Releases: tinyhumansai/opencompany
Release list
OpenCompany v0.1.7
The Safer Keys & Smarter Console Upgrade
v0.1.7 is here with a big step forward in day-to-day operator ergonomics: cleaner connection management, stronger credential safety guarantees, and chat + room UI improvements that make conversations easier to run and audit. 🎉
Highlights
Chat & collaboration 💬
Chat is more capable and more traceable in this release: you can now stage multiple attachments per message (drag, drop, and paste), and chat-produced files are linked back into replies for smoother follow-through from “make it” to “here it is.” Cross-conversation context also gets smarter: seats can carry context across conversations, while private asides are collapsed/elided more consistently to keep shared threads clean.
(#2119, #2295, #2247, #2323) — Thank you @theamazinghenk, @oxoxDev, and @sanil-23!
Accounts, connections, and provider pages 🔌
The console’s provider and account surfaces were reshaped for clarity and consistency: Composio credentials are now presented as single-select rows, and both the LLM and Search areas align to a consistent “provider list” shape. The Account page also got a layout + UX pass (including “two ways to connect” flows), and one edge case in create/reset availability is now pinned to the build that actually offers it.
(#2278, #2279, #2280, #2304, #2302, #2121) — Thank you @graycyrus and @sanil-23!
Credential safety & key rework foundations 🔐
Credential handling is tighter across the stack: inference endpoints, names, and logs are protected so credentials don’t “survive” in unsafe places, and the repository now includes a docs-only phased implementation plan for the broader keys rework effort. This lays clear groundwork for safer key flows and clearer operational guarantees as the system evolves.
(#2281, #2307, #2306) — Thank you @graycyrus!
Skills, setup, desktop, and deploy reliability 🛠️
A set of targeted fixes improve admin and runtime reliability: Skills UI now reports and renders more consistently (including better MCP row behavior and cleaner admin notices), setup probes discovered models with typed failure classification, and first-run flow can skip model selection when the host already supplies one. Desktop wiring and deploy attribution were also hardened to reduce configuration drift and improve correctness in tenant pod attribution.
(#2314, #2315, #2316, #2317, #2318, #2325, #2324, #2328) — Thank you @oxoxDev!
CI, tests, and core layout 🧪
Release + CI flows were updated for branch ruleset compatibility (including a PR CI gate and adjusted production cutting), and end-to-end testing coverage was reinforced around provider setup paths. The repo also saw a significant internal re-org: keys rework work was merged and host sources moved under crates/opencompany-core, with follow-up commits to bring main back to green and modernize test harness setup.
(#2299, #2298, #2301, #2309, #2310, #2312, #2270, #2311, #2327, #2329) — Thank you @senamakel, @graycyrus, and @sanil-23!
Documentation clarity 📚
Docs were refreshed to make operator expectations explicit—most notably, the README now states plainly whether OpenCompany “phones home,” and documentation continued to evolve alongside the provider list and key rework initiatives.
(#2154) — Thank you @graycyrus!
Contributor Credits
Thank you to all contributors who shipped v0.1.7: cyrus, Cyrus Gray, github-actions[bot], Jarno de Vries, oxoxDev, sanil-23, Steven Enamakel.
Full Compare
Contributors
Assets 10
OpenCompany v0.1.6
The Reliability & Release Automation Upgrade
Between stronger runtime invariants and a smoother path to shipping, v0.1.4 → v0.1.6 makes OpenCompany more durable under restarts, safer under concurrency, and easier to release with confidence. 🥂
Highlights
Runtime durability & authority safety 🔐
A big slice of this release is about making authority decisions and approvals survive the real world: restarts, partial cycles, and edge-case state transitions. Standing grants now properly rehydrate at boot (with lifetime validation), consumed single-use grants won’t re-arm after a crash window, and task review verdicts are refused unless a card is actually in the right state (in_review).
(#2226, #2267, #2251) — Thank you @oxoxDev!
Workflows & blockers that don’t strand 🧭
Workflow continuation and blockers handling got sturdier, especially around “resume” and “skip” flows. Global workflows can now be correctly resolved when a resume reloads the graph, preventing stranded runs after an operator answers a clarifying question. And for task-backed blockers, “Skip” now truly settles the underlying task card without dispatching another attempt—while keeping the rest of the node/task actions consistent.
(#2236, #2268) — Thank you @oxoxDev!
