. Plus, Code Insights now exports to Neo4j for graph-powered analysis.

Added

• ✨ Guardrails for Agents (#127)

  • Define input/output guardrails in agent YAML
  • Real-time guardrail violation display in Agent Playground
  • Show guardrail names instead of UUIDs in visual builder
  • Label support for guardrails in YAML

• ✨ Custom Tools for Agents (#124)

  • Agents can now use custom tools defined in Tools UI
  • Inline tool calls and thinking displayed in Agent Playground
  • Seed script for sample custom tools and demo agent
  • tool_ref support for library tools in agent YAML

• ✨ Deployment Copilot Enhancements (#128)

  • Sequential subagent orchestration for structured pipeline generation
  • Pipeline View drawer for visualizing generated artifacts
  • Scripts phase with dedicated generator
  • Enhanced artifact detection patterns for GitHub Actions and Terraform

• ✨ Code Insights Graph Export (#123)

  • Neo4j graph export for code relationship visualization
  • MongoDB Atlas Vector Search for code embeddings (replaced CocoaIndex)
  • LlamaIndex CodeSplitter for syntax-aware code chunking

Changed

• ♻️ Simplified Artifact Parsing
  • Explicit language tags for cleaner artifact detection
  • Multi-document YAML parsing support
  • Hyphenated tag support (e.g., github-actions, docker-compose)

Fixed

• 🐛 Deployment Copilot

  • Artifact parsing for multi-document YAML files
  • GitHub Actions and Terraform detection patterns
  • Subagent loading and configuration

• 🐛 Agent Playground

  • DeepAgent event handling for token and done events
  • ReAct agent tool invocation in streaming mode
  • Sample tools handle both UI and agent execution

• 🐛 Guardrails & YAML

  • V2 YAML parser missing guardrail library fields
  • Guardrail node data field names
  • Frontend YAML generation for guardrails
  • Silent failure when MongoDB not connected for vectorization

Documentation

• SSE event type handling checklist (lesson learned)
• Guardrails for agents with test examples
• LangChain streaming and serialization patterns

✨ Highlights

Test Data Generator 2.0 - Major overhaul of the test data generation system. A new Rust sidecar delivers blazing-fast data generation, cloud integration enables remote execution, and dozens of bug fixes make the entire flow rock-solid. Edit completed jobs, create databases on the fly, and enjoy a polished UI with searchable dropdowns.

Added

• ✨ Rust Test Data Generator Sidecar

  • High-performance Rust-based data generation engine
  • Seamless integration with existing Python backend
  • Robust faker patterns with fallback for unknown methods

• ✨ Test Data Generator Cloud Integration (#108)

  • Execute generation jobs remotely
  • Execution mode persistence and routing

• ✨ Edit Job Feature

  • Edit completed test data jobs for re-runs
  • Run history with row selection and visual feedback

• ✨ Database Enhancements

  • "Create new database" option in database selection
  • Searchable combobox for database selection

Changed

• ♻️ Test Data Generator Code Simplification

  • Cleaner, more maintainable codebase
  • Faker method defaults aligned with Rust sidecar

• 🔧 Faker Method Expansion

  • Added emailAddress, commerce.department, and others
  • Fixed jsonTemplate parsing

Fixed

• 🐛 Progress & Status Tracking

  • Progress UI now updates correctly on completion
  • SSE progress field name mapping fixed
  • Job status updates when SSE stream completes
  • Run status reflects latest run data
  • totalInsertedRecords computation corrected

• 🐛 UI & UX Fixes

  • Project link now shows in test data editor
  • WizardLayout remounts properly when run data changes
  • Fixed stale closure in handleStatusChange causing workloads reset
  • Run history row selection visual feedback
  • Total records display in header
  • SelectItem empty value error across all pages (#109)

• 🐛 Backend & Integration

  • MongoDB connection string extraction for Rust sidecar
  • Job transformation to Rust sidecar format
  • Test data generation with missing fakerMethod
  • Background task execution for SSE streaming
  • Product document upload and generation error handling (#110)
  • Allow stopping TestDataGen regardless of managed status

✨ Highlights

Performance & Polish Release - This release focuses on squashing bugs, eliminating performance bottlenecks, and cleaning up technical debt. The Ask PeerAI module now loads instantly instead of making 100+ API calls, Code Insights got a memory optimization, and we consolidated duplicate utilities across the codebase.

Added

• ✨ Code Insights Enhancements
  • Source toggle for documentation selection
  • Enhanced detection features and process mining prompts
  • Better workflow and bottleneck analysis

Changed

• ♻️ SSE Parsing Consolidation

  • New sse-core.ts with shared parsing utilities
  • streaming.ts and sse-parser.ts now use common core
  • ~50 lines of duplicate code eliminated

• ♻️ Relative Time Utilities Consolidation

  • formatRelativeTime and formatRelativeTimeShort centralized in utils.ts
  • format-utils.ts re-exports from central location
  • Single source of truth for time formatting

• 🔧 Build-time Version Injection

  • APP_VERSION now uses Vite's __APP_VERSION__ instead of hardcoded string
  • One less file to update during releases (bump-version.sh now updates 10 files, not 11)

Fixed

• 🐛 N+1 Query Fix for Ask PeerAI (PR #100)

  • Opening Ask PeerAI was making 100+ individual GET /api/v1/ask/sources/{id} requests
  • Now uses pre-loaded data from parent component - zero extra API calls
  • Page loads significantly faster

• 🐛 AbortController Memory Leak (PR #93)

  • Controllers now cleaned up when tasks reach terminal state
  • clearCompleted() also removes orphaned controllers

• 🐛 MongoDB URL Encoding (PR #94)

  • Passwords with special characters (@, :, /) now work correctly
  • Uses encodeURIComponent for auth and URL API for parsing

• ⚡ Code Insights Performance (PR #95)

  • Added useMemo for stream event conversion
  • Prevents expensive recomputation on every render

Security

• 🔒 CVE-2026-24486 - Upgraded python-multipart to 0.0.22
  • Fixes arbitrary file write vulnerability (non-default config only)

✨ Highlights

Database Layer Overhaul - We rebuilt the MongoDB foundation from the ground up. A new base class hierarchy eliminates code duplication, and critical async fixes ensure rock-solid database operations. This is the kind of refactoring that pays dividends for years.

Changed

• ♻️ MongoDB Client Consolidation

  • New base_database.py with BaseMongoDBClient, SingleDBClient, and MultiDBClient classes
  • Shared connection logic, timeouts, and TLS configuration in one place
  • ~120 lines of duplicated code eliminated across 3 database clients
  • Ask PeerAI's dual-database pattern (vector + metadata) properly documented

• 🔒 Branch Protection Workflows

  • Beta branch: Only accepts PRs from alpha (enforced by GitHub Action)
  • Main branch: Only accepts PRs from beta (production gate)
  • Proper release flow: alpha → beta → main

Fixed

• 🐛 PyMongo Native Async Migration (Motor → PyMongo 4.16+)

  • find() - Returns cursor directly, NOT a coroutine (14 occurrences fixed)
  • aggregate() - Returns coroutine that must be awaited (5 occurrences fixed)
  • client.close() - Now async, requires await (15 occurrences fixed)
  • These patterns are now documented in CLAUDE.md to prevent future bugs

• 🐛 Specs API Performance

  • Added caching for specs list and get endpoints
  • Reduces redundant database queries on app startup

Documentation

• 📚 CLAUDE.md - Added PyMongo async gotchas with error messages to look for
• 📚 FOR-DEVELOPERS.md - New "MongoDB Client Consolidation" war story
• 📚 README.md - Updated MongoDB driver info (Motor deprecated May 2025)

✨ Highlights

Project-Based Workflows - PeerAI Coder gets organized! Create projects with dedicated workspaces, persistent sessions, and settings that follow your workflow. Plus, a bunch of fixes that make the whole experience smoother.

Added

• 📁 Project Management for PeerAI Coder

  • Create named projects with dedicated workspaces and sessions
  • Project settings page for backend configuration and working directory
  • Project name displays in editor header - no more "PeerAI Coder" everywhere
  • Session persistence across app restarts

• 📋 Copy to Clipboard

  • One-click copy button on assistant messages
  • Finally, no more manual text selection!

• ⏭️ Skip Button for API Key Sync

  • 5-second timeout on startup API key sync
  • Skip button appears if sync takes too long
  • No more infinite "Syncing API keys..." spinner

Fixed

• 🐛 File Operations Actually Work Now

  • File write/create operations now display content in the editor
  • OpenCode only sends file paths, so we read content from disk after writes
  • Large files (>5MB) gracefully skipped to prevent memory issues

• 🐛 Streaming Display for Thinking/Tool Events

  • Thinking blocks and tool calls now render correctly during streaming
  • Activity log properly groups consecutive tool calls by name

• 🐛 Session Message Persistence

  • Tool outputs now saved with session history
  • Messages survive app restarts without losing context

• 🐛 MongoDB Field Naming Consistency

  • Fixed camelCase aliases in all queries and updates
  • No more duplicate snake_case fields in the database
  • Session API returns proper camelCase field names

• 🐛 Startup Reliability

  • Fixed stuck "Syncing API keys" with timeout mechanism
  • Graceful degradation when backend is slow to respond

✨ Highlights

PeerAI Coder - Your new AI-powered coding companion is here! An IDE-like experience that connects to multiple AI coding backends (OpenCode, Claude SDK, Copilot SDK) with real-time streaming, session persistence, and a slick developer-first UI.

