Bug Fixes

• Keep synthetic onboarding to one primary model attempt: no provider retry, fallback, wait Tool, model repair loop, or semantic Completion Gate repair.
• Preserve deterministic onboarding verification and canonical terminal delivery so the streaming bubble and Stop control settle with the Run.
• Count Direct Chat Tool calls within the current user turn instead of substituting the Session-wide aggregate.
• Stop unreadable ledger-bound artifact references from consuming repeated model repair attempts.
• Suppress false Sandbox Gateway blocked-file warnings for unchanged materialized Skill scripts.
• Preserve provider-confirmed Vercel READY completion when the deployment URL is not readable from Runtime.

Verification

• Backend Runtime/onboarding/WebSocket: 179 passed
• Frontend: 122 passed
• Frontend production build: passed
• Scoped Ruff: passed
• Architecture Guard P0: passed
• git diff --check: passed

Upgrade

Pull tag v1.11.4-fix.1 and rebuild/restart the deployment. Database migrations continue to run automatically on application startup.

v1.11.4 — More Reliable Long-Running Tasks, Tool Runtime, and Feishu Group Collaboration

v1.11.4 focuses on sustained execution in real-world workflows. Agents can preserve context, recover execution, and deliver results more reliably across Feishu groups, Direct Chat, scheduled tasks, and long-running Tool calls, while Session and Tenant isolation boundaries are further strengthened.

What's New and Improved

Feishu Group Collaboration and Approvals

• Proactive group discovery and targeted delivery: Authorized Feishu groups can be discovered and selected as Schedule or Trigger targets before sending a message to an Agent. Group identifiers remain resolved server-side according to Tenant, Agent, and authorization relationships; raw chat_id values are not exposed to the model.
• Quiet group listening: Agents can follow group context and respond only when a message genuinely requires a reply. Messages that need no response do not create noise or hold a public group Run indefinitely.
• More complete group context: Structured human @ mentions are preserved, stale read state no longer hides Sessions, and external group conversations remain visible in the correct conversation entry point.
• New Feishu approval tools: Agents can look up approval definitions, upload attachments, create approvals, and query approval status. Write operations are bound to the current Tenant, Run, Tool receipt, and user confirmation, preventing stale confirmation or cross-task reuse.
• Clearer reply feedback: After confirming a reply, an Agent can add a lightweight reaction to the original Feishu group message. Reaction failures remain cosmetic and cannot block or duplicate the text reply.

Durable Runtime and Tool Execution

• Recoverable Tool call contracts: Tool identity, route, call identity, timeout, and execution result are fixed when a call is accepted and remain consistent across retries, Worker recovery, and Checkpoint replay, reducing duplicate execution and route drift.
• More reliable argument validation and repair: Advertised Tool Schemas are validated before Provider invocation. Repairable errors receive bounded feedback, while Runs that exhaust their repair budget fail clearly and release their execution lane instead of looping indefinitely.
• Explicit handling for unknown outcomes: MCP calls, asynchronous Tools, and external writes enter an explicit recovery path when they time out or have an unknown result. A timed-out call is no longer assumed to mean that no side effect occurred.
• Evidence-based completion gate: Before a task finishes, the Runtime checks whether the goal is supported by verifiable evidence. A normal semantic-evaluation failure cannot overwrite confirmed Tool facts.
• Complete Tool history: Tool calls, results, and totals remain available after long-task compaction, history pagination, and Session switching.

Session-Isolated Sandboxes and Long-Running Workspaces

• Session-level output isolation: Code execution and file writes occur in isolated Session workspaces. Tenant and destination paths are validated at publication time to prevent one conversation from overwriting another conversation's artifacts.
• One coherent workspace for long Skill Runs: An Agent Loop reuses its sandbox and Skill snapshot, allowing an in-Run Skill repair to continue without context drift caused by compaction or rematerialization.
• Safer artifact publication: Workspace changes are validated before publication. Identical content is accepted idempotently, and concurrent conflicts retain diagnostic evidence. A workspace publication failure no longer suppresses an already generated chat response.
• More practical Code Executor defaults: python3 is supported consistently. The default execution timeout is now 180 seconds, with a default maximum of 300 seconds; explicit custom settings remain authoritative.

