Repository navigation
Releases
v1.4.0 — People workspace, Ollama, native Windows
v1.4.0 — People workspace, Ollama, native Windows
Bodhiorchard now works for the whole company, not just engineering. HR and Sales staff get a People workspace with leave, work from home, the team calendar and their own records, and never see the engineering side. This release also adds Ollama as a fourth AI provider, native Windows support, and a round of security fixes.
Highlights
- Leave and work from home. Balances per leave type, half days, optional holidays, and a request dialog that counts working days before anything is sent.
- Loss of pay. Leave past someone's balance goes through as unpaid days. The employee sees it before sending, the approver sees it before deciding, and the days are given back if the request is cancelled or a holiday is added.
- Approvals from Slack. Several reporting managers per person; any one of them decides and the first decision wins, with HR covering when no manager is left. Approvers get Approve and Reject buttons in a Slack DM.
- Leave policies per location. Leave types, quotas, working days, timezone, leave year, proration for new joiners and work-from-home rules per policy. Public holidays are imported by country and region, then edited.
- Offices and remote work. Each office follows a leave policy; remote people have no office and never show as working from home.
- Team calendar and daily Slack post. Names only, never the leave type or reason, and nobody on leave is @mentioned.
- Encrypted employee records. Bank and ID numbers and documents are encrypted with AES-256-GCM under a separate HR key. Reveals and downloads are logged, and each employee can see who opened their records.
- Workspace roles. New HR and Sales roles. People-only users are refused engineering routes at the API, MCP, WebSocket and multiplayer layers, and nobody can grant a role with more access than they hold.
- Ollama provider. Run the agents on a local or hosted Ollama endpoint, with no CLI and no data leaving your network.
- Native Windows. Hybrid mode runs on Windows without WSL, and scans work without Redis.
- Smaller additions. A two-player Backlash mini-game, file attachments on bug reports, restoring discarded BUDs from the board, uploading wireframe HTML as designs, BUD todos over MCP, and the latest Claude models as defaults.
Security
- Full Docker installs now generate their own
SECRET_KEY,ENCRYPTION_KEYand multiplayer bridge secret on first start and keep them on a volume. Stored tokens are re-encrypted off the old shared default key automatically, which signs everyone out once. - Deactivated users lose access straight away, and refresh tokens no longer work as access tokens.
- Request bodies are capped at 26 MB.
- Setup endpoints are closed once setup is done, and the directory picker is confined to the home directory.
Upgrading
- Migrations run on start as usual.
- Employee records have their own encryption key,
HR_ENCRYPTION_KEYS. Full Docker generates it on first start and keeps it on thesecrets_datavolume. The npx installer (now 1.1.0) adds it to.envon your nextbodhiorchard startorupdate, andnpm run setupadds it tobackend/.envfor Hybrid. Back it up with your data: records encrypted with it can't be read without it. - Production-mode installs (
docker-compose.prod.yml/deploy.sh) must setHR_ENCRYPTION_KEYSthemselves and refuse to start without it. Generate one withecho "k1:$(openssl rand -base64 32)". - Existing custom roles keep engineering access. Give HR staff the new HR role, and Sales or other non-engineering staff the Sales role.
## Pluggable AI provider — Claude, GitHub Copilot, or OpenAI Codex
Pluggable AI provider — Claude, GitHub Copilot, or OpenAI Codex
Each organization now chooses which agent CLI runs its codebase-aware work. Pick your provider in the setup wizard and Settings → AI Configuration; every agent task — specs, chat, design, estimation, scans, triage — runs on it.
Highlights
- Three providers, one selector. Claude Code, GitHub Copilot, or OpenAI Codex, switchable per org. All reach the same BUD/code-graph tools through the built-in MCP bridge.
- Flexible auth per provider. Hybrid mode inherits your host login (
claude login,gh/Copilot, orcodex login) so agents run on your existing subscription — or paste an API key/token, stored encrypted (Fernet AES-128). - In-wizard connection test for the selected provider before you commit.
- Provider-native repo setup. The first scan writes the right config for your provider — Claude →
.claude/, Codex →.codex/hooks.json+config.toml, Copilot →.github/hooks/, plusAGENTS.md— all wired to the same dev-activity / BUD-todo hooks so tracking works everywhere. - Live progress for Codex & Copilot — runs now stream tool activity instead of showing a static spinner.