Added

• 🤖 PeerAI Coder - Multi-Backend AI Coding IDE

  • Connect to OpenCode (local), Claude SDK, or GitHub Copilot SDK
  • Real-time SSE streaming with proper buffer flushing - no more batched responses!
  • Session management with chat history, file workspace, and terminal integration
  • Hot-swappable backends - switch between AI providers without losing context

• 💬 Rich Chat Experience

  • Thinking visualization - see the AI's reasoning process unfold
  • Tool call display with collapsible details
  • File attachments with automatic document parsing (PDF, DOCX, images)
  • Markdown rendering with syntax-highlighted code blocks

• 📁 Integrated Workspace

  • File explorer with create/edit/delete operations
  • Monaco-powered code editor with full syntax highlighting
  • Integrated terminal for running shell commands
  • All changes persist to session workspace

• 🎨 IDE-Style Layout

  • Resizable 4-panel layout: chat, file explorer, editor, terminal
  • Activity drawer with real-time event stream and Raw JSON debug panel
  • Session sidebar for managing multiple coding sessions
  • Dark mode optimized for long coding sessions

Fixed

• 🐛 SSE Streaming Buffering
  • Replaced StreamingResponse with sse-starlette's EventSourceResponse
  • Events now flush immediately instead of TCP-batching
  • Responses stream incrementally as they're generated

✨ Highlights

Reliability & Resilience - This release focuses on bulletproofing the core infrastructure. MongoDB connections now retry automatically, the API client handles errors intelligently, and memory leaks are a thing of the past.

Added

• 🔄 MongoDB Connection Retry Logic

  • Automatic retries with exponential backoff (2-10s, up to 3 attempts)
  • Handles transient network failures gracefully
  • New health_check_with_retry() for critical operations
  • Connection pool tuning: min 10, max 50 connections
  • Increased timeouts for better stability (10s server selection, 5s connect/socket)

• 🎯 Typed API Errors

  • New ApiError class with status code information
  • Helper methods: isRetryable(), isClientError(), isServerError(), isNotFound(), isUnauthorized()
  • Exported for consumers who need granular error handling

• 💾 Request Cache Improvements

  • Size limit of 500 entries prevents unbounded memory growth
  • LRU-like eviction: frequently accessed entries stay longer
  • Automatic cleanup of expired entries every 60 seconds
  • Monitoring: size and inFlightCount getters
  • dispose() method for clean test teardown

Changed

• ♻️ SidecarProvider Refactoring
  • Split monolithic 248-line component into focused custom hooks
  • use-sidecar-health-check.ts: Adaptive polling (2s when unhealthy, 30s when healthy)
  • use-api-key-sync.ts: API key synchronization from Keychain to backend
  • Proper async cleanup prevents React warnings on unmount
  • Status changes now propagate for all fields (health, database, version)

Fixed

• 🐛 Smarter Retry Logic

  • Only retry specific 5xx errors: 500, 502, 503, 504
  • Network errors: only retry fetch-related TypeErrors (not programming errors)
  • Timeout errors (AbortError) no longer retried - caller decides
  • 4xx client errors never retried (request is invalid)

• 🐛 Event Listener Cleanup

  • Fixed potential listener leak when component unmounts during async setup
  • Background verification tasks properly cancelled on disconnect

• 🐛 Memory Leak Prevention

  • Request cache no longer grows unbounded
  • Batch eviction (oldest 10%) reduces eviction overhead

✨ Highlights

Doc Bench - A comprehensive documentation quality evaluation tool that uses AI-powered multi-dimensional scoring. Import docs from files, folders, or GitHub, apply 50+ rubric templates, and get actionable recommendations to improve your documentation.

Added

• 📚 Doc Bench - Documentation Quality Evaluation

  • Multi-dimensional scoring across 5 quality dimensions: Completeness, Clarity, Accuracy, Examples, Findability
  • AI-powered evaluation with real-time streaming progress
  • Evidence-based scoring with document citations
  • Prioritized recommendations with effort estimates

• 📁 Flexible Document Import

  • Upload individual files (Markdown, HTML, PDF, RST, TXT, DOCX, PPTX)
  • Bulk folder import with recursive file discovery
  • GitHub repository import with branch selection
  • Automatic document tree visualization

• 🎯 50+ Rubric Templates

  • Framework-specific: React, Angular, Vue, Django, FastAPI, Spring Boot, and more
  • Architecture patterns: Microservices, REST API, GraphQL, Event-Driven
  • Documentation types: API Reference, SDK, CLI Tools, Internal Docs
  • AI-generated rubrics based on your document content

• 📊 Rich Results Dashboard

  • Radar chart visualization of quality dimensions
  • Hierarchical rubric score tree with expand/collapse
  • Evidence panel with document excerpts and relevance indicators
  • Recommendations panel with priority filtering

• 🔄 Evaluation Features

  • Incremental persistence (progress saved after each rubric)
  • Re-run evaluation on completed assessments
  • View progress even after closing the drawer
  • Export results as PDF, Markdown, or JSON

Changed

• Consistent UI dialogs using shadcn AlertDialog across all features

✨ Highlights

AI-Powered Agent Builder - Describe what you want in plain English, and our meta-agent builds the agent for you. Watch the magic happen in real-time with the new Activity Drawer.

Real-Time Activity Drawers - New timeline views showing every step of agent building and execution. See tool calls, subagent delegations, and streaming events as they happen.

Added

• 🤖 AI Agent Builder with Meta-Agent Backend

  • Describe your agent in natural language and let AI build it
  • Real-time streaming of the build process via SSE
  • Automatic provider selection based on configured API keys
  • Support for simple, react, and deep agent architectures
  • New /api/v1/agent-builder/generate endpoint with streaming

• 📊 Agent Builder Activity Drawer (BuilderActivityDrawer.tsx)

  • Real-time timeline of meta-agent build progress
  • Status badges showing phase completion (analyzing, designing, generating)
  • Collapsible sections with detailed event data
  • Auto-scroll on new events with smooth animation

• 📊 Agent Playground Activity Drawer (PlaygroundActivityDrawer.tsx)

  • Timeline view of all SSE events during execution
  • Color-coded event types: start (blue), tool calls (purple), subagents (orange), complete (green), error (red)
  • Expandable details for tool arguments and results
  • Clear button to reset activity log

• 📈 Enhanced Execution Metrics

  • Token counts (input/output/total) displayed in metrics panel
  • Iteration count for react agents
  • LLM call count tracking

• 🎨 New Hooks for Provider Management

  • useConfiguredProviders - detect which AI providers have API keys
  • useAvailableModels - fetch models for configured providers
  • useAgentBuilder - orchestrate meta-agent communication

Fixed

• 🐛 Agent Execution 422 Error - Fixed API key resolution for agent execution
• 🐛 Streaming Tokens Not Displaying - Tokens now stream progressively instead of appearing all at once
• 🐛 Agent Playground Draft Agents - Draft agents now appear in the agent selector
• 🐛 Agent Builder "Test" Button - Fixed "Agent not found" error when testing newly created agents
• 🐛 RunCard Icon Overlap - Fixed star/delete icons overlapping with agent name in history sidebar
• 🐛 Activity Drawer Auto-Scroll - Fixed auto-scroll not working on new events
• 🐛 Unsaved Changes Dialog - Fixed timing issues with dirty state detection
• 🐛 False Dirty State on Load - Agent no longer shows unsaved changes immediately after loading

Changed

• ♻️ YAML-Only Agent Storage - Agents now stored as YAML source of truth, improving portability
• 💄 Agent Builder Wizard UI - Improved layout and visual feedback during generation
• 📝 Documented Tauri Dialog Gotcha - Added warning about async window.confirm() behavior in Tauri

Technical Details

The meta-agent uses a multi-phase approach:

1. Analysis - Understands requirements and determines agent type
2. Design - Plans tools, prompts, and architecture
3. Generation - Produces valid YAML configuration
4. Validation - Ensures generated config is syntactically correct

Activity events follow the pattern established in ADR generation, providing consistent UX across features.

🔧 Stability & Resource Management

This patch release addresses critical process lifecycle issues that could lead to orphaned background processes consuming system resources after quitting the app.

Fixed

• 🐛 Orphaned Python Sidecar Processes (src-tauri/src/sidecar.rs, lib.rs)

  • Fixed zombie python-backend processes (~185MB each) remaining after app quit
  • Root cause: PyInstaller's 2-process architecture (bootloader + actual Python) wasn't being fully terminated
  • Solution: Kill process tree properly - children first, then parent, then process group
  • Added RunEvent::Exit handler for reliable cleanup even on abrupt termination
  • Port-based fallback using lsof ensures surgical cleanup without affecting unrelated processes
  • Dev mode: Pattern-based kill with port number (uvicorn.*--port <PORT>)
  • Windows: Now uses taskkill /T flag for proper process tree termination

• 🐛 Stale Data After SSE Streaming (various components)

  • Fixed UI not refreshing with latest data after streaming operations complete
  • Data now properly invalidates and refreshes when SSE streams finish

• 🐛 DocsTab Legacy Run Documentation (DocsTab component)

  • Fixed documentation not displaying for legacy/older run formats
  • Improved backward compatibility with historical data

Technical Details

The sidecar cleanup now follows this sequence:

1. pkill -TERM -P <pid> - Kill children first (prevents orphaning)
2. kill -TERM <pid> - Graceful shutdown to parent
3. kill -TERM -<pid> - Kill entire process group
4. Wait up to 3s with verification loop checking both parent AND children
5. kill -KILL for any remaining processes
6. Port-based lsof fallback for stragglers

✨ Highlights

Performance Overhaul: 70% Faster Startup, 40% Smaller Bundle - This release completes Phase 2 of our optimization roadmap, delivering dramatic improvements across the entire stack. Route-based code splitting now lazy-loads all 78 pages, backend routers load on-demand, MongoDB queries use proper text indexes, and heavy visualization libraries are dynamically imported. The result? A snappier, more responsive PeerAI Studio.