Chat and Task Visibility

• Restored Web Chat streaming: Only persisted, recoverable text is streamed. After a page refresh, network transition, or Worker handoff, the final response can still be restored from the same source of truth.
• Improved Direct Chat history navigation: History is paginated by complete folded Tool Turns, additional pages load at the top edge, and the visible position remains anchored when older messages are inserted. Live auto-follow no longer competes with history pagination.
• Strict Session state isolation: Run state, delayed events, controls, and Tool cards remain attached to their owning Session. Background tasks continue after switching conversations and display their results in the correct place.

Scheduling, Deployment, and Platform Foundations

• Schedules and Triggers share one scheduled occurrence: Timezone handling, Cron evaluation, queueing, idempotency keys, persistence, and Runtime Intake use the same occurrence, reducing duplicate execution and incorrectly timed triggers.
• Vercel deployments wait for an authoritative terminal state: Deployment operations track the exact Deployment and settle only after the Provider reports an authoritative state such as READY, ERROR, or CANCELED.
• More stable default Agent initialization: Deleting, stopping, or renaming the default Agent no longer creates unintended duplicates. Initialization remains one-time and recoverable.
• Stricter multi-tenant data boundaries: APIs, DAOs, organization data, model configuration, Tool history, and enterprise information consistently enforce Tenant Scope. Required backfills and migration convergence prevent cross-tenant reads and hidden historical records.
• Improved release observability: Post-release health-check results can be delivered through Feishu notifications. Database pools, API Workers, and slow-login thresholds can also be tuned for deployment scale.

Bug Fixes

• Fixed long-running tasks becoming unable to continue after context compaction, oversized Tool results, or truncated model output.
• Fixed Runs remaining active after exhausting the Tool repair budget, as well as waiting states failing to release their execution lane.
• Fixed Gemini Tool Results being paired with the wrong Assistant Turn and improved recovery from unknown Provider failures.
• Fixed long-running Vercel BUILDING states being marked as failed too early and inconsistent settlement after status-read failures.
• Fixed sandbox CAS conflicts during concurrent publication, isolated output counting against input quotas, and workspaces being released too early inside an Agent Loop.
• Fixed group Sessions, Tool history, and enterprise information becoming hidden after switching Tenants, as well as WebSockets using the wrong Tenant Scope.
• Fixed Direct Chat history failing to load at the top edge, long Tool Turns being split across pages, scroll jumps after pagination, and background Tool cards disappearing.
• Fixed stale Feishu group read markers hiding Sessions, public-group waiting states consuming unrelated messages, and duplicate event-creation risk after calendar contact lookup.
• Fixed Bubblewrap compatibility on CentOS 7 and Linux 3.10 and added safe proxy propagation into isolated sandboxes.
• Fixed duplicate-column failures during fresh-database migration and converged the v1.11.4 Runtime, Tenant, and Tool migrations onto a single Alembic Head.

Upgrade Guide

Database migrations run automatically during application startup. No manual Alembic command is required. Back up the database and persistent files before upgrading production.

Docker

git fetch --tags
git checkout v1.11.4

cd deploy
docker compose up -d --build
docker compose ps
docker compose logs -f --tail=200 backend frontend

Wait until the API, Runtime Worker, Scheduler, and channel services are healthy before reopening production traffic.

Source Deployment

git fetch --tags
git checkout v1.11.4

cd backend
pip install -r requirements.txt

cd ../frontend
npm install --legacy-peer-deps
npm run build

cd ..
systemctl restart clawith.target

Kubernetes / Helm

Update every Clawith service to the v1.11.4 image, then run:

helm upgrade clawith helm/clawith/ -f values.yaml
kubectl rollout status deploy/clawith-api
kubectl rollout status deploy/clawith-worker

Notes

• Keep versions aligned: The Backend API, Runtime Worker, Scheduler, channel services, and frontend should use the same v1.11.4 build to avoid Runtime or Tool contract mismatches.
• Automatic migrations: Confirm in startup logs that database migration completed and only one latest Head remains. Do not start multiple old and new instances that can run migrations concurrently.
• Feishu permissions: Existing Feishu applications that use group reply reactions or approval capabilities must have the corresponding permissions approved by an administrator. Missing permissions do not affect ordinary text replies but will limit those capabilities.
• Sandbox configuration: SANDBOX_DEFAULT_TIMEOUT now defaults to 180 seconds and SANDBOX_MAX_TIMEOUT to 300 seconds. Review existing Tenant- and Agent-level Tool settings during the upgrade; explicit custom values are preserved.
• Proxy configuration: For sandbox tasks that require controlled network access, use the dedicated SANDBOX_HTTP_PROXY, SANDBOX_HTTPS_PROXY, and SANDBOX_NO_PROXY settings. Avoid exposing unnecessary internal network addresses to the sandbox.