- Provider-aware Agent Prompts. Model/effort dropdowns reflect the active provider (Copilot runs on auto-selected models).
Notes
- Claude behavior is unchanged; existing orgs keep running exactly as before.
- Ollama (local models) is next on the roadmap.
v1.3.3
What's Changed
- fix(minigames): bound submitted score to each game's own max by @mickyarun in #258
- feat(minigames): make all three games server-authoritative via Colyseus by @mickyarun in #259
- chore(frontend): enable npm-merge-driver in .gitattributes by @mickyarun in #260
- fix(minigames): score fishing hook from the client's in-cast time by @mickyarun in #261
- fix(minigames): show fishing result instantly, stop snapping the bobber by @mickyarun in #262
- fix(minigames): match optimistic fishing score to the reported instant by @mickyarun in #263
- feat(minigames): ramp Pollen Pop — faster, denser, more varied over time by @mickyarun in #270
- feat(race): dynamic pack-framing spectator camera by @mickyarun in #271
- fix(minigames): cap Pollen Pop end density, speed it up, jitter the cadence by @mickyarun in #272
- feat(minigames): randomise the Lake Fishing bobber per cast by @mickyarun in #273
- feat(minigames): lives + escalating levels for Fishing and Pollen Pop by @mickyarun in #274
- feat(minigames): firefly anti-bot guards + clear inflated leaderboard scores by @mickyarun in #275
- feat(buds,bugs): unblock PM/manager workflows and add bug numbers by @mickyarun in #276
- feat(sp): rework Skill Points to credit the people who did the work by @mickyarun in #277
Full Changelog: v1.3.2...v1.3.3
v1.3.2 — Podman-friendly frontend DNS
Fixes
Self-hosting on Podman now works out of the box. The frontend nginx container previously hardcoded Docker's embedded-DNS address (127.0.0.11) for resolving the backend upstream. Under Podman that resolver lives on the network gateway instead, so every /api/* request failed with 502 Bad Gateway and the setup screen couldn't reach the backend.
nginx now auto-detects the container's DNS resolver from /etc/resolv.conf at startup, so the same image runs unchanged on Docker, Podman, and rootless Podman — no manual config edits. The runtime DNS re-resolution that keeps nginx pointed at the right backend after a --force-recreate is preserved.
Full diff: #257
v1.3.1 This release introduces the Company Quiz Game — a daily, AI-generated knowledge game built from your own product data — plus timezone-aware mini-game nudges, org-wide high-score shout-outs, and two garden reliability fixes.
✨ Highlights
🧠 Company Quiz Game (#252)
A daily knowledge game that tests how well your team understands what you've built:
AI-drafted, human-approved — questions are generated from your feature registry into a review queue; admins edit/approve/reject, and only approved questions ever go live. No unreviewed AI output reaches the team.
Three formats — multiple-choice, scramble, and fill-in-the-blank, auto-rotating so it stays fresh.
Runs on your schedule — opens on admin-selected weekdays (default Mon & Fri) at a set time in your org's timezone, with a live countdown and an answer-+-explanation reveal when the window closes.
Competitive & rewarding — accuracy + speed scoring, a monthly leaderboard, and the month's top scorer earns SP (a genuine tie splits the prize). Slack DMs announce when a quiz opens (with the SP prize on the line) and post the standings at reveal.
Surfaced right on the dashboard garden, alongside the mini-games. Configure it under Settings → Quiz Game.
🎮 Mini-games
Daily Slack nudge now fires at 09:00 in each org's own timezone (#248)
New high scores are broadcast to the whole org on Slack (#250)
🐛 Fixes
Garden: self-heal invisible repo trees caused by a corrupt IndexedDB cache (#249)
Garden: stop canvas input from bleeding through open dialogs (#251)
📝 Upgrade notes
The Company Quiz Game adds new database tables; the Alembic migration runs automatically on backend startup. It's enabled by default and can be turned off per-org in Settings → Quiz Game.
Full Changelog: v1.3.0...v1.3.1
v1.3.0 — Living Garden & Race Day 🌳🏁
This release is all about the 3D world coming alive: a top-to-bottom performance and visual rewrite of the Living Garden, a brand-new mini-games arcade with leaderboards, a richer racing experience, and a refreshed public landing site + docs.