Performance

• ⚡ Route-Based Code Splitting (src/App.tsx)

  • Implemented lazy loading for all 78 pages using React.lazy + Suspense
  • Initial bundle reduced by ~40% - only loads code for current route
  • PageSkeleton fallback provides instant visual feedback during navigation
  • Dynamic imports for all route components with proper chunk naming

• ⚡ Backend Lazy Router Loading (src-python/app/main.py)

  • Only health and logs routers loaded at startup
  • Remaining 20+ routers load on first non-health request
  • Startup time reduced by 2-3 seconds
  • Middleware-based lazy initialization pattern

• ⚡ MongoDB Text Search Indexes (src-python/app/core/index_manager.py)

  • Added compound text indexes for prompts, agents, tools, guardrails, ontologies
  • Query performance improved 50-70% for search operations
  • Eliminated in-memory filtering of large collections
  • Proper $text search instead of regex scanning

• ⚡ Context Value Memoization (SidecarProvider.tsx, LicenseProvider.tsx)

  • Wrapped context values in useMemo to prevent cascade re-renders
  • Children no longer re-render on unrelated state changes
  • Significant reduction in React reconciliation work
  • Applied to all major context providers

• ⚡ API Request Caching (src/lib/api/client.ts)

  • Implemented TTL-based request cache with 30-second default
  • In-flight request deduplication prevents duplicate API calls
  • Pattern-based cache invalidation on mutations
  • 50-70% reduction in redundant network requests

• ⚡ Analytics Query Pagination (src-python/app/routers/analytics.py)

  • Added proper limit/offset pagination with max 1000 records
  • Prevents OOM on large analytics datasets
  • Returns total count for client-side pagination UI
  • Date-based filtering pushed to database layer

• ⚡ Non-Blocking Async I/O (multiple routers)

  • Replaced synchronous file operations with asyncio.to_thread
  • Event loop no longer blocked by file reads/writes
  • Improved concurrent request handling
  • Applied to tools.py, logs.py, agent_hub.py

• ⚡ Dynamic Heavy Library Imports (use-mermaid.ts, use-markmap.ts)

  • Mermaid (~350KB) and Markmap (~200KB) now loaded on-demand
  • Initial page load 550KB+ lighter
  • Singleton pattern prevents duplicate initialization
  • Loading states shown during async import

Changed

• 📊 Optimization Progress: Phase 2 Performance now 100% complete (8/8 tasks)
• 📈 Overall Progress: 21 of 37 optimization tasks completed (57%)

Metrics

✨ Highlights

Spec Copilot v2: Your AI-Powered Specification Studio - A complete reimagining of the specification builder with a Notion-like canvas editor, inline AI copilot, and intelligent source materials management. Write specifications faster with AI that understands your context and enhances your content in real-time.

Added

• 🎨 Canvas Editor (src/components/spec/canvas/)

  • Block-based editing: Notion-style blocks for text, headings, lists, code, tables, and mermaid diagrams
  • Drag-and-drop: Reorder blocks effortlessly with visual drop indicators
  • Rich rendering: Live preview of markdown, syntax-highlighted code, and rendered mermaid diagrams
  • Section management: Expandable/collapsible sections with quick navigation

• 🤖 Inline AI Copilot (src/components/spec/canvas/CopilotPopover.tsx, CopilotLine.tsx)

  • Selection toolbar: Select any text and instantly enhance, expand, simplify, or add examples
  • Slash commands: Type / anywhere to access AI commands like generate, explain, or convert
  • Streaming responses: See AI-generated content appear in real-time with visual feedback
  • Context-aware: AI understands your entire spec for coherent enhancements

• 📎 Source Materials Panel (src/components/spec/sources/SourcePanel.tsx)

  • Unified management: Single panel for documents, images, text snippets, and URLs
  • Drag-and-drop upload: Drop files directly into the panel
  • Clipboard paste: Cmd+V to paste images or text instantly
  • Document parsing: Automatic text extraction from PDFs and DOCX via LlamaCloud/Unstructured
  • Smart preview: Graceful handling of unparsed documents with helpful messages
  • Local state management: No more full-page refreshes when adding/removing sources

• 💬 Enhanced Chat (src/components/spec/SpecChat.tsx)

  • Section context: Chat responses aware of currently selected section
  • Visual feedback: Animated indicators during AI processing
  • Quick actions: One-click buttons for common requests

• 📝 New Backend Endpoints (src-python/app/routers/specs.py)

  • POST /specs/{id}/sections/{id}/blocks/{id}/enhance - AI enhancement for individual blocks
  • POST /specs/{id}/sources/upload - Base64 document upload with parsing
  • POST /specs/{id}/sources/image - Image upload endpoint

• 📊 Code Insights Expansion (src/pages/CodeInsightsPage.tsx)

  • 10 new document types: API Gateway, Batch Processing, Cache Layer, Event Handler, Feature Flag, GraphQL, Message Queue, Rate Limiter, Search Index, State Machine
  • 2 new templates: Microservices Architecture, Event-Driven System

Fixed

• 🐛 PDF/DOCX Parsing: Fixed critical bug where document parser was called with wrong arguments
• 🐛 Canvas Rendering: Fixed mermaid diagrams, nested lists, and table rendering in canvas view
• 🐛 Memory Leaks: Resolved memory leaks in frontend event listeners and backend async operations
• 🐛 Source Panel Refresh: Panel now updates instantly without triggering full page reload
• 🐛 Spinner Stuck: Fixed upload spinner remaining after successful upload
• 🐛 Command Palette Icons: Synced icons between command palette and sidebar navigation

Changed

• ⚡ Improved AI Prompts: Enhanced prompts for better, more contextual AI responses
• 📋 Copy to Clipboard: Added copy functionality for AI-generated content

✨ Highlights

Repo Insights: Git-Fame Style Analytics - Introducing a powerful new utility that brings git-fame style contributor analytics to PeerAI Studio. Analyze any public or private repository to discover who contributes what, estimate development effort with COCOMO modeling, visualize collaboration networks, and identify file hotspots. Beautiful interactive charts bring your repository's history to life.

Added

• 📊 Repo Insights - Complete Contributor Analytics (src/pages/RepoInsightsPage.tsx, src-python/app/services/repo_insights_service.py)

  • Contributor Leaderboard: Rank contributors by commits, lines of code, and files owned with percentage breakdowns
  • COCOMO Effort Estimation: Calculate person-hours, person-months, and estimated cost based on lines of code
  • Activity Heatmap: GitHub-style calendar view showing commit activity over time
  • Commit Frequency Chart: Interactive area chart tracking commits, additions, and deletions over weeks
  • File Hotspots Treemap: Visual map of most-changed files with color-coded churn scores (green=stable, red=high churn)
  • Collaboration Network: Force-directed graph showing which contributors work on shared files
  • Language Distribution: Donut chart with progress bars showing language breakdown by lines of code
  • Code Churn Analysis: Identify potential technical debt through add/delete ratios
  • Export Options: Download results as JSON, CSV, or Markdown report

• 🔐 Private Repository Support

  • GitHub personal access token input for analyzing private repositories
  • Tokens are never stored - used only for the current session

• ⚡ Analysis Pipeline with Real-Time Progress

  • 8-step analysis pipeline with visual progress indicators
  • SSE streaming for live updates during analysis
  • Repository caching option for faster re-analysis
  • Sample mode for large repositories (analyze recent commits only)

• ℹ️ Rich About Dialog (src/components/AboutDialog.tsx)

  • New About dialog accessible from Help menu
  • Displays app version, system info, and links

Changed

• 💄 Enhanced Visualizations
  • Collaboration Network: Vibrant 15-color node palette with gradient edges and hover glow effects
  • Language Distribution: Horizontal progress bars with full language names visible
  • COCOMO Estimate: Colorful gradient cards (blue/emerald/amber) with smart number formatting (k/M suffixes)
  • File Hotspots: Treemap with color-coded churn indicators and top files table

Fixed

• 🐛 Results Tab Race Condition (src-python/app/routers/repo_insights.py)
  • Fixed timing issue where Results tab stayed disabled after analysis completed
  • Database now updates before SSE event is sent to client
  • Analysis status correctly shows "Completed" when results exist

Technical Details

New Files

Backend (Python)

• src-python/app/models/repo_insights.py - Pydantic models for all data types
• src-python/app/services/repo_insights_service.py - Git analysis, metrics calculation, export
• src-python/app/routers/repo_insights.py - REST API + SSE streaming endpoints

Frontend (React)

• src/pages/RepoInsightsPage.tsx - Main page with tabs
• src/hooks/use-repo-insights.ts - CRUD operations hook
• src/hooks/use-repo-insights-analysis.ts - Analysis with step-based progress
• src/lib/api/repo-insights.ts - API client
• src/lib/types/repo-insights.ts - TypeScript types

Visualization Components

• ContributorLeaderboard.tsx - Sortable table with sparklines
• COCOMOEstimate.tsx - Effort estimation cards
• ActivityHeatmap.tsx - GitHub-style calendar
• CommitFrequencyChart.tsx - Area chart over time
• FileHotspotsTreemap.tsx - Treemap visualization
• CollaborationNetwork.tsx - Force-directed graph
• LanguageDistribution.tsx - Donut chart with bars
• CodeChurnAnalysis.tsx - Tech debt indicators
• AnalysisSteps.tsx - Progress step display
• AnalysisHistory.tsx - Job list sidebar
• RepoConfigPanel.tsx - Configuration form
• ExportPanel.tsx - Export buttons

✨ Highlights

Blazing Fast & Feature-Rich Code Insights - This release supercharges startup performance and transforms Code Insights into a powerful, wizard-driven analysis tool. Database operations that blocked the app for seconds now complete in milliseconds. The new pipeline view gives you real-time visibility into analysis progress, while editable configuration forms put you in control.