Thank you to all contributors and community members for testing, feedback, and improvements throughout this release.

v1.11.3 — Runtime safeguards and stability improvements

What's New

Core Features

• Enforce L3 approval before Runtime workspace deletion
  • Adds a mandatory high-level approval gate for deleting Runtime workspaces, reducing the risk of accidental or unauthorized data loss.
• Resume originating run after deletion approval
  • Automatically resumes the paused run once deletion is approved, streamlining operational workflows without manual intervention.
• Group runs can reuse standard workspace tools
  • Enables group runs to access ordinary workspace tools, improving reuse and reducing duplication across workflows.

UI/UX Enhancements

• Hide internal system tools from company settings
  • Removes non-user-facing tools from settings, decluttering the interface and reducing misconfiguration risk.
• Clarify runtime budget in agent settings
  • Improves the label/description to make budget limits clearer for operators.

Optimizations & CI/CD

• Automate production deployment after releases
  • Speeds up and standardizes production rollouts post-release, lowering human error and lead time.
• Split Drone CI and tag-triggered deployment
  • Separates build/test from release deployments for safer, more predictable pipelines and clearer traceability.

Bug Fixes

• Runtime stability and compatibility
  • Finalize v1.11.3 Runtime protocol fixes to resolve edge-case inconsistencies and improve interoperability.
  • Ensure Runtime migration remains compatible with existing Alembic storage, preventing upgrade interruptions on deployed systems.
  • Make Runtime maintenance recoverable across legacy group states, improving resilience during background tasks.
  • Keep fresh deployments compatible with logical deletion, avoiding startup or cleanup issues on new installs.
• Upgrade and validation
  • Keep upgrade validation usable across a previously broken release path, ensuring smoother validation and rollback checks.
• Integrations
  • Maintain PPTX document read compatibility with python-pptx, fixing failures when parsing PowerPoint content.
• Chat
  • Fix chat history pagination to use a compound cursor instead of bare timestamps, eliminating duplicate/missing message pages under load.

Upgrade Guide

Standard steps (no manual database commands required; migrations run automatically on startup).

Docker

• Update your environment to use the v1.11.3 image.
• Run:

  docker compose pull
  docker compose up -d
  docker compose logs -f --tail=200
  

• Wait for services to become healthy.

Source

• From your deployment directory:

  git fetch --tags
  git checkout v1.11.3
  pip install -r requirements.txt
  npm install --legacy-peer-deps
  npm run build
  systemctl restart clawith.target
  

• Verify logs and health checks.

Kubernetes

• Update image tags to v1.11.3 in your manifests or Helm values.
• Apply and wait for rollout:

  kubectl apply -f k8s/
  kubectl rollout status deploy/clawith-api
  kubectl rollout status deploy/clawith-worker
  

• Confirm pods are Ready and passing health checks.

Notes

• Runtime/Agent compatibility: This release includes final Runtime protocol fixes for the 1.11.3 line. Ensure all Runtime agents are updated in lockstep with the server to avoid protocol mismatches.
• Automatic migrations: Database migrations run automatically on application startup; no manual steps are required.
• Approval policy: Organizations relying on automated workspace teardown should account for the new L3 approval requirement to avoid unexpected pauses. Configure approvers as needed before rollout.

v1.11.2 — Group creation enhancements and runtime reliability

What's New

Core Features

• Select members during group creation
  • You can now pick group members at the time you create a group, streamlining setup and reducing follow-up steps.
• Unified workspace formats for groups and agents
  • Groups now support the same workspace formats as agents, providing a consistent collaboration experience and simplifying content sharing.
• Preserve history on deletion of models and agents
  • Historical activity and references remain intact when models or agents are deleted, improving auditability and preventing accidental loss of context.