Console tools: safer pages & cleaner OAuth handoffs 📄
Operator-facing tooling got safer and less confusing. pages_read now bounds output and supports continuation via byte offsets, and pages_delete is protected against removing pages that are still linked elsewhere. On the integration side, Composio authorization now deduplicates in-flight OAuth handoffs—so retries won’t accidentally spawn a second connect flow and orphan the first.
(#2255, #2246) — Thank you @oxoxDev!
Inference: provider routing docs & “Managed-only” restart clarity 🧠
Inference configuration got a significant documentation-and-correctness pass focused on multi-provider reality (dialects, routing assumptions, and provider-specific behaviors), along with a concrete runtime fix: if a company is configured such that only Managed is set up, the system now reports that a restart is required instead of silently appearing “configured” while staying stuck on the offline echo brain.
(#2262, #2289) — Thank you @graycyrus!
Spend controls, reachability invariants, and tougher tests 🧪
Budget enforcement is now real at dispatch-time: the monthly inference cap is enforced before every built-in inference call (including confined turns), closing gaps where companies could keep spending after hitting their declared cap. MCP server connections are also guarded by a reachability invariant—preventing “healthy/connected” servers that no teammate can actually access. Alongside these fixes, test coverage was expanded to lock in concurrency, policy edge-cases, and authority boundaries across store/server/workflows.
(#2265, #2269, #2219, #2237) — Thank you @oxoxDev!
Release ops: version bump + one-dispatch release flow 🚀
Shipping is now far more streamlined. This range includes the v0.1.5 version bump, and a revamped CI release pipeline that moves from hand-tagging across multiple workflows to a single dispatch flow that promotes, stages, and then ships to production with clearer gates and artifacts.
(#2296, #2297) — Thank you @oxoxDev and @senamakel!
New Contributors
- Thanks @github-actions[bot] for contributing “chore(release): v0.1.6 [skip ci]”, helping keep the release process moving smoothly.
Hope you’ll stick around and join the community—contributor rewards are waiting for you.
Contributor Credits
Thank you to everyone who contributed to this release: cyrus, Cyrus Gray, github-actions[bot], oxoxDev, Steven Enamakel.
Full Compare
Contributors
Assets 10
v0.1.4
The Crates & Console Reliability Upgrade
v0.1.4 is here with a cleaner Rust workspace foundation and a noticeably smoother desktop console experience. From build + CI alignment to fixes that make links and auth flows behave correctly in the packaged app—this release tightens the whole operator loop. 🎉
Highlights
Workspace & packaging foundations 🧱
OpenCompany now follows a clearer crates/… layout, aligning with sibling TinyHumans libraries: a virtual workspace root, shared workspace package/dependency configuration, and explicit crates for core, TUI, and the desktop app shell. This makes builds, tooling, and long-term modularity much easier to reason about.
To cap it off, all version declarations were reconciled back into agreement for the v0.1.4 cut, eliminating drift introduced during the workspace restructure. (#2258, #2285) — Thank you @senamakel and @oxoxDev!
Desktop console link handling & shell integration 🖥️
Outward links in the packaged desktop app now reliably open in the operator’s default browser, instead of silently doing nothing in the Tauri webview. This work includes wiring the console to delegate external navigation appropriately, and ensuring the shell capability is placed where the desktop app crate actually lives after the crate move.
The end result: all those “open in a new tab” links behave like operators expect—without requiring guesswork or manual copy/paste. (#2283, #2284) — Thank you @oxoxDev!
Workflow scheduling correctness ⏱️
The workflow scheduler now enumerates workflows from the global baseline in addition to a company’s seed files and overlays, so scheduled globals don’t disappear silently. Company precedence rules and global opt-outs remain respected, and the scheduler can correctly report scheduled work even when a company isn’t wired to run it yet. (#2261) — Thank you @oxoxDev!
E2E coverage for Composio authority flows 🧪
The console E2E suite was stabilized by anchoring Composio authority checks to the actual key field the console exposes, fixing timeouts and expectation drift in connection-flow tests. This brings the lanes back to meaningful signal and protects against regressions in role-based token UI behavior. (#2271) — Thank you @oxoxDev!
Contributor Credits
Thanks to oxoxDev and Steven Enamakel for shipping v0.1.4!