Added

• 🔮 Code Insights Wizard Flow (src/pages/code-insights/)

  • New step-by-step wizard for creating analyses
  • Template selection for common analysis patterns
  • Document type configuration with enable/disable toggles
  • Token usage tracking with cost estimates

• 📊 Pipeline View & Progress Tracking (src/components/code-insights/)

  • Real-time pipeline visualization showing scan → parse → analyze → generate stages
  • Live progress indicators with file counts and percentages
  • Streaming activity log with expandable details
  • Visual feedback for each processing phase

• ✏️ Editable Configuration Forms (src/pages/code-insights/tabs/EditTab.tsx)

  • Full form-based editing for analysis configurations
  • Inline editing of connection settings, paths, and branches
  • Document type toggles with folder customization
  • Tag management with add/remove functionality

Changed

• ⚡ Parallel Index Creation (src-python/app/core/ask_database.py, future_of_work_repository.py)

  • Indexes now checked before creation - skip if already exist
  • Missing indexes created in parallel with asyncio.gather()
  • 33+ sequential awaits reduced to single parallel batch
  • Subsequent startups see "indexes already exist" - instant startup

• ⚡ Smart List Query Projections (src-python/app/routers/code_insights.py)

  • List queries now exclude large result fields (scanResults, parseResults, etc.)
  • Only metadata fields fetched for list view
  • Dramatically faster page loads for Code Insights dashboard

Fixed

• 🐛 Progress Card Visibility (src/pages/code-insights/)

  • Progress card now properly hides after analysis completes
  • No more lingering "in progress" states after completion

• 🐛 App Startup Blocking

  • Fixed systemic slowdown from 33+ sequential await create_index() calls
  • Ask PeerAI (24 indexes) and Future of Work (9 indexes) now initialize in parallel

Technical Details

✨ Highlights

Auto-Updates Go Live - The auto-updater is now fully functional. We've wired up the Tauri v2 updater plugin on the Rust side, added a "Check for Updates" menu item, and enabled the hourly background checks. When a new version drops, you'll see a sleek notification with one-click install. Plus, the new minimize-to-tray toggle in Settings means PeerAI Studio can stay ready in your system tray.

Added

• 🔄 Complete Auto-Update Pipeline (src-tauri/Cargo.toml, lib.rs)

  • Integrated tauri-plugin-updater on Rust backend
  • Configured update endpoints for GitHub releases
  • Added updater permissions to capabilities
  • Enabled signed update artifacts for secure delivery

• 📋 "Check for Updates" Menu Item (src-tauri/src/menu.rs)

  • New menu item in PeerAI Studio app menu
  • Triggers immediate update check on click
  • Shows notification banner if update available

• 🖥️ Desktop Settings Card (src/pages/settings/GeneralSettingsPage.tsx)

  • New "Desktop" section in Settings > General
  • Minimize-to-tray toggle with live Tauri backend sync
  • Clean UI with Laptop icon and descriptive text

• 📊 Enhanced Code Insights (src/components/code-insights/)

  • Comprehensive stats dashboard with file counts and metrics
  • Improved module tree visualization
  • Better streaming progress indicators

Fixed

• 🐛 Auto-Update Was Disabled (src/App.tsx)

  • Removed checkOnMount={false} and checkInterval={0} props
  • Updates now check on startup (5s delay) and hourly thereafter

• 🐛 Tauri Commands Missing Result Pattern (src-tauri/src/lib.rs)

  • set_minimize_to_tray now returns Result<(), String>
  • get_minimize_to_tray now returns Result<bool, String>
  • Follows Tauri v2 best practices for error handling

Technical Details

✨ Highlights

PeerAI Studio Goes Native - This release transforms PeerAI Studio from a web app in a window into a true desktop citizen. Native menus, system tray integration, global shortcuts, and desktop notifications make the app feel right at home on macOS. Plus, a powerful command palette puts every feature at your fingertips with Cmd+Shift+P.

Added

• 🍎 Native macOS Menu Bar (src-tauri/src/menu.rs)

  • Standard Edit menu with Cut, Copy, Paste, Select All
  • View menu with fullscreen toggle and developer tools
  • Window menu for minimize, zoom, and window management
  • Help menu linking to documentation and support

• 📍 System Tray Integration (src-tauri/src/tray.rs)

  • Persistent tray icon with PeerAI branding
  • Health indicator shows backend status at a glance
  • Quick actions: Show window, Quit application
  • Minimize-to-tray setting for background operation

• ⌨️ Global Keyboard Shortcuts (src/hooks/useGlobalShortcuts.ts)

  • Cmd+Shift+P opens command palette system-wide
  • Works even when app is in background
  • Built on Tauri's global-shortcut plugin

• 🎯 Command Palette (src/components/GlobalCommandPalette.tsx)

  • Spotlight-style quick navigation to all features
  • Fuzzy search across 30+ modules and actions
  • Compact system status bar shows service health
  • Keyboard-driven workflow: navigate, select, go

• 🔔 Native Notifications (src/lib/notifications.ts)

  • Desktop alerts for long-running task completion
  • Migration progress, agent execution, data generation
  • useNotifications hook for easy integration
  • Respects system notification settings

• 🔄 Auto-Update System (src/providers/UpdateProvider.tsx)

  • Checks for updates hourly in background
  • Non-intrusive notification when update available
  • One-click install and restart
  • Built on Tauri updater plugin

• 📊 System Status Dashboard (src/pages/settings/SystemPage.tsx)

  • Neo4j connection status with version info
  • SonarQube integration health check
  • MongoDB connection monitoring
  • Sidecar backend status and port display
  • Separate App Version and Sidecar Version display

• 🪟 Windows Build Pipeline (.github/workflows/build-windows.yml)

  • Manual GitHub Actions workflow for Windows builds
  • Cross-compilation from GitHub runners
  • Produces MSI and NSIS installers

• 🛠️ Developer Experience

  • ./scripts/dev.sh - Full-stack development launcher
  • ./scripts/dev-backend.sh - Backend-only with hot-reload
  • ./scripts/bump-version.sh --sidecar - Separate sidecar versioning

Changed

• Version Management
  • App and sidecar versions now tracked independently
  • Health endpoint returns sidecar version
  • Settings page shows both versions clearly

Technical Details

✨ Highlights

API Client Architecture 2.0 - The frontend API layer gets a complete makeover. The monolithic 2,698-line api-client.ts is now a lean facade backed by 18 domain-specific modules. Each module owns its endpoints, types, and logic. Import what you need, tree-shake the rest. This is clean architecture for the modern frontend.

Added

• 🎯 Domain-Driven API Modules (src/lib/api/)

  • projects.ts - Project management endpoints
  • agents.ts - Agent configuration and execution (moved from index.ts)
  • ask-peerai.ts - RAG chat, collections, sources, jobs
  • product-runs.ts - Product copilot with SSE streaming
  • migrations.ts - Data migration workflow
  • datamodels.ts - Schema modeling operations
  • deployments.ts - Deployment management
  • functional-tests.ts - Test case execution
  • test-data.ts - Synthetic data generation
  • services.ts - External service integrations
  • future-of-work.ts - Workforce analytics
  • code-insights.ts - Code analysis pipeline
  • analytics.ts - Usage analytics and metrics
  • agent-hub.ts - Agent marketplace operations
  • repo-visualizer.ts - Repository visualization
  • adrs.ts - Architecture Decision Records
  • playground.ts - API playground
  • feedback.ts - User feedback collection
  • token-usage.ts - LLM token tracking

• 🔗 Unified API Composition

  • New api object composes all domain modules
  • Clean import: import { api } from '@/lib/api'
  • Intuitive access: api.projects.list(), api.agents.execute()

Changed

• 📦 Backward Compatible Facade
  • api-client.ts reduced from 2,698 to 385 lines (86% reduction)
  • getApiClient() now delegates to domain modules
  • Zero breaking changes for existing code

Technical Details

Migration Notes

• No action required - existing getApiClient() calls work unchanged
• Recommended: Gradually migrate to import { api } from '@/lib/api'
• New pattern: api.domain.method() instead of getApiClient().domain.method()

✨ Highlights

Modular Architecture Revolution - A major structural refactoring that transforms monolithic files into clean, modular components. The massive 1,759-line ask_peerai.py is now a well-organized package. The sprawling api-client.ts becomes a composable module system. This release lays the foundation for faster feature development and easier code maintenance.

Added

• 🧱 Backend Core Utilities

  • config_loader.py - Generic JSON config loading with error handling
  • response_utils.py - Document normalization helpers (normalize_id, get_field)
  • exceptions.py - Typed HTTP exceptions (EntityNotFoundError, ValidationError, StorageError)
  • 34 new unit tests for core utilities with 100% coverage

• 📦 Modular Ask PeerAI Router

  • Split 1,759-line monolith into focused sub-routers:
    • collections.py - Collection CRUD operations
    • sources.py - Data source management
    • chat.py - Chat sessions, jobs, config, and debug endpoints
  • shared.py - Reusable utilities and response converters

• 🔌 Modular Frontend API Client

  • New src/lib/api/ module structure with domain-specific files
  • client.ts - Base ApiClient with retry logic and timeout handling
  • Domain modules: agents, prompts, tools, guardrails, ontologies, settings
  • Composable architecture for easy extension

• 🎨 Shared UI Components

  • FormFields.tsx - Reusable form components (InputField, TextareaField, SelectField, SwitchField, SliderField)
  • Database settings shared components and types

Changed

• 🐍 Backend Optimizations
  • main.py now uses shared config_loader (removed ~50 lines of duplication)
  • products_copilot.py uses shared utilities from app.core.filters
  • Extracted product_normalizer.py service with normalize_doc() and calculate_stats()

Technical Details

Migration Notes

• No breaking API changes - all endpoints maintain backward compatibility
• The modular structure is purely internal refactoring
• Existing imports continue to work

✨ Highlights

The Great Simplification - A comprehensive codebase refinement touching 59 files across frontend and backend. We've removed 1,069 lines of code while preserving every feature. Cleaner code means faster development, easier debugging, and better maintainability for the entire team.