Optimizations

• Faster, single-pass group greetings
  • Simple group greeting flows run in a single pass, reducing latency and improving responsiveness during onboarding.
• Faster CI image builds
  • Image builds now use a faster Alpine mirror, shortening build times in CI pipelines and reducing release turnaround.

UI/UX Enhancements

• Planned group collaboration stays visible
  • Planned collaborative actions remain visible to all group members, improving transparency and coordination.
• Cached group announcements persist across tab switches
  • Announcement content no longer disappears when switching tabs, ensuring consistent visibility across the app.

Bug Fixes

• Restored automatic execution for Agent task schedules
  • Fixes a regression where scheduled agent tasks were not executing automatically, restoring expected automation behavior.
• Restored reliable runtime and end-to-end behavior
  • Addresses multiple stability regressions to ensure predictable runtime and E2E test behavior in v1.11.2.
• Prevented deterministic compaction failures from poisoning retries
  • Runtime retries are no longer affected by deterministic compaction failures, improving overall task reliability.
• Stopped invalid request budgets at compact input boundary
  • Adds guardrails to block invalid request budgets before compaction, preventing downstream errors and wasted cycles.
• Prevented image payloads from failing before provider dispatch
  • Image payloads are correctly prepared and dispatched to providers, reducing pre-dispatch failures.
• Preserved backend error semantics across user-visible surfaces
  • Error semantics are consistent across backend and UI layers, improving clarity and debuggability.
• Preserved human assignments before group planning
  • Human role assignments remain intact when initiating group planning, preventing unintended loss of ownership.
• Kept completed group plans from disappearing silently
  • Completed plans no longer vanish without feedback, improving traceability and user confidence.
• Kept oversized file generation recoverable
  • Oversized file generation now fails gracefully and remains recoverable, reducing disruption in long-running tasks.

Upgrade Guide

Docker

• Update services to v1.11.2 and restart:

docker compose pull
docker compose up -d

Source

• Upgrade to the v1.11.2 tag, rebuild the frontend, and restart services:

git fetch --tags
git checkout v1.11.2

# Install/update dependencies
npm ci
npm run build

# Restart your application processes (example)
systemctl restart clawith || pm2 restart all

Kubernetes

• Deploy the v1.11.2 image and restart workloads:

# Update image tag in your manifests to v1.11.2, then:
kubectl apply -f deploy/
kubectl rollout restart deployment/clawith
kubectl rollout status deployment/clawith

Notes

• Node runtime compatibility
  • Release checkout now enforces a supported Node runtime. Ensure your CI and build environments use a current, supported Node LTS.
• CI and tooling
  • Release notes workflow and GPT model updated; test branches are constrained to deployable source to improve pipeline reliability.
  • Faster Alpine mirror is used during image builds; no configuration changes are required for standard setups.
• No manual database migrations
  • Database migrations run automatically on application startup. No manual steps are required.

v1.11.1 — Runtime Startup and Upgrade Fixes

• Fixed a Backend startup failure caused by the missing checkpoint_migrations table.
• Alembic migrations and LangGraph Checkpoint initialization now run automatically during startup.
• Durable Runtime is enabled by default and is no longer affected by AGENT_RUNTIME_V2_ENABLED=false in legacy .env files.
• Improved concurrency safety during Checkpoint initialization.
• Synchronized Backend and Frontend version display.

Upgrade Guide

Docker Deployment

git pull origin main

cd deploy
# Rebuild and restart services
docker compose down && docker compose up -d --build

Source Deployment

git pull origin main

cd backend
cd ..

cd frontend
npm install
npm run build
cd ..

./restart.sh

Kubernetes / Helm

helm upgrade clawith helm/clawith/ -f values.yaml

Self-Organizing Multi-Agent Group Chat

Humans and multiple Agents can now collaborate in the same group:

• Create and manage groups, members, and multiple independent sessions.
• Use structured @ mentions to precisely activate one or more Agents.
• Multiple Agents can plan responsibilities, work in parallel, and hand tasks off publicly within the group.
• See which Agents are currently planning or working, with support for cancelling active runs.
• Use group announcements, per-Agent group memory, and a shared text workspace.
• Receive real-time messages, recover missed messages after reconnecting, track unread sessions, and restore run status.
• View backend error details, error codes, and Run IDs directly when a task fails, making retries and troubleshooting easier.