Full Compare
Contributors
Assets 10
v0.1.3
The Safer, Smoother Operator Console Upgrade
v0.1.3 tightens OpenCompany’s safety rails while making the console feel faster, clearer, and more “ready to run a company.” Lots of governance hardening, a more navigable console, and big momentum in hive-mind collaboration. 🎉
Highlights
Console UX & navigation polish 🧭
The console got a major layout and information-architecture pass: tabbed pages, a reworked agent page, clearer “Agents/Automations” vocabulary, a /company route prefix, and a pinned Room rail with a cleaner two-column layout. Search is now wired into the title row for quick jumps across agents, channels, files, and conversation messages, and notifications/approvals are consolidated into a dedicated tabbed surface.
Onboarding and everyday account flows are smoother too: the New-workflow dialog is simplified when a copilot can draft, adding teammates is streamlined (with better landing), members no longer see settings controls they can’t use, and sign-out is available from both the account menu and settings. (#1988, #1998, #2046, #2049, #2087, #2118, #2133, #2134, #2159, #2160, #2162, #2214, #2225, #2239, #2245, #2254, #2256, #2263, #2108, #2228) — Thank you @graycyrus, @oxoxDev, @sanil-23, @senamakel, and @CodeGhost21!
Chat & collaboration reliability 💬
Chat attribution and thread handling were strengthened: shared-desk seeds are now attributed correctly to their speakers (with the follow-up consolidation), blockers are threaded off the message they answer, and the system better reports “@name” mentions that reached nobody while preventing edge cases from disrupting core channels like #general.
Message sending UX is more explicit too, including clearer failed-send rendering and routing for which conversation is open. (#1956, #2075, #2161, #2217) — Thank you @CodeGhost21, @graycyrus, and @oxoxDev!
Hive-mind desks & multi-agent deliberation 🐝
This release makes the hive substantially more tangible: deliberation UI, an activity graph, a grammar installer, private asides, and cross-desk referral improvements sharpen both visibility and coordination. On the runtime side, hive-mind desks and memory drivers expand what teams can model and remember, and the bundle work adds more realistic examples that exercise agent-to-agent communication.
Also included: multiple bugfixes and wording tweaks from live runs, plus recording crossings so that “a question to a person” becomes a conversation with them. (#2097, #2109, #2116, #2129, #2166, #2204, #2206, #2207, #2244, #2168) — Thank you @senamakel and @sanil-23!
Security, governance & policy hardening 🛡️
A wide sweep of authorization and governance work lands in v0.1.3: admin authority is now required across company lifecycle routes, skill writes, approval decisions, permission revokes, and memory-engine repointing. Scope handling and read-plane scoping were tightened, capability refusal interpretation improved, and multiple routes and matrices were pinned with tests to prevent regressions.
Policy behavior also becomes clearer and safer: read-shaped HTTP calls are free, emergency stop now halts operations (and is enforced at boundaries), and new “shadow-only” measurement work explores consequence floors without changing decisions yet. Documentation was updated to reflect what still blocks calls and how the consequence floor is intended to work. (#1989, #2047, #2077, #2086, #2091, #2125, #2127, #2132, #2137, #2140, #2145, #2151, #2152, #2163, #2164, #2170, #2174, #2175, #2176, #2178, #2180, #2184, #2199) — Thank you @oxoxDev!
Ledger, runtime & data integrity 📒
The ledger and persistence layers got stricter and more resilient: ledger writes are refused if they can’t be read back, concurrent ledger declarations are serialized, and multiple surfaces now bound reads/writes to prevent oversized or racing operations. Workspace improvements include safer attachment handling, atomic refusal of racing note writes, and better mobile layout behavior.
On runtime and operations: git is installed in the runtime image for workspace checkpoints, desktop updater key handling is fixed, and multiple boot/config edges are now refused early (e.g., explicit backend URL requirements). (#2029, #2040, #2063, #2088, #2112, #2120, #2141, #2186, #2230, #2243) — Thank you @oxoxDev, @CodeGhost21, and @theamazinghenk!
Inference, tools & harness correctness 🧠
Inference and tool-call handling improved meaningfully: providers’ model vocabularies are discovered rather than assumed, and tool calls written as <invoke> markup are recovered more reliably while guarding chat-only turns against leaked tool-call markup. The harness became more strict and fail-closed in key places, including registry tool gating and refusing dispatch when spend caps can’t be measured.