🌳 Living Garden — rewritten for speed and beauty
A ground-up overhaul of the PlayCanvas dashboard world (#247):
Much faster. Hardware instancing, smarter draw-call batching, a parallelized scene build, and incremental updates — a single repo change now regrows just that one tree instead of rebuilding the whole world. Holds 60fps even at 24+ repos.
Prettier. New single-source color theme, spring-morning lighting, camera post-processing, grass that sways in the wind, drifting pollen, and a dense painterly grass field.
A world worth visiting. Natural wood-bark trees, a teak-plank swimming pool, a packed-earth village with a campfire, benches, yards, and a washing line.
🎮 Mini-games arcade (new!)
A new way to take a break in the garden (#247):
Two games — Lake Fishing and Pollen Pop — with a daily play streak.
Per-game leaderboards, now first-class tabs on the Leaderboard page right next to XP and the race circuits.
A daily Slack nudge to keep your streak alive and call out who's leading.
Games are pure fun: they never award XP — XP stays reserved for real development work.
🏁 Racing
Invite colleagues by Slack DM to race, with a reusable member picker and a more robust test setup (#235).
Boost pads and hurdles added to tracks — with knockdowns, jumps, and speed-scaled launch arcs (#236).
A new circular/organic-loop track with laps, a roadside village, and AI dev bots to race against (#242).
🌐 Landing site & docs
The cramped single-page landing is now a 6-page SEO marketing site with animation (#240).
README restructured as a proper landing page, with reference docs moved into docs/ (#243), refreshed promo imagery and screenshots (#244), and shipped features (Slack bot, GitHub webhook, test-gen, CI/CD) now marked Live today (#246).
🐛 Fixes
Vuetify components now correctly inherit the Hallmark body font (#239).
Full changelog: v1.2.0...v1.3.0
🌿 v1.2.0 — A fresh coat of paint, and a starting gun 🏁
This release reskins Bodhiorchard end-to-end around a new forest design system, adds automatic light/dark theming, and ships Racing v2.
🎨 A whole new look — the Forest design system
Every screen redesigned around a "living garden" visual language.
- Forest palette — a deep-forest dark canvas, one disciplined leaf-green accent, and a reserved harvest-gold for reward moments (XP, shipped BUDs, the leaderboard podium).
- New typography — Bricolage Grotesque for headings & the wordmark, Geist for the UI, JetBrains Mono for code.
- Light & Dark, automatically — the app now follows your OS colour scheme and switches live (e.g. at sunset). Prefer a fixed mode? Pick Auto · Light · Dark in the user menu and it's remembered.
Highlights
- BUD detail keeps the phase tabs and section toolbar pinned as you scroll a long document.
- Leaderboard gets a gold podium for #1 and a focused, readable ranking.
- Design Systems cards now render the extracted palette as real colour swatches.
- Methodology is now a polished, cinematic landing page.
- Consistent passes across the BUD board, Dashboard, Features, Bugs, Skills, Members, and Settings, plus accessibility touches (visible focus rings, reduced-motion support, theme-aware tooltips & scrollbars).
🏁 Racing v2
Race your teammates through the 3D garden.
- Lobby → track → podium — invite racers, line up in the lobby, race, and see podium results.
- Stamina mechanic — manage stamina across the race, with a quick intro and a live stamina bar.
- Invite flow — invite one or more colleagues, with live invite toasts and a watch banner for spectators.
- A finish banner, proximity hotkeys, and a per-distance race leaderboard.
👥 Teams
- Group org members into teams and map them to the repositories they own, so new work auto-routes to the right squad based on which repos a BUD touches.
🛠 Under the hood
- Priority-aware smart BUD assignment with a yield-offer flow.
- Numerous light-theme correctness fixes and component cleanups.
v1.1.0
v1.0.1 - Teams: route work to the squad that owns the code
Auto-assignment now narrows to the team that owns a BUD's impacted repos
before applying the existing capacity, skill, recency, and round-robin
scoring. Plus async background jobs for the Danger Zone skill rerun and
proper navigation feedback when opening a BUD.