Agent Runtime Refactor

The Agent runtime has been rebuilt with LangGraph:

• Automatic context compaction.
• Long-running tasks can recover after a page refresh, network interruption, or Worker restart.
• LangGraph Checkpoints preserve the authoritative execution state.
• Persistent Commands, a tool execution ledger, and delivery records prevent work from being lost or executed more than once.
• Tool calls with uncertain side effects are safely paused and can resume after explicit reconciliation.
• Direct Chat, Tasks, Triggers, Heartbeats, A2A, and external channels now share one durable execution path.
• Direct Chat and Group can both restore progress, tool activity, and final results from authoritative Runtime state.

Session Result Delivery Optimization

When an Agent creates a scheduled task, Webhook, reminder, or long-running task from a Session, the result is now delivered back to the Session where the task originated instead of defaulting to the Primary Session:

• The Origin Session is preserved when the task is created, and completed results are delivered there first.
• Web Sessions receive the new message in the original Session together with the corresponding real-time notification.
• Third-party channels such as Feishu, Slack, and DingTalk use the channel and external conversation information stored on the Origin Session to deliver results back to the original external conversation.
• Agent-to-Agent tasks continue to return to their original A2A Session, preserving collaboration context.
• The Primary Session is used only when the Origin Session has been deleted, is no longer valid, or cannot receive the result. If no Primary Session exists, one is created automatically.
• When a delivery outcome is uncertain, the system reconciles it using idempotency records before considering a Primary Session fallback, preventing duplicate delivery.

Collaboration Relationships Upgraded to Agent Directory

Agents can now discover and contact available collaborators more reliably:

• Find visible human members and other Agents according to access permissions.
• Start A2A or human messaging with stable member IDs instead of relying on name matching.
• Maintain human and Agent directory entries for Custom Agents.
• Public, Custom, and Private Agents continue to follow explicit visibility and isolation boundaries.

Plaza Upgraded to Experience Plaza

Plaza has been upgraded into a new Experience Plaza where team knowledge can be continuously organized, reviewed, and reused:

• Switch between Team Experiences and My Experiences.
• Distill conversations into drafts and edit them directly from Chat.
• Create revision drafts, publish, review, retire, and delete experiences.
• Use Markdown content, tag search, query expansion, and duplicate detection for similar experiences.
• Generate traceable citations when Agents use an experience while keeping historical references readable.
• Share published experiences across the tenant while preserving separate management permissions for drafts and retired content.

Experience and Usability Improvements

• Fixed various bugs.
• Improved numerous product details.

Upgrade Notes

This release includes database schema migrations and LangGraph Checkpoint initialization changes. Back up your database and stored data before upgrading, and make sure that:

• Database migrations complete successfully at the single latest Head.
• The Checkpoint Schema is initialized before the Runtime Worker starts.
• Available Group Planning and Context models are configured in Enterprise Settings.
• Runtime Workers, the Scheduler, and channel services use the same release configuration.

Upgrade Guide

Docker Deployment

git pull origin main

cd deploy
# Rebuild and restart services
docker compose down && docker compose up -d --build

Source Deployment

git pull origin main

cd backend
cd ..

cd frontend
npm install
npm run build
cd ..

./restart.sh

Kubernetes / Helm

helm upgrade clawith helm/clawith/ -f values.yaml

Thank you to all our contributors and community members for your support, feedback, and contributions to this release.