Changed

• 🧹 Frontend Components Refined - Extracted reusable patterns across the UI

  • AgentConfigPanel: New LibrarySelectorTabs, library hooks for prompts/tools/guardrails
  • CodeInsightsWizard: Extracted StatusBadge, PipelineStepCard, PhaseDetails components
  • DatabasePage: New ActionButton, MaskedTextarea, MaskedInput helpers
  • ArtifactEditor: Extracted FormField, SelectField, CollapsibleArraySection
  • DeploymentSettingsForm: New ResourceCard, PresetButton, SectionHeader

• ⚡ Hooks Optimized - Consistent patterns across all React hooks

  • Consolidated refs in CRUD hook, removed redundant JSDoc
  • Synchronized filter functions (async → sync where appropriate)
  • Consistent error extraction using shared utilities

• 🐍 Python Backend Streamlined - Cleaner routers and services

  • projects.py: Extracted _get_project_doc(), _save_project_doc(), _build_app_modernization_activities()
  • adrs.py/migrations.py: Consolidated duplicate CRUD patterns
  • Models: Proper inheritance eliminates redundant model_config declarations
  • Services: Simplified schema_analyzer, api_key_resolver, adr_agent_service

• 📡 API Client Simplified - DRYer request handling

  • New postFormData helper consolidates 5 file upload methods
  • Extracted EvaluationResult type for INVEST/STRIPE evaluations
  • ~120 lines removed from api-client.ts alone

Improved

• List Components: All list views now share consistent renderContent() + hasFilters pattern
• Settings Controls: Removed unused refs, extracted validation status components
• Type Definitions: Removed unused types, consolidated similar definitions
• Utility Functions: Simplified conditionals, removed dead code paths

Technical Details

Quality Assurance

• ✅ All TypeScript strict checks pass
• ✅ All ESLint rules pass (0 warnings)
• ✅ All Ruff Python checks pass
• ✅ 99 pytest tests pass

✨ Highlights

Agent Hub UX Refresh - Tabs for asset types make browsing faster and more intuitive. See exactly what's available at a glance with counts for Agents, Templates, Hooks, and Connectors.

Added

• 🗂️ Asset Type Tabs - New primary navigation for Agent Hub

  • Tab Bar: All (31) | Agents (19) | Templates (3) | Hooks (4) | Connectors (5)
  • Live Counts: Badge shows total packages per type, updates in real-time
  • One-Click Filter: Switch between asset types instantly
  • Icons Per Type: Visual distinction - Bot, FileCode, Webhook, Plug icons

• 🔍 Smarter Search - Context-aware search placeholder

  • Placeholder updates based on active tab: "Search agents...", "Search templates...", etc.

• ⭐ Featured First Sort - New sort option to surface featured packages

Changed

• Filter UX Simplified - Asset type moved from dropdown to tabs

  • Primary navigation (what it is) via tabs
  • Secondary filters (category, trust, sort) remain as dropdowns
  • Cleaner visual hierarchy, less cognitive load

• Featured Section Repositioned - Now appears below main grid on "All" tab

  • Only shows when no filters applied
  • Includes Featured and Trending sections

Fixed

• Permission Type Validation - Added api_access and notification to enum
  • Connectors (Jira, GitHub, Slack, etc.) now validate correctly
  • Hooks (approval gates) with notification permissions work

✨ Highlights

Agent Hub is Here! - Your enterprise marketplace for AI automation. Discover, install, and manage pre-built agents, templates, hooks, and connectors. Build your AI automation supply chain with trust tiers, version control, and governance built-in.

Added

• 🏪 Agent Hub Marketplace - Enterprise AI agent discovery and management

  • Browse & Search: Filter by category, trust tier, asset type with grid/list views
  • 7-Tab Detail View: Overview, Permissions, Config, Dependencies, Usage, Security, Versions
  • One-Click Install: Install agents directly to your Agent Builder library
  • Star Ratings: Rate and review agents to help the community

• 🛡️ Trust & Governance - Enterprise-grade security controls

  • 4 Trust Tiers: PeerAI Certified, Partner Verified, Community, Local
  • Enhanced Permissions: Data access declarations, action scopes, risk levels
  • Approval Gates: Flag high-risk permissions requiring approval
  • Audit Logging: Track install, uninstall, execute, publish, deprecate actions

• 📦 Multiple Asset Types - Beyond just agents

  • Agents: Full AI automation packages (20+ samples included)
  • Templates: Reusable prompt patterns for user stories, API design, incident reports
  • Hooks: Pre/post execution guardrails - PII detection, rate limiting, approval gates, cost tracking
  • Connectors: MCP-standard integrations - Jira, GitHub, Slack, ServiceNow, Confluence

• 📚 Collections - Curated bundles for fast onboarding

  • Starter Packs: Essential agents for new users
  • Persona Bundles: Backend Developer, Frontend Developer, DevOps Engineer
  • Workflow Collections: Legacy Modernization, Security Essentials
  • One-Click Install: Deploy entire collection with a single action

• 📝 Publishing Wizard - Share your agents with the team

  • 5-Step Flow: Package Info → Config Upload → Permissions → Test Cases → Review
  • YAML Config Support: Upload your agent YAML with syntax validation
  • Permission Declaration: Declare what your agent needs access to
  • Test Case Bundling: Include test cases for quality assurance

• 🔄 Version Management - Full lifecycle control

  • Version History: Timeline view of all releases with changelogs
  • Promotion Workflow: Draft → Pending Review → Verified → Certified
  • Fork Capability: Clone and customize existing agents
  • Rollback Support: Revert to previous versions when needed
  • Kill Switch: Disable problematic agents globally

• 📊 Usage Analytics - Know how your agents perform

  • Execution Stats: Total runs, success rate, average duration
  • Dependency Tracking: Model, API, tool, and connector dependencies
  • Security Tab: SECURITY.md display, vulnerability tracking

• 🎮 Playground Integration - Test before you install

  • Try in Playground: Test any agent without installing
  • Config Passthrough: Agent config automatically loaded in playground

Technical Details

• 31 new files added (9,176 lines of code)
• Backend: Full CRUD API with install/uninstall, ratings, fork, promote, deprecate, kill-switch
• Frontend: 3 new pages, 8 new components, 2 new hooks, comprehensive type definitions
• Seed Data: 31 sample packages across all asset types and trust tiers

✨ Highlights

Agent Playground is Here! - A powerful new environment for testing and debugging AI agents built with Agent Builder. Stream outputs in real-time, inspect tool calls, track token usage, and maintain a full history of your test runs.

Added

• 🎮 Agent Playground - Complete agent testing environment

  • 3-Panel Layout: History sidebar, main execution area, toggleable metrics panel
  • Dual Agent Sources: Test both Agent Builder agents AND YAML config-based agents
  • Real-time Streaming: Watch agent output stream token-by-token with live cursor
  • Smart Variable Detection: Auto-detects {{variable}} patterns from YAML prompts and renders input fields
  • Markdown Rendering: Beautiful output display with syntax-highlighted code blocks
  • Keyboard Shortcuts: Cmd+B (sidebar), Cmd+M (metrics), Cmd+Enter (run/stop)

• 📊 Execution Metrics Panel - Deep insights into agent behavior

  • Token Tracking: Prompt tokens, completion tokens, total usage
  • Performance Stats: Elapsed time, iteration count, LLM call count
  • Tool Call Inspector: Expandable cards showing tool name, arguments, and results
  • Thinking Steps: View extended thinking process for supported models

• 📜 Run History - Persistent test run management

  • MongoDB Storage: Full run persistence with all inputs, outputs, and metrics
  • Star/Favorite: Mark important runs for quick access
  • Search & Filter: Find runs by agent name or content
  • Quick Replay: Click any run to restore its full state

• 📈 Token Usage Recording - Track LLM costs across all agent executions

  • Automatic token counting per execution
  • Provider and model breakdown
  • Cumulative usage tracking

• ✏️ CodeEditor Component - New syntax-highlighted code editor

  • Monaco-style editing experience
  • Language auto-detection
  • Copy-to-clipboard support

• 🌐 Remote Repository Support - CodeInsights now supports remote repos

  • Clone and analyze GitHub repositories
  • SSH and HTTPS authentication support

Fixed

• Agent Playground: Route order fix preventing "Agent 'configs' not found" error
• Agent Playground: Resilient agent loading with Promise.allSettled for partial failures
• CodeInsights: Multiple fixes for progress display, SSE parsing, and results rendering
• CodeInsights: MongoDB + file fallback pattern consistency

Changed

• Improved code block styling with consistent dark theme and language badges
• Updated peerai-tools to use local wheel instead of git dependency

✨ Highlights

Functional Test Copilot is Here! - Introducing a powerful new way to validate API compatibility during migrations. Capture real traffic, auto-generate BDD test scenarios, and compare responses between old and new APIs—all with AI-powered intelligence to catch issues before they reach production.