Highlights
- Teams as a first-class concept (Members → Teams). A team is a
named set of org members mapped to one or more repos. Three new
tables with composite FKs that make cross-org membership a DB-level
impossibility. - Team-scoped auto-assignment. The role picker intersects with team
membership when a BUD hasimpacted_repos. Falls back to org-wide
with a banner warning when no team owns the impacted repo — never
stalls silently. - Per-repo TODO splitting at DEVELOPMENT entry. Multi-repo BUDs
now distribute TODOs across each repo's team developers instead of
dumping every TODO on the lead. - Members page consolidated into two tabs (People / Teams);
/settings/teamsredirects so old bookmarks keep working. - BUD board Team filter alongside the existing assignee / priority
filters. - Danger Zone skill rerun moved to the async-job pattern — no more
axios client-side timeouts on 20-repo orgs. - Opening a BUD shows immediate feedback via a layout-level route
progress bar + a page-shaped skeleton loader.
Migration
One new migration runs on backend start; no data backfill required.
Existing orgs get empty Teams — admins create them via Members → Teams.
Tenant safety
The three new Teams join tables enforce tenant alignment via composite
FKs back to teams, org_to_user, and tracked_repositories.
Cross-org membership or repo ownership is rejected at the database
layer regardless of application bugs.
What's Changed
- fix(skills): run Danger Zone skill rerun as a background job by @mickyarun in #218
- fix(buds): add navigation feedback so opening a BUD never feels stuck by @mickyarun in #224
- feat(teams): team-scoped BUD auto-assignment + Members → Teams tab by @mickyarun in #225
Stable Initial Version
What's Changed
- Relicense AGPL-3.0 → Apache 2.0 + add IP-independence NOTICE by @mickyarun in #48
- Public-launch follow-ups: BUD canonicalization, public-readiness docs, landing page by @mickyarun in #49
- README polish: move independence note to footer, mark Anthropic direct API as supported by @mickyarun in #50
- README + landing-page narrative restructure: How It Runs, AI Engines, Manifesto, FAQ, JSON-LD schema by @mickyarun in #51
- Position as Agent-Driven Development methodology (not agentic SDLC) by @mickyarun in #52
- feat: MCP-ify BUD design read/write + cascade reassignment to TODOs by @mickyarun in #53
- feat: surface BUD ↔ Feature links in Requirements tab + audit timeline by @mickyarun in #54
- fix+feat: BUD design wireframe preview, linked-features UI, PM agent grounding by @mickyarun in #55
- prompt: wireframes are plain HTML — drop Vue/Vuetify, retire v-app autowrap by @mickyarun in #56
- feat: BUD ↔ Feature links UI, wireframe preview fixes, todo regen on chat edit by @mickyarun in #57
- refactor: regenerate BUD todos only at DEVELOPMENT-phase transition by @mickyarun in #58
- scan: domain-overlap guard + compound-bigram cluster labels by @mickyarun in #60
- chore: replace customer-specific placeholders with generic ones by @mickyarun in #61
- feat(buds): Notion-style TODO board + agent-driven generation via job queue by @mickyarun in #59
- removed the deployment file by @tanu-atoa in #62
- chore(backend): clear all mypy --strict errors + open-source PR gates by @mickyarun in #63
- fix(ci): unblock Dependabot PRs (hatchling pin + DCO bot exemption) by @mickyarun in #90
- chore(deps)(deps): update tzdata requirement from >=2024.1 to >=2026.2 in /backend by @dependabot[bot] in #84
- chore(deps)(deps): update python-dotenv requirement from >=1.0.0 to >=1.2.2 in /backend by @dependabot[bot] in #81
- chore(ci)(deps): bump actions/setup-node from 4 to 6 by @dependabot[bot] in #64
- chore(ci)(deps): bump googleapis/release-please-action from 4 to 5 by @dependabot[bot] in #66
- chore(deps)(deps): update bcrypt requirement from <4.1 to <5.1 in /backend by @dependabot[bot] in #77
- chore(deps)(deps): bump the multiplayer-minor-patch group across 1 directory with 2 updates by @dependabot[bot] in #70
- chore(deps)(deps): bump the frontend-minor-patch group across 1 directory with 8 updates by @dependabot[bot] in #74
- perf(design): real --resume + iteration_model + no-link fallback by @mickyarun in #89