Tooling, publishing, and ops edges were tightened too: shell timeout promises are honored, high-risk shell guards and CSV export ceilings are enforced, webhook deliveries get deadlines, and multiple idempotency paths (including Composio executions) are stabilized. (#2045, #2093, #2095, #2126, #2128, #2139, #2153, #2188, #2192, #2208, #2212, #2249) — Thank you @oxoxDev and @senamakel!
Setup, integrations & connectivity 🔌
Setup and connectivity flows are more guided and robust: model selection is consolidated into a single picker with clearer “get a key” and “way past it” paths, propose_roster is rate-limited per company, and hosted tenants can default the tiny.place URL whil...
Contributors
Assets 10
v0.1.2
The Operator Confidence Upgrade
v0.1.2 is all about making OpenCompany more trustworthy in production: operators can resolve blockers with the full set of outcomes, workflow runs report spend honestly, and the desktop/console are easier to keep updated and observable. 🎉
Highlights
Blockers & operator controls 🧭
Operators can now answer a parked blocker using the full four-way verdict set (not just a binary approve/deny), bringing the console/server flow in line with the underlying runtime semantics and making “amend/skip/retry/cancel” actually reachable from the operator surface. (#2038) — Thank you @oxoxDev!
Workflow run-state & metering integrity 📊
Workflow runs now report spend and state more honestly—especially in failure cases—so real work (tokens, tool calls, cost) doesn’t disappear, panels don’t contradict each other, and orphaned runs are cleaned up instead of lingering. (#2053) — Thank you @graycyrus!
Memory engines: Cortex as a configurable option 🧠
OpenCompany can now register CortexDB as a selectable memory engine (without changing defaults), making evaluation and rollout a configuration choice rather than a larger integration project. The accompanying spec was also corrected to withdraw an earlier finding after discovering enrichment had been switched off during measurement. (#2065, #2076) — Thank you @CodeGhost21!
Desktop updates: finally automatic 🔄
The desktop app now supports auto-updates end-to-end, including updater wiring and the required public signing key so releases can be verified and actually shipped. (#2066, #2078) — Thank you @graycyrus and @sanil-23!
Observability: Sentry for host + console 🔎
Crash and error reporting is now available for both the Rust host and the React console via operator-owned Sentry configuration, with the console side unblocked for hosted tenants by passing the Sentry DSN and build commit at image build time. (#2067, #2073) — Thank you @graycyrus and @sanil-23!
Console product scoping 🏢
The console is now scoped to a single company per install, removing managed-only surfaces from the UI and turning the company switcher into a static label—simplifying the experience without changing Rust-side behavior or migrating stored data. (#2074) — Thank you @oxoxDev!
New Contributors
- graycyrus — Thank you for your first contribution! You improved workflow run-state accuracy and spend accounting (and related cleanup), strengthening trust in what the console reports. (#2053)
Hope you’ll stick around—OpenCompany has contributor rewards, and we’d love to see what you build next.
Contributor Credits
Thank you to everyone who shipped v0.1.2: CodeGhost21, cyrus, Cyrus Gray, Ghost Scripter, graycyrus, oxoxDev, sanil-23.
Full Compare
Contributors
Assets 10
v0.1.1
The Blocker-to-Answer Upgrade
v0.1.1 sharpens OpenCompany’s day-to-day operating loop: workflows pause and resume more reliably, approvals survive restarts, and the console gets cleaner navigation and clearer task states. Shipping a sturdier “ask → answer → continue” cycle feels great. 🎉
Highlights
Workflows & runtime resilience ⚙️
This release strengthens how work waits and continues: blocked steps can park instead of failing, re-enter when answered, and even restart from checkpoints. We also tightened node journaling, postconditions, verdict honesty, peer consultation scoping, and paused-company behavior so runs do what they promise—even across restarts.
(#1757, #1781, #1861, #1862, #1863, #1865, #1866, #1882, #1883, #1937, #1990, #1991, #2006, #2027, #2033, #2036) — Thank you @oxoxDev and @sanil-23!
Approvals, blockers, and “answer to resume” ✅
Approvals are now more explicit and durable: parked approvals can resume after a restart, expired approvals correctly settle the run, and the console can approve a blocker and reliably deliver output. The system also clarifies which message is the root for resuming, and avoids unlinking parked approvals on partial lookup failures.