Added

• 🧪 Functional Test Copilot - Complete API regression testing solution

  • 4-Step Wizard Flow: Capture → Review → Generate → Execute
  • HAR File Upload: Drag-and-drop support for HTTP Archive files up to 50MB
  • Capture Script Templates: Ready-to-use snippets for Chrome DevTools, Playwright, mitmproxy, and tcpdump
  • Smart API Detection: Automatically filters static assets, health checks, and metrics endpoints

• 📋 API Call Review - Comprehensive traffic analysis

  • Filterable List View: Filter by HTTP method (GET, POST, PUT, DELETE) and status codes
  • Bulk Selection: Select all, deselect all, or cherry-pick specific calls
  • Detailed Inspector: View headers, query params, request/response bodies with syntax highlighting
  • Exclude Patterns: Mark endpoints to skip during test generation

• ⚡ AI-Powered Test Generation - Automatic BDD scenario creation

  • Gherkin Scenarios: Human-readable Given/When/Then specifications
  • pytest-bdd Code: Production-ready Python test code with fixtures
  • Smart Assertions: Auto-inferred status codes, content types, and JSON structure checks
  • SSE Streaming: Real-time progress updates during generation

• 🔄 Dual API Execution - Side-by-side comparison engine

  • Configurable Endpoints: Set base URLs for old and new APIs
  • Auth Support: Bearer tokens, API keys, and basic auth
  • JSON Diff Viewer: Visual comparison with added/removed/modified highlighting
  • Response Time Tracking: Performance regression detection

• 💡 AI Suggestions Panel - Intelligent test improvement recommendations

  • Exclude Endpoint: Suggests removing health checks, auth, and static assets
  • Ignore Field: Detects volatile fields (timestamps, IDs, tokens) causing false failures
  • Strengthen Assertion: Recommends adding content-type and response time checks
  • Fix Flaky Test: Identifies tests failing only due to dynamic data
  • One-Click Actions: Accept or dismiss suggestions instantly

• 🎯 Execution Results Dashboard - Comprehensive test reporting

  • Pass/Fail Summary: At-a-glance test status with pass rate percentage
  • Diff Explorer: Drill into individual test results with tabbed views
  • Assertion Results: Detailed breakdown of each assertion outcome
  • Version History: Track execution runs over time

• 🌱 Seed Data Script - Development-ready sample data

  • 4 sample test runs with varied states (passed, failed, ready, capturing)
  • Realistic HAR entries, scenarios, and execution results
  • MongoDB Local and Atlas support with --atlas flag

✨ Highlights

Test Data Generation is Here! - Generate realistic, production-quality test data with our new AI-powered Test Data Generator. Whether you need to populate a database for testing or export sample data files, this feature has you covered with a beautiful 6-step wizard workflow.

Added

• 🧪 Test Data Generator - Complete synthetic data generation solution

  • 6-Step Wizard Flow: Job Details → Destination → Entities → Volume → Progress → Results
  • Multi-Database Support: MongoDB, PostgreSQL, MySQL with connection testing
  • File Export Options: Generate CSV, JSON, or SQL files for local use
  • Entity Management: Select tables/collections with column-level configuration
  • Volume Control: Set target row counts per entity with visual estimation
  • Real-Time Progress: Live progress tracking with entity-level status updates
  • Activity Logging: Detailed execution logs in a slide-out drawer

• ☁️ Cloud API Integration - Enterprise-grade data generation

  • Cloud job creation and workload management
  • Parallel workload execution with progress polling
  • Automatic Faker method inference based on column names and types
  • Smart defaults for common patterns (email, name, phone, address, etc.)

• 📊 Results & Export - Comprehensive output options

  • Sample data preview with paginated table view
  • One-click export to CSV, JSON, or SQL formats
  • Copy SQL INSERT statements to clipboard
  • Entity-specific or bulk export options

• 🎯 Smart Faker Mapping - Intelligent data generation rules

  • Name-based inference: email → faker.internet.email, phone → faker.phone.number
  • Type-based fallbacks: UUID, integers, decimals, booleans, dates
  • Custom generation rules per column

Fixed

• 🐛 API Call Performance - Fixed rapid repeated API calls during progress simulation
  • Moved side effects outside React state updater callbacks
  • Added refs for proper interval and completion state tracking
  • Prevents multiple API calls in React 18+ concurrent rendering mode

✨ Highlights

Projects Supercharged - This release transforms the Projects feature into a comprehensive project management powerhouse. Define your input sources, configure your target architecture, and manage everything through a beautiful ADR-style interface.

Added

• 📂 Input Sources Management - Connect all your project artifacts

  • Git Repositories: Link source code with branch and repository type (UI, Backend, Database, Fullstack, Monorepo)
  • Log Sources: Configure log files, URLs, or S3 locations for analysis
  • Documentation: Connect Confluence, SharePoint, or file-based docs
  • Jira Integration: Link tickets with project key and custom JQL filters
  • Feedback Sources: Import user feedback from CSV, JSON, or API endpoints
  • Full CRUD operations with status tracking (pending, connected, syncing, ready, failed)

• 🏗️ Target Architecture Configuration - Design your technology stack

  • 5 architectural layers: UI, Backend, Database, Data Pipeline, Deployment
  • Collapsible sections with enable/disable toggles per layer
  • Primary framework selection (React, Vue, Angular, FastAPI, Spring, etc.)
  • Additional technologies multi-select (TypeScript, GraphQL, Redis, Kafka, etc.)
  • Available for App Modernization and Greenfield Development templates
  • "Enable Target Architecture" button for existing projects

• 📋 ADR-Style Project Editor - Professional project management interface

  • Tabbed navigation: Edit, Sources, Activity, History
  • Edit tab with sidebar: Basic Info, Target Tech, Settings
  • Sources tab for full source management with add/edit/delete
  • Activity tab showing project progress (existing ActivitiesPanel)
  • History tab with timeline of all project changes
  • Quick status dropdown in header for workflow management

• 🧙 Enhanced Project Wizard - Smarter project creation

  • Conditional steps based on template selection
  • Sources step for configuring input artifacts
  • Architecture step only shown for modernization/greenfield templates
  • Enhanced review step with sources and architecture summary

Changed

• Project detail view now uses ADR-style tabbed interface instead of simple form
• Streamlined sidebar navigation (removed duplicate Sources entry)

✨ Highlights

Project Activities & Security Integration - This release introduces a powerful human-in-the-loop workflow system for projects and adds comprehensive security scanning integration. Projects now support guided activities that orchestrate AI copilots, while the new Security Settings page connects your codebase to SonarQube/SonarCloud for continuous code quality monitoring.

Added

• 🎯 Human-in-the-Loop Activities for Projects - Revolutionary workflow orchestration

  • Activities within project phases guide users through complex multi-step processes
  • Each activity links directly to AI copilots (Data Model, Migration, Product, Deployment)
  • Automatic status tracking: activities update to "Completed" when copilot runs finish
  • Activity history drawer shows full audit trail of all runs
  • "Return to Project" navigation after completing any copilot
  • Auto-sync detects existing copilot runs and links them to activities
  • Project templates now include pre-configured activities for common workflows

• 🔐 Security Settings Page - SonarQube & SonarCloud Integration

  • New dedicated Security page in Settings for code quality scanning
  • Support for both SonarQube (self-hosted) and SonarCloud (cloud-hosted)
  • Secure token storage in macOS Keychain
  • Project key configuration for multi-project support
  • Connection testing to validate credentials
  • Foundation for future security scanning features

• 🔗 Activity-Copilot Deep Integration - Seamless context passing

  • URL parameters carry projectId, activityId, and returnTo between pages
  • All copilots (Product, Migration, Deployment) now support activity linking
  • Backend models extended with activity_id field for persistence
  • Activity status API for real-time updates from copilots

Fixed

• 🐛 Migration Executor Crash - Fixed crash when source_schema is None
• 🐛 Sidecar Shutdown - App now properly terminates sidecar process on close
• 🐛 Activity Status Updates - Status now updates immediately on copilot completion

Fixed

• 🐛 PyInstaller Sidecar Code Signing - Fixed sidecar failing to load in production builds
  • Added disable-library-validation entitlement for hardened runtime compatibility
  • Allows PyInstaller's extracted Python.framework to load with different signatures
  • Resolves "code signature not valid for use in process" error on macOS

Fixed

• 🐛 Release Build Loading Issue - Fixed app getting stuck on loading spinner
  • Terms acceptance now uses localStorage instead of API polling
  • App no longer hangs indefinitely when sidecar is slow or unavailable
  • Instant terms check on startup for better user experience
  • Backend sync happens in background (non-blocking)

✨ Highlights

Code Quality Release - Major codebase simplification removing 725+ lines of duplicate code through shared utilities, factories, and components. This release focuses on maintainability and developer experience.

Added

• 🧩 Shared CardSkeleton Component - Unified loading states across list pages

  • New CardSkeleton and CardSkeletonGrid components in src/components/shared/
  • Consistent loading experience for Prompts, Ontologies, Tools, and Guardrails pages
  • Configurable skeleton count for different page layouts

• 🔧 build_update_dict Utility - Simplify Python router update methods

  • New utility in app/core/utils.py for building update dictionaries
  • Automatic enum value extraction
  • Support for nested Pydantic model lists with by_alias option
  • Reduces boilerplate in CRUD routers by ~70%

• 🏭 _field() Factory - Streamlined schema field definitions

  • New factory function in migration_analyzer.py
  • Reduces field definitions from ~50 lines to single-line calls
  • Keyword arguments for cleaner, more readable code
  • Smart defaults (nullable=True, is_pk=False, etc.)

Changed

• ♻️ ProductAgentService Streaming - Consolidated artifact generation

  • New _generate_artifacts_streaming() helper method
  • Unified streaming pattern for PRD, Epics, Features, Stories, Test Cases, NFRs, and Journey Maps
  • Reduced ~400 lines of duplicate streaming logic
  • Consistent error handling and progress events

• 📝 Router Simplification - Cleaner update endpoints

  • Prompts, Ontologies, Tools, and Guardrails routers now use build_update_dict
  • Removed repetitive if field is not None checks
  • More maintainable and less error-prone code

Developer Experience

• All linting passes (ESLint, TypeScript, Ruff)
• No breaking changes to APIs or components
• Fully backward compatible

Fixed

• Add missing terser dependency for production builds

✨ Highlights

This release brings Ask PeerAI directly to your fingertips with a floating widget, plus enterprise-grade production features for observability and user engagement.