- chore(ci)(deps): bump amannn/action-semantic-pull-request from 5 to 6 by @dependabot[bot] in #65
- chore(ci)(deps): bump github/codeql-action from 3 to 4 by @dependabot[bot] in #67
- chore(ci)(deps): bump actions/setup-python from 5 to 6 by @dependabot[bot] in #69
- chore(ci)(deps): bump actions/checkout from 4 to 6 by @dependabot[bot] in #68
- feat(security): layered defense for --dangerously-skip-permissions claude subprocess by @mickyarun in #92
- fix(claude-guard): post-merge bug fixes — MCP unblock, NPROC, env-exfil by @mickyarun in #93
- chore(ci): drop CodeQL quality queries, keep security-extended by @mickyarun in #94
- fix(security): block partial-SSRF in GitHub install-repo client by @mickyarun in #95
- refactor(bud-detail): split into focused subcomponents + composables by @mickyarun in #97
- feat(bud): phase-gated section edits + role fallback chain by @mickyarun in #98
- feat(bud-chat): session resume for all states + log-leak + concurrency + design auto-trigger fix by @mickyarun in #99
- fix(security): truncate password input to 72 bytes before bcrypt by @mickyarun in #100
- fix(bud-design): per-design cancel, per-repo banners, tracker loop fixes by @mickyarun in #101
- feat(bud-chat): durable resume, cancel, and restart recovery for AI Editor chats by @mickyarun in #102
- feat(pr-merge,features): phase 5 Redis-stream redesign + lineage UI by @mickyarun in #103
- feat(skills): custom skills per agent type, per-BUD stage overrides, override-aware resolver by @mickyarun in #104
- fix(skills): prune orphan agent_skills rows mis-tagged by step-4 fallback by @mickyarun in #105
- feat(scan,pr-merge): route subsequent scans through the diff-based engine by @mickyarun in #106
- fix(rbac): manager team:manage + developer buds:edit seed by @mickyarun in #107
- feat(design-systems): user customisations preserved across re-extraction by @mickyarun in #120
- fix(scan): include .bodhiorchard/mcp_bridge.py in bootstrap PR by @mickyarun in #121
- fix(nginx): proxy /mcp/ to backend for example-repo bridges by @mickyarun in #122
- fix(hooks): resolve tracked repo by basename when laptop path misses by @mickyarun in #123
- fix(xp): resolve PR author to BUD assignee when github_username misses by @mickyarun in #124
- feat(xp): stage-gated XP awards split among BUD contributors by @mickyarun in #125
- feat(code-review): classify git-auth failures + retry-once + explicit cwd by @mickyarun in #126
- fix(rbac): grant QA role buds:test for evidence upload + result entry by @mickyarun in #127
- fix(github): preserve valid inline review comments on 422 by @mickyarun in #128
- fix(github): post code-review on every re-run, not just the first by @mickyarun in #129
- fix(storage): surface S3 errors as FileStorageError + bump success log by @mickyarun in #133
- feat(qa-evidence): pydantic-settings storage
- filename sanitiser + scoped UI errors by @mickyarun in #134
- feat: external-LLM mode + remote MCP endpoint + per-phase auto-generate by @mickyarun in #136
- fix(mcp-connect): show full URL when VITE_API_BASE_URL is a bare path by @mickyarun in #137
- feat(mcp): write-side BYO-AI tools + append-only BUD edit history by @mickyarun in #138
- fix(webhook): use webhook_event= kwarg to avoid structlog event collision by @mickyarun in #132
- fix(bud-assignment): phase-scope continuity to stop cross-phase bleed by @mickyarun in #135
- fix(security): safe_join helper + targeted path-injection guards by @mickyarun in #142
- fix(qa-evidence): always-visible delete button on evidence tile by @mickyarun in #140
- feat(qa): allow reverting manual result to pending by @mickyarun in #141
- feat(tech-spec): figma URL on Design tab + local-Claude tech-spec generation by @mickyarun in #143
- feat(figma-tech-spec): add figma URL field and make tech spec sections universal by @mickyarun in #146
- feat(mcp): accept bud_number on get_bud_by_id by @mickyarun in #147
- fix(narrow-synthesis): containment tier + deactivate filter prevent wrong-deactivation by @mickyarun in #148
- fix(dev): add missing dev-start.sh referenced by package.json by @mickyarun in #149
- feat(bud-board): add title and assignee filters to kanban header by @mickyarun in #150
- feat(slack): feature Q&A + harden triage duplicate det...