可以压缩成升级说明里的一个小节：

Durable Runtime

新增环境变量：

AGENT_RUNTIME_V2_ENABLED=true

这是 Durable Runtime 的总开关。关闭时：

• Native Web Chat 将不可用（不会回退到旧 Runtime）。
• 即使已完成数据库 Migration 和 Checkpoint 初始化，也不会执行聊天请求。

推荐升级顺序：

1. alembic upgrade head
2. python -m app.scripts.setup_langgraph_checkpoints
3. 设置 AGENT_RUNTIME_V2_ENABLED=true
4. 启动 Backend API 和 Runtime Worker

Docker Compose（单体部署）

environment:
  AGENT_RUNTIME_V2_ENABLED: ${AGENT_RUNTIME_V2_ENABLED:-true}
  AGENT_RUNTIME_V2_AGENT_IDS: ${AGENT_RUNTIME_V2_AGENT_IDS:-}
  AGENT_RUNTIME_V2_SOURCE_TYPES: ${AGENT_RUNTIME_V2_SOURCE_TYPES:-task}

多容器部署时，backend-api 和 backend-worker 必须使用相同的 AGENT_RUNTIME_V2_ENABLED 配置。

v1.10.3 — Agent-to-Agent Messaging Session Consistency

What's New

Optimizations

• Agent-to-Agent Messaging Database Session Handling: Refined database session management for agent-to-agent (A2A) messaging events. This optimization reduces the risk of transactional inconsistencies during message triggers, improving backend reliability and ensuring clean session boundaries for agent interactions.

Bug Fixes

• A2A Messaging Stability: Fixed an issue where database sessions in agent-to-agent messaging could become mismanaged, preventing potential side effects such as lingering database transactions or message delivery failures. This results in more robust and predictable agent messaging.

Upgrade Guide

Docker Deployment

git pull origin main

cd deploy
# Rebuild and restart services
docker compose down && docker compose up -d --build

Source Deployment

git pull origin main

cd backend
cd ..

cd frontend
npm install
npm run build
cd ..

./restart.sh

Kubernetes / Helm

helm upgrade clawith helm/clawith/ -f values.yaml

Notes

• Agent Messaging Reliability: Tenants using agent-to-agent messaging will benefit from improved backend consistency and reduced risk of message-related faults.
• No manual database migration required: Schema migrations run automatically on application startup.

v1.10.2 — Transaction Granularity & Sandbox Stability Enhancements

What's New

Core Features

• Database ContextVar DAO Layer & Transaction Granularity Optimization: Introduced a ContextVar-based DAO abstraction, leading to cleaner, safer transaction handling throughout the application. This reduces risk of cross-request interference and improves backend reliability in concurrent environments.
• Expanded Soul.md Capacity for Agent Context: Increased the character limit for soul.md in agent context-building from 2,000 to 30,000 characters, allowing for richer agent context and more complex behavior modeling.

Optimizations

• Sandbox Process Tree Cleanup on Timeout: Significantly improved bwrap sandbox process cleanup to ensure all subprocesses are reliably terminated after execution timeout. This prevents zombie processes and resource leaks in environments with high code execution activity.
• Tool Call Pairing Integrity in LLM Routing: Enhanced LLM payload handling to ensure tool call pairs are always valid, preventing mismatches during agent tool calling and reducing backend errors.
• Workspace Deletion Permissions Refinement: Clarified and enforced workspace deletion permissions, ensuring only authorized users may delete workspaces or workspace files.

UI/UX Enhancements

• Sidebar Focus List Expansion Option: Users can now view more than 12 items in the sidebar Focus list, with an expand option for easier navigation.
• Chat Timestamp Localization: Chat message timestamps now strictly align with the selected application language, improving global user experience and clarity.

Bug Fixes

• A2A Infinite Loop Prevention: Addressed an issue where agent-to-agent message triggers could recurse infinitely, ensuring stable message routing and preventing backend exhaustion.
• System Email Validation for Invitations: Added pre-validation for system email addresses before sending invites, preventing invalid invitation cycles and reducing bounce rates.
• Agent Settings Permissions Consistency: Fixed permission handling for agent settings, preventing unauthorized edits and ensuring consistent access control.
• Browser Extract Schema Compatibility: Migrated browser extract schema to use Pydantic RootModel[Any], resolving SDK typing issues for proper API serialization/deserialization.
• Workspace File Delete Authorization: Ensured workspace file deletions honor manager permissions, fixing cases where unauthorized deletes could occur.
• History Message Order Correction: Fixed double-reverse logic in chat gateway message ordering, restoring correct transcript sequencing in chat histories.

Upgrade Guide

Docker Deployment

git pull origin main

cd deploy
# Rebuild and restart services
docker compose down && docker compose up -d --build

Source Deployment

git pull origin main

cd backend
cd ..

cd frontend
npm install
npm run build
cd ..