Added

• 🤖 Ask PeerAI Widget - Instant AI assistance from anywhere in the app

  • Floating widget button in bottom-right corner (click to expand)
  • Tabbed interface: Chat with AI + Submit Feedback in one place
  • Streaming responses with markdown rendering and code highlighting
  • "New Chat" button to start fresh conversations
  • Contextual help powered by your Documentation Knowledge Base
  • Beautiful dark/light theme support

• 📚 Documentation Knowledge Base Settings - New settings page for RAG configuration

  • View vector search index status and configuration
  • Health check endpoint for vector database connectivity
  • Debug endpoint for troubleshooting search issues
  • Real-time index statistics display

• 💬 Feedback System - Capture user feedback with ease

  • Feedback widget with emoji reactions (😊 😐 😞)
  • Optional screenshot attachment support
  • Category selection (bug, feature, question, other)
  • Full feedback API with MongoDB storage

• 📊 Token Usage Tracking - Monitor your LLM API consumption

  • Track tokens per request (prompt + completion)
  • Cost calculation with up-to-date pricing
  • Usage history with filtering and aggregation
  • Support for all major providers (OpenAI, Anthropic, Google)

• 📈 Analytics Dashboard - Admin insights at a glance

  • New /admin/analytics dashboard page
  • Separate analytics MongoDB database support
  • Track feature usage, user engagement, and trends
  • Beautiful charts and metrics visualization

• 🔍 Sentry Integration - Production-grade error tracking

  • Backend error capture with full stack traces
  • Optional frontend SDK integration
  • Environment and release tagging
  • Settings page configuration in Observability section

• 🔐 macOS Code Signing & Notarization - Enterprise-ready distribution

  • Proper code signing with Developer ID certificate
  • Apple notarization for Gatekeeper approval
  • New entitlements.plist for required permissions
  • SHA256 checksums for release artifacts

Changed

• Widget panel widened to 700px for better readability
• Improved RAG search with higher topK values
• Better question generation for follow-up suggestions
• Enhanced code block styling in chat responses
• Release workflow now includes notarization step
• Production builds use terser minification

Fixed

• Vector health endpoint ImportError
• Code block overflow in chat messages
• Chat scroll area respects full viewport height

Added

• Multi-Version Release Notes in Update Dialog - When updating across multiple versions (e.g., v0.11.1 → v0.13.1), the update dialog now shows release notes for all intermediate versions
  • Collapsible accordion-style UI for each version
  • Version badges with relative timestamps ("2 days ago", "1 week ago")
  • Quick links to each version's GitHub release page
  • Latest version expanded by default, others collapsed
  • Displays count of versions being updated ("3 versions to update")

Changed

• Update checker now fetches and filters all studio releases from GitHub API
• UpdateCheckResult interface extended with intermediateReleases array
• New ReleaseInfo interface for structured release data

Added

• Comprehensive Codebase Documentation - 50 new markdown files in docs/ folder
  • Full architecture documentation with MermaidJS diagrams
  • Three-layer system coverage: React frontend, FastAPI backend, Tauri desktop
  • Agent runtime deep-dive: configuration, types, tools, execution, streaming
  • Component library docs: hooks, providers, UI components catalog
  • Security documentation: keychain integration, API key management
  • Licensing system: key format, CRC16 validation, feature gating
  • Settings architecture: auto-save patterns, module configuration
  • CI/CD workflows: GitHub Actions, release automation
  • All 5 copilots documented: ADR, Data Model, Migration, Product, Deployment
  • Complete API reference with curl examples and SSE streaming patterns
  • Testing guides: Vitest (frontend) and Pytest (backend) patterns
  • Contributing guidelines: code style, PR process, gitmoji conventions
  • Deployment options: Homebrew cask, manual DMG installation

Documentation Structure

docs/
├── README.md                 # Navigation home
├── getting-started.md        # Quick start guide
├── architecture/             # System architecture (4 files)
├── agent-runtime/            # LLM agent system (6 files)
├── components/               # React components (4 files)
├── security/                 # Security model (3 files)
├── licensing/                # License system (3 files)
├── settings/                 # Settings system (3 files)
├── updates/                  # Auto-update (1 file)
├── ci-cd/                    # CI/CD pipelines (3 files)
├── copilots/                 # AI copilots (6 files)
├── api/                      # API reference (3 files)
├── setup/                    # Dev setup (3 files)
├── testing/                  # Test guides (3 files)
├── contributing/             # Contribution (3 files)
└── deployment/               # Distribution (3 files)

Fixed

• Ask PeerAI Data Sources Panel - Fixed width constraint not being applied (Sheet's default max-width was overriding custom width)
• Ask PeerAI Source Card - Three-dot menu button now always visible instead of hover-only
• Ask PeerAI Source Deletion - Fixed bug where deleted sources remained in database due to conditional deletion logic

Added

• Ask PeerAI - RAG-powered document Q&A module
  • Document ingestion: PDF upload, web URL crawling, text paste
  • Vector search with ChromaDB and configurable chunk sizes
  • Multiple embedding providers: OpenAI, Anthropic, Google
  • Chat interface with streaming responses and inline citations
  • Related questions and source preview
  • Collections for organizing documents
  • Background job queue for document processing
  • Exa integration for real-time web search augmentation
  • Dedicated database configuration for vector storage
  • Command palette (⌘K) for quick actions
  • Session management with chat history

Added

• Shared Base Entity Types - New BaseEntity type system for consistent entity definitions

  • BaseVisibility and BaseStatus types shared across all entities
  • BaseListParams and BaseListResponse for standardized API responses
  • All entity types (Prompt, Tool, Ontology, Guardrail, Project, ADR) now extend BaseEntity

• Generic CRUD Service (Python) - New CrudService[T] class with PEP 695 type parameters

  • Unified MongoDB/file storage abstraction
  • Consistent filtering, search, and pagination across all entities
  • Shared filter utilities in app/core/filters.py

• Generic CRUD Hook (React) - New useCrudResource hook

  • Eliminates duplication across entity-specific hooks
  • Standardized loading states, error handling, and toast notifications
  • Uses useRef pattern to prevent infinite loops with object dependencies

• Shared EntityCard Component - Generic card with three display variants

  • Supports compact, grid, and list layouts
  • Configurable via entity-ui-config.ts for visibility/status styling
  • Extensible with renderExtra prop for entity-specific content

Changed

• Refactored CRUD Hooks - Reduced from ~1,145 lines to ~200 lines

  • use-prompts.ts, use-tools.ts, use-ontologies.ts, use-guardrails.ts now thin wrappers

• Refactored Backend Routers - Reduced from ~2,457 lines to ~1,000 lines

  • prompts.py, tools.py, ontologies.py, guardrails.py, projects.py use CrudService

• Improved DocumentPreview Dialog - Copy/Close buttons moved to proper dialog footer

• Improved Source Documents UI - Paste Text button now adjacent to Upload Document

Fixed

• Python lint errors: Import Callable from collections.abc, use PEP 695 type parameters
• Archived status color now uses text-gray-500 for consistent color manipulation

Added

• Collapsible Navigation Categories - Organize sidebar with expandable sections

  • AI Development, Testing & Data, Architecture, Configuration categories
  • Persistent collapse state across sessions
  • Smooth expand/collapse animations with Radix Collapsible
  • Intuitive chevron indicators for visual feedback

• Reusable List Page Components - New shared component library

  • ListPageHeader - Consistent page headers with icon, title, description, actions
  • ListPageStats - Horizontal stats bar with colored values and dividers
  • formatRelativeDate utility for consistent date formatting

Changed

• Unified List View Experience - All module list pages now share consistent UX

  • Full-width container layout replacing master-detail sidebar pattern
  • Grid/List view toggle with smooth transitions
  • Horizontal stats bar with real-time counts
  • Consistent search, filter, and action button placement
  • Updated pages: Prompts, Tools, Guardrails, Ontologies, Projects, ADR, Data Model, Migration, Product

• Enhanced Card Components - All card components support multiple display modes

  • variant prop with "compact", "grid", "list" options
  • Responsive layouts optimized for each view mode
  • Consistent hover states and click interactions

• Improved Navigation Sidebar

  • Hidden scrollbar for cleaner appearance
  • Better visual hierarchy with category groupings
  • Reorganized modules into logical categories

Fixed

• ESLint import order warnings across multiple files
• Consistent routing pattern for all modules (/module, /module/new, /module/:id)

Added

• Data Migration Copilot - AI-powered any-to-any database migration wizard

  • Support for 9 source databases: Oracle, SQL Server, PostgreSQL, MySQL, MongoDB, Cosmos DB, DynamoDB, Couchbase, HBase
  • Support for 3 target databases: MongoDB, PostgreSQL, MySQL
  • 8-step guided wizard for end-to-end migration planning and execution

• Step 1: Connect - Database connection configuration

  • Dynamic connection forms based on database type
  • Support for connection strings or manual configuration (host, port, credentials)
  • Test connection functionality with latency display
  • SSL/TLS configuration options

• Step 2: Assessment - Source schema analysis

  • Automatic schema discovery with table/collection detection
  • Row counts, storage sizes, and column information
  • Index and constraint detection (PK, UNIQUE, FK, NULL)
  • Re-analyze capability for schema refresh

• Step 3: Constraints - Migration behavior configuration

  • NULL value handling options
  • BLOB/CLOB large object handling
  • Date/time format and character encoding settings
  • Batch size and parallel workers configuration
  • ID handling (preserve original vs generate new)

• Step 4: Schema Mapping - Entity and field mapping editor

  • Visual source → target mapping interface
  • Multiple mapping strategies: Root Document, Embedded, Embedded Array, Referenced, Denormalized, Hybrid
  • Field-level mapping with type transformation
  • Compatibility scoring for type conversions

• Step 5: Volume Estimation - Migration size and duration planning

  • Total records and storage size calculations
  • Source vs target size comparison
  • Data distribution visualization by table
  • Resource recommendations (CPU, memory, storage)
  • Estimated migration duration

• Step 6: Target Configuration - Cloud deployment settings

  • Provider selection (Atlas, RDS, self-hosted)
  • Cluster tier selection with specs and pricing
  • High availability options (replica sets, sharding)
  • Security features (encryption, backups, monitoring)
  • Monthly cost estimation