(#1816, #1825, #1891, #1895, #1925, #2004, #2008, #2017, #2030, #2034) — Thank you @oxoxDev, @sanil-23, and @senamakel!
Console UX, navigation, and connection surfaces 🧭
The console got a cohesive UI pass: consistent headers, improved layout/padding, a clearer Overview header, a consolidated title row, and a revamped sidebar structure that cleanly exposes Room/Company/Connections/Flows. We also moved OAuth + MCP into a top-level Connections section, improved task details truthfulness, improved money rendering consistency, and added wallet-mode provisioning flows (including safer create/reset and archive cleanup).
(#1785, #1828, #1893, #1910, #1912, #1914, #1931, #1943, #1953, #1969, #1977, #1980, #1983, #1984, #1987, #1995, #2025, #2050, #2051, #2061, #2064, #2069, #2070) — Thank you @graycyrus, @oxoxDev, @CodeGhost21, and @sanil-23!
Chat, threads, and live receipts 💬
Threads are treated more consistently as true context boundaries (not just a UI fold), with fixes across dispatch relays, thread identity, and where nested review feedback renders. Live receipts for sent instructions landed and were hardened through the sync→async detach handoff, and read-only channels now correctly hide composer and reaction UI.
(#1852, #1881, #1890, #1896, #1934, #1935, #1972, #1981, #1986, #2001, #2021, #2024, #2042, #2044) — Thank you @oxoxDev and @sanil-23!
Harness, tool routing, Composio, and inference correctness 🧠
Tooling and inference got a big correctness pass: capability routing is more native-first (Composio only when connected accounts require it), Composio calls were standardized, and edge cases in inference parsing (including reasoning/array content) were handled. We also tightened evidence planning behavior, prevented unwanted card minting from board-op chat, improved empty-turn explanations, ensured models’ text-written tool calls can still run, and added read-side task/run tools.
(#1759, #1779, #1780, #1859, #1922, #1941, #1946, #1949, #1951, #1954, #1994, #2010, #2011, #2013, #2016, #2019, #2026, #2068) — Thank you @oxoxDev and @sanil-23!
Storage, workspace, memory, and durability 🗄️
We closed several races and durability issues across workspace/store paths (including rollback and adoption/orphan edge cases), improved boot-time memory probing, and made Brain’s memory list responsive via virtualization. Context stores now share one ranked lexical search, and the system retries stranded archive cleanup more robustly.
(#1801, #1807, #1826, #1876, #1915, #1928, #1968, #1973, #2015, #2023) — Thank you @oxoxDev, @theamazinghenk, and @CodeGhost21!
Onboarding, nudges, and desktop first-run 🚪
First-run is more guided and less error-prone: non-admin members won’t get trapped behind onboarding gates, desktop now opens the wizard instead of seeding a company, and a week-1 nudge encourages new users to save their first workflow (without being auto-acked as operational noise). The onboarding gate is also now explicitly tied to a named company, an integration, and one successful run.
(#1844, #1845, #1875, #1878, #1902) — Thank you @oxoxDev and @sanil-23!
Policy, ledger integrity, tasks, and planning guardrails 🧾
We clarified and tightened operational rules: read-shaped HTTP calls are now free, read-only gating messaging is more ex...
Contributors
Assets 5
v0.1.0
The Release-Ready Console Upgrade
OpenCompany just took a big step from early RC territory into a more dependable, policy-aware, release-ready runtime. 🎉
Highlights
🧠 Delegation & budget-pause correctness
Delegation pause handling is now stricter and more honest: the “skipped CEO relay” logic is correctly gated on a real DESK pause (not a sticky one), and the harness/delegation layer no longer offers “add credits & resend” style redemptions in budget-pause situations where the server can’t actually honor them. This reduces confusing UX loops and prevents wasted inference attempts when a provider is already exhausted. (#1846, #1886) — Thank you Ghost Scripter and @CodeGhost21!
🛡️ Policy: declared hosting tools stay in sync
Policy now explicitly declares the hosting tools that were implicitly introduced via the vendor pin, ensuring they’re tracked in the consequence declaration set and don’t accidentally slide through “read-only” heuristics due to shared hosting_ prefixes. This makes tool governance clearer and safer as the runtime evolves. (#1884) — Thank you @CodeGhost21 (Mega Mind)!