./restart.sh

Kubernetes / Helm

helm upgrade clawith helm/clawith/ -f values.yaml

Notes

• Database Transaction Logic: Custom integrations or plugins interacting with the backend database should review transaction scope logic for compatibility with the new ContextVar-based DAO layer.
• Sandbox Process Cleanup: No configuration changes are needed for sandbox improvements, but heavy code execution tenants may notice improved resource management.
• Soul.md Expansion: Applications or agents using large context files can now take advantage of the raised character limit for richer agent capabilities.
• No manual database migration required: Schema migrations run automatically on application startup.

v1.10.1 — Chat Model Switcher & Entrypoint Permission Optimizations

What's New

Core Features

• Live Chat Model Switching via WebSocket: Enables users to change the active chat model in real time through websockets, improving flexibility and responsiveness in ongoing chat sessions.

Optimizations

• Faster Entrypoint Permissions Check: Refactored and optimized entrypoint permissions verification, providing faster and leaner permission handling during request routing and task dispatch.
• Deployment Config Adjustments: Updated deployment configuration for improved reliability and compatibility with diverse environments.

Bug Fixes

• Chat Model Switcher Stability: Resolved issues related to toggling chat models via websocket, ensuring seamless switching without session drops or inconsistent UI states.
• Entrypoint Permissions Issue: Fixed minor permission validation defects that could block valid requests in specific workflows.
• Config Consistency: Addressed deployment config edge cases related to environment-specific overrides and fallback handling.

Upgrade Guide

Docker Deployment

git pull origin main

cd deploy
# Rebuild and restart services
docker compose down && docker compose up -d --build

Source Deployment

git pull origin main

cd backend
alembic upgrade heads
cd ..

cd frontend
npm install
npm run build
cd ..

./restart.sh

Kubernetes / Helm

helm upgrade clawith helm/clawith/ -f values.yaml

Notes

• Live Model Switching: No special configuration is required for enabling the websocket-based chat model switcher; feature is enabled by default.
• Entrypoint Permissions: Permission check routines have changed under the hood. If you maintain custom permission middleware or gateway logic, audit integration points for compatibility.
• No manual database migration required: Schema migrations run automatically on application startup.

v1.10.0 — Async Agent Messaging, Atlas Onboarding & Robust File/Code Streaming

What's New

Async Agent-to-Agent Communication

• A2A (Agent-to-Agent) async messaging enabled by default: Modernizes inter-agent communication, ensuring agents can message each other asynchronously. Existing tenants are auto-repaired on startup for seamless transition and compatibility.
• Optimized trigger logic and error handling: Improves reliability when invoking agent triggers, handling edge cases more gracefully across communication workflows.

Onboarding Experience — Atlas Design System

• Complete onboarding rewrite using Atlas design system: Revamped 4-screen onboarding with paper/night foundations, cosmographic visuals, personality chips, animated SVG brand marks, and responsive layouts.
• OriginPlate and UniverseMap branding: Login and multi-screen flows now match latest mockups with upgraded illustrations, decorative motifs, and increased accessibility.
• Phase-wise UI enhancements: Phases 1–3 implemented for core onboarding journey, improving engagement and brand cohesion.

Streaming & Workspace File Delivery

• Real-time file delivery injection in A2A chat sessions: Files are now sent directly into agent-to-agent conversations, enhancing collaborative workflows.
• Live code execution streaming: Code output is streamed to the right-side Code panel in real-time, including improved error handling, truncated output on timeout, and user-facing retry hints.
• Chromium PDF sandboxing improvements: Improved Linux compatibility by adding --no-sandbox argument, ensuring stable PDF generation for workspace files.

UI/UX Enhancements

• Atlas login/dialog polish: Login screens unified with refined chrome, cosmography plates, compass motifs, and improved brand mark SVG.
• Multi-select personality chips and dynamic transitions: Boosts agent creation flexibility and onboarding clarity.
• Notification bar stabilization: Top notification now stays fixed, with sticky elements offset below for consistent experience.
• Agent and enterprise settings refactoring: Settings tabs and detail page shells recalibrated for clarity.