• Step 7: Planning - Migration plan generation

  • Phase-based timeline with task breakdown
  • Risk assessment with severity ratings and mitigations
  • Pre-migration checklist with actionable items
  • Export plan functionality

• Step 8: Workbench - Real-time migration execution

  • Live progress tracking with overall and per-table progress bars
  • Real-time stats: rows migrated, throughput (rows/sec), time remaining
  • Execution logs with auto-scroll, copy, and export functionality
  • Pause/Resume/Cancel controls
  • Post-migration validation results

• Migration List View - Dashboard for all migrations

  • Grid and list view modes
  • Search and filter by status/project
  • Stats overview (total, completed, failed, in-progress)
  • Quick access to migration details

Changed

• Updated modules configuration with Data Migration Copilot entry
• Enhanced API client with migration endpoints
• Added migration-specific type definitions and profiles

Fixed

• SSE endpoints now use GET method for EventSource compatibility
• SSE streams properly resolve promises on completion for correct async/await behavior
• Real-time log display formatting with proper inline layout

Added

• Product Management Copilot - AI-powered product artifact generation and management

  • Generate complete product hierarchies: PRDs → Epics → Features → User Stories → Test Cases
  • AI-driven artifact creation with streaming progress and real-time feedback
  • Hierarchical tree navigation for exploring artifact relationships
  • Support for NFRs (Non-Functional Requirements) and User Journey Maps

• INVEST/STRIPE Evaluation Framework - Quality assessment for product artifacts

  • INVEST criteria evaluation for User Stories and Epics (Independent, Negotiable, Valuable, Estimable, Small, Testable)
  • STRIPE criteria evaluation for Test Cases (Specific, Traceable, Repeatable, Independent, Precise, Executable)
  • Detailed scoring with feedback and improvement suggestions
  • Persistent evaluation results stored with artifacts

• Document Upload & Parsing - Context-aware generation from existing documents

  • Support for PDF, DOCX, and TXT file uploads
  • Integration with LlamaCloud and Unstructured for intelligent document parsing
  • Document preview with parsed content visualization
  • Use uploaded documents as context for artifact generation

• Comprehensive Artifact Editor - Full-featured editing for all artifact types

  • Epic fields: businessContext, successCriteria, targetUsers, businessValue
  • Feature fields: userRequirements, technicalRequirements, acceptanceCriteria, dependencies
  • User Story fields: As a/I want/So that format, acceptanceCriteria, storyPoints, notes
  • Test Case fields: type selector (functional/integration/e2e/performance/security), step editor
  • Collapsible sections for complex array fields

• Version History & Diff - Track changes across artifact versions

  • Visual diff comparison between versions
  • Restore previous versions with one click
  • Version metadata with timestamps

• Export Capabilities - Share and document your product artifacts

  • Export to Markdown format
  • Export to PDF with professional styling
  • Copy to clipboard functionality

• AI Suggestions Panel - Context-aware improvement recommendations

  • Automated suggestions for artifact improvements
  • Accept/reject suggestions with one click
  • Confidence scores for each suggestion

• Product Templates Settings - Customize artifact section templates

  • Epic sections (14 templates): Title, Business Context, Goals, Key Features, Acceptance Criteria, etc.
  • Feature sections (17 templates): Including new Business Context, Success Metrics, WSJF Scoring
  • Story sections (13 templates): Including Linked Work Items, Stakeholders
  • Test Case sections (11 templates): Including AC Alignment, Notes

• New Reusable Components

  • StringArrayInput - Numbered list editor for string arrays with add/remove
  • TestStepsEditor - Collapsible step editor for test cases with action/expected result

Changed

• Enhanced seed data with realistic product management examples (E-commerce, Mobile Banking, Healthcare, SaaS Analytics)
• Improved activity drawer pattern for streaming operation visibility
• Updated API client with product management endpoints

Added

• API Explorer - Built-in HTTP client for testing and debugging backend endpoints
  • Searchable endpoint catalog organized by category (System, Settings, Prompts, Tools, etc.)
  • Full-featured request builder with method selector, URL input, params/headers/body editors
  • Response viewer with status badge, timing metrics, headers, and formatted body
  • Keyboard shortcut ⌘ + Enter to send requests quickly
  • Support for GET, POST, PUT, DELETE, PATCH methods
  • Backend proxy router (http_client.py) for request handling

Changed

• Updated documentation with LLM agent infrastructure details
• Added Activity Drawer pattern documentation for reuse in other features
• Expanded environment variables documentation with API key references

Added

• LLM Agent Infrastructure - Complete agent runtime for AI-powered workflows

  • YAML-based agent configuration with variable substitution ({{variable}} syntax)
  • Three agent types: Simple (single prompt), ReAct (tool-using), and Deep (orchestrator)
  • Streaming execution with Server-Sent Events (SSE)
  • Built-in tools: Exa neural search, web fetch, structured output
  • Agent executor UI for testing and debugging agent configurations

• ADR Copilot - Multi-agent AI assistant for Architecture Decision Records

  • Three specialized subagents working in sequence:
    • context_researcher - Researches decision topics using Exa search
    • options_analyzer - Analyzes alternatives and generates trade-off matrices
    • adr_writer - Synthesizes research into professional ADR sections
  • Real-time streaming of generation progress with section-by-section updates
  • Context-aware prompts using actual ADR data (technologies, constraints, decision)

• Agent Activity Drawer - Real-time visibility into agent execution

  • View prompts sent to agents with full variable substitution
  • See agent responses and intermediate outputs
  • Debug events for troubleshooting generation issues
  • Copy button for easy extraction of prompts/responses
  • Collapsible event details with syntax highlighting

• OpenRouter Integration - Access multiple AI models through single provider

  • Support for Claude, GPT, Gemini, and other models via OpenRouter
  • Configurable in Settings > AI Models
  • Default provider for ADR Copilot agents

• Tool API Keys Settings - Manage API keys for agent tools

  • New settings page at Settings > Tool API Keys
  • Exa API key configuration for neural search capabilities
  • Secure storage in macOS Keychain

Changed

• ADR generation now uses LLM agents by default (was stub mode)
• Simplified GenerationProgress component with cleaner UI
• Expanded Agent Activity drawer width for better readability
• Improved log viewer with functional Clear button

Fixed

• decision_technology field now properly saved to database on CREATE and UPDATE
• Agent prompts now substitute variables correctly (technologies, constraints, etc.)
• OpenRouter model IDs corrected (use dot notation: claude-sonnet-4.5)
• Agent activity streaming properly tracks section completion
• Circular import resolved in agents module
• Python lint errors in test files (unused imports, variables)

Added

• Data Model Copilot - AI-powered schema transformation and migration planning
  • 5-step wizard: Inputs → Analysis → Mapping → Generate → Review
  • Support for multiple source databases: PostgreSQL, MySQL, SQL Server, Oracle, MongoDB, CosmosDB, DynamoDB
  • Target database support: MongoDB, PostgreSQL, MySQL
  • Visual database selector with category badges and icons
  • Schema analysis with SSE streaming progress
  • Entity mapping with configurable strategies (root, embedded, embedded-array, referenced, denormalized, hybrid)
  • Field-level mapping editor with type compatibility indicators
  • AI Suggestions panel with accept/dismiss workflow
  • Schema generation with real-time progress streaming
  • Generated schema preview with Compass-style tree view
  • JSON schema viewer with syntax highlighting and line numbers
  • Entity sidebar navigation for browsing generated collections
  • Creative summary view in Review step with expandable sections
  • Migration journey visualization showing source → target transformation
  • Strategy distribution charts and field compatibility analysis
  • Export options: DDL statements and JSON schema
  • Version history with snapshot/restore capability
  • Sample data seeding script for testing (scripts/seed-datamodels.py)

Changed

• Enhanced wizard stepper component with equidistant step layout
• Improved tab content layouts using absolute positioning for proper scrolling

Fixed

• RunDetail layout issues - Details tab now scrolls properly
• Generated Schema tab whitespace issue resolved
• Python import sorting in main.py (ruff I001)

Added

• generate-changelog.sh script for auto-categorizing commits by gitmoji
• CHANGELOG.md with Keep a Changelog format for detailed release notes
• .github/release.yml for PR categorization in auto-generated notes

Changed

• Release workflow now extracts notes from CHANGELOG.md (falls back to commits)
• bump-version.sh now updates CHANGELOG.md and reminds about release notes
• Normalized clipboard operations to use shared clipboard-utils.ts utility
• Updated documentation for PDF export prerequisites and Tauri plugins

Added

• PDF export for ADRs using WeasyPrint with professional styling
• Tauri clipboard plugin for proper native clipboard access
• Section navigation in ADR preview

Fixed

• ADR export - clipboard and PDF generation with WeasyPrint
• ADR export using Tauri plugins for HTTP/dialog/fs operations
• ADR preview scroll using absolute positioning for bounded height

Changed

• Improved ADR preview layout with better section organization
• Release workflow now fetches notes from main repo for homebrew-tap

Added

• ADR Copilot - Architecture Decision Records feature
• Atomic technology comparison for ADRs
• ADR constraints section and section tabs
• ADR versioning with Save Version dialog
• Status badge with quick change dropdown in ADR header

Changed

• Refactored ADR page to full-page list view (like Agents)
• Tags now display on same line in ADR preview
• Show current version in Save Version dialog

Changed

• Made update dialog wider with taller release notes area
• Auto-generate release notes from git commits in release workflow

Added

• Check for Updates feature in Settings > System

Fixed

• Various bug fixes and stability improvements

Fixed

• Minor bug fixes

Added

• Agent Builder with visual drag-and-drop editor
• Agent YAML import/export
• MCP (Model Context Protocol) publishing for agents

Fixed

• Bug fixes for Projects feature

Added

• Projects feature with phases and progress tracking
• Project creation wizard (3-step)
• Project templates (App Modernization, Greenfield Development, etc.)

Added

• Guardrails pattern testing endpoint

Fixed

• Various stability improvements