🧪 Console E2E reliability & strict-mode fixes
End-to-end tests got sturdier and less flaky: thread reply matching is now exact (so quoted/echo content can’t break strict-mode locators), and the detached-POST spec now cuts the connection based on an observation rather than a wall-clock guess—improving signal and making failures easier to diagnose. (#1885, #1888, #1889) — Thank you Ghost Scripter and @CodeGhost21!
🚀 Release pipeline: signed macOS DMG + dynamic notes
The release process now produces a more “real” desktop artifact: macOS DMGs are signed and notarized, and release notes generation is automated so releases are easier to ship consistently. This is a major step toward a polished distribution pipeline for the operator console. (#1892) — Thank you @senamakel (Steven Enamakel)!
🧱 Brain fixtures fixed for updated TurnOutcome
Test fixtures were updated to include newly-required TurnOutcome fields, restoring clean compilation for --all-targets and keeping CI green as the brain contract expands. (#1900) — Thank you @senamakel (Steven Enamakel)!
Contributor Credits
Thanks to everyone who made this release happen: Ghost Scripter, Mega Mind, and Steven Enamakel.
Full Compare
Contributors
Assets 5
v0.0.1-rc.2
senamakel
Steven Enamakel
The Release Automation Tune‑Up
This release tightens up the machinery that ships OpenCompany—making the path from change to release smoother and more reliable. 🎉
Highlights
Release tooling & notes 🛠️
Release note generation scripts were updated to improve how changes are captured and presented between versions, helping keep each release clearer and more consistent. Thanks to Steven Enamakel!
CI/CD workflows 🚀
GitHub Actions release workflows were refined, including updates to the main release pipeline and the macOS desktop release workflow—strengthening automation around building and publishing. Thanks to Steven Enamakel!
Contributor Credits
Thank you Steven Enamakel for 5 commits in this release range.
Full Compare
v0.0.1-rc.1
senamakel
Steven Enamakel
The Operator Control Upgrade
This release sharpens the day-to-day operator experience—from smarter model selection to clearer run outcomes and safer controls—so your company’s agents can keep moving with confidence. 🎛️
Highlights
Inference & model picking 🧠
OpenRouter support is now significantly more ergonomic: the runtime can pick OpenRouter models per tier, and the console adds a dedicated model picker backed by a cached catalog and verified tier defaults. Error messages and connectivity probes are also more actionable, especially when model IDs are invalid or a specific harness/provider is the root cause. (#1803, #1838, #1811) — Thank you @oxoxDev!
Safer tool grants & operator controls 🔐
Tool access is now expressible in three clear states—inherit, deny-all, or narrowed via globs—threaded consistently through runtime, server, harness, and console. This makes it easier to lock tools down intentionally without losing the ability to revert to defaults. (#1804) — Thank you @oxoxDev!
Approvals that don’t stall ⏳
Approvals gained a surfaced deadline plus an “extend” lever end-to-end (policy, runtime, server, console, and docs), helping operators keep long-running or parked work moving without losing the audit trail. (#1805) — Thank you @oxoxDev!
Clearer run outcomes & observability 🔎
Runs can now settle as Declined when a refusal is by design—so “declined” no longer reads as “succeeded,” and downstream task/card behavior aligns with reality. Console and observatory visuals were updated to match the neutral semantics. (#1809) — Thank you @oxoxDev!
Chat channels with auto responders 💬
You can now create leadless channels where a selector picks the best-fit responder per message—enabling flexible routing without forcing a single channel lead. The API/docs also capture the responder modes and selector contract clearly. (#1835, #1872) — Thank you @sanil-23!
Budget pause UX & recovery 💳
Budget pauses are now more durable and more recoverable: the console renders a clearer budget-pause card with an Add-Credits CTA (including thread-parented notices), while the runtime/server/harness improve pause propagation, replay, and redemption safety across orchestration paths. (#1846) — Thank you @oxoxDev!
Harness reliability & routing polish 🧰
The harness better respects ACP stop reasons (avoiding false “success” on truncated/empty turns), and test fixtures were tightened to ensure CompanyRecord fields are correctly populated. On the console side, the #/work alias is now constrained to the intended bare route for more predictable navigation. (#1853, #1880, #1797) — Thank you @oxoxDev!
Contributor Credits
Thanks to everyone who helped ship this release: @oxoxDev, @sanil-23, @cyrus, @cyrus Gray, @jarno de Vries, and @steven Enamakel.