Chat & Pagination Improvements

• Cursor-based pagination for chat history: Allows smooth scrolling through long chat sessions, reduces page load times, and supports scalable transcript navigation for end users.

Authentication & Provider Management

• Global Single Sign-On (SSO) custom domain toggle: Administrators can now switch SSO redirect behavior platform-wide, including adaptive UI theming.
• OAuth multi-tenant flow and provider support: Added platform-level OAuth providers for Google and GitHub, improving identity integration for organizations.
• Google Workspace SSO routing hardening: Refined org member links and provider routing to support enterprise teams using Google Workspace.

Workspace & Tool Reliability

• Workspace file deletion restricted to managers: Tightens workspace security by limiting destructive actions to those with management rights.
• S3/GCS endpoint auto-detection and compatibility: Removes ‘SignatureDoesNotMatch’ errors; GCS endpoints now auto-configure for correct V4 signing.
• AgentTool relationship backfill and dynamic loading: Ensures all configured agents have proper tool records; disables tools respected in LLM payloads.

Optimizations

• Reduce DB connection pool exhaustion: Lowers risk of backend overload during LLM calls, ensuring more stable service.
• High-availability (HA) runtime improvements: Backend deployment logic cleaned up for smoother scaling and reliability.
• Dynamic tool log persistence and optimized skill seeding: Tool logs now persisted for channels with faster skill relationship loading, improving auditability and first-run experience.
• Sandbox and workspace fallback logic: Allows local fallback when sandbox environment (bwrap) is unavailable, relaxes subprocess restrictions for broader compatibility.
• Improved release workflow and auto-tagging: Protected branch deployment, auto PR tagging, and smoother release ops.

Bug Fixes

• Workspace file deletion: Only users with manager permissions can delete workspace files, preventing unauthorized data loss.
• DB migration & tool record issues: Alembic migration conflicts resolved, tool backfill now uses commit() for consistency, skips missing AgentTool records, and honor user-disabled tools in LLM call payloads.
• Chat message/file injection errors: Corrected DetachedInstanceError and import paths for chat/file delivery, preventing communication and file transfer failures.
• Live event handling in Agent Detail: Fixed ghost user bubble artifacts caused by agentbay_live events.
• Sandbox streaming & timeout: Proper capturing of code execution stream output on timeout, descriptions now respect config limits (default 60s, max 1h).
• PDF rendering fallback logging: Improved diagnostic messages and error traces for PDF generation under Linux.
• UI/UX Minor Fixes: Numerous adjustments across Atlas screens — logo, ring gaps, cosmography, section labels, indicator lines, and login plate visuals revised for coherence.
• SSO, OAuth, and deployment: Vercel env var type updated to ‘encrypted’, Google Workspace SSO provider routing adjusted, global SSO and reset password theme fixes.
• GCS/S3 signature errors: GCS signature configuration auto-corrects endpoint and resolves API mismatch.
• Markdown rendering and workflow: Improved markdown rendering and refined release workflow triggers.

Upgrade Guide

Docker Deployment

git pull origin main

# Rebuild and restart services
docker compose down && docker compose up -d --build

Source Deployment

git pull origin main

# Rebuild frontend
cd frontend && npm install && npm run build
cd ..

# Restart backend / frontend services

Kubernetes / Helm

helm upgrade clawith helm/clawith/ -f values.yaml

Notes

• Atlas onboarding and agent creation screens: UI/design foundation changed substantially. Custom themes or branding may require review.
• Agent-to-Agent async messaging (A2A) is now standard. Legacy tenant configs are auto-repaired; review downstream automations if you rely on custom agent communication logic.
• OAuth/SSO behavior and domain redirects: New global toggle and improved routing; check your organization’s identity provider setup for compatibility.
• Code execution sandboxes: Timeout is now read from config, max timeout raised to 1h. Ensure configs are up-to-date if you leverage extended runtimes.
• Workspace permissions: Only managers may delete workspace files. Review role assignments to ensure proper access control.
• Release workflow improvements: Protected branch and PR auto-tagging are now supported. Update any internal release scripts if needed.
• GCS/S3 endpoint auto-detection: GCS storage integrations will now self-configure for correct signature version. If you use custom endpoints, verify compatibility.
• No manual database migration required: Schema migrations run automatically on application startup.