* 🗃️ chore: add agent task system database schema
Add 6 new tables for the Agent Task System:
- tasks: core task with tree structure, heartbeat, scheduling
- task_dependencies: inter-task dependency graph (blocks/relates)
- task_documents: MVP workspace document pinning
- task_topics: topic tracking with handoff (jsonb) and review results
- task_comments: user/agent comments with author tracking (text id: cmt_)
- briefs: unresolved notification system (text id: brf_)
All sub-tables include userId FK for row-level user isolation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🗃️ chore: add self-referential FK on tasks.parentTaskId (ON DELETE SET NULL)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🐛 fix: use foreignKey() for self-referential parentTaskId to avoid TS circular inference
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🗃️ chore: add FK on task_topics.topic_id → topics.id (ON DELETE SET NULL)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🐛 fix: resolve pre-existing TS type-check errors
- Fix i18next defaultValue type (string | null → string)
- Fix i18next options type mismatches
- Fix fieldTags.webhook possibly undefined
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🗃️ chore: add FK on tasks.currentTopicId → topics.id (ON DELETE SET NULL)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🗃️ chore: add FK constraints for assignee, author, topic, and parent fields
- tasks.assigneeUserId → users.id (ON DELETE SET NULL)
- tasks.assigneeAgentId → agents.id (ON DELETE SET NULL)
- tasks.parentTaskId → tasks.id (ON DELETE SET NULL)
- tasks.currentTopicId → topics.id (ON DELETE SET NULL)
- task_comments.authorUserId → users.id (ON DELETE SET NULL)
- task_comments.authorAgentId → agents.id (ON DELETE SET NULL)
- task_topics.topicId → topics.id (ON DELETE SET NULL)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🗃️ chore: change task_topics.topicId FK to ON DELETE CASCADE
Topic deleted → task_topic mapping row removed (not just nulled).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ♻️ refactor: use inline .references() for currentTopicId FK
No circular inference issue — only parentTaskId (self-ref) needs foreignKey().
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🗃️ chore: add FK on task_comments.briefId and topicId (ON DELETE SET NULL)
- task_comments.briefId → briefs.id (SET NULL)
- task_comments.topicId → topics.id (SET NULL)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ♻️ refactor: merge briefs table into task.ts to fix circular dependency
brief.ts imported task.ts (briefs.taskId FK) and task.ts imported
brief.ts (taskComments.briefId FK), causing circular dependency error.
Merged briefs into task.ts since briefs are part of the task system.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🗃️ chore: add FK on tasks.createdByAgentId → agents.id (ON DELETE SET NULL)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ♻️ refactor: use incremental diff for snapshot messages to prevent OOM
Replace full messages/messagesAfter duplication per step with baseline + delta approach:
- Step 0 and compression resets store full messagesBaseline
- Other steps store only messagesDelta (new messages added)
- Strip llm_stream events from snapshot (not useful for post-analysis)
- Strip messages from done.finalState (reconstructible from delta chain)
- Strip duplicate toolResults from context.payload
- Reduce context_engine_result event size by removing messages and toolsConfig
- Add reconstructMessages() utility for rebuilding full state from delta chain
- AiAgentService constructor now accepts runtimeOptions for DI
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ♻️ refactor: add incremental toolset delta for snapshot
- Store operationToolSet as toolsetBaseline in step 0 only (immutable)
- Track activatedStepTools changes via per-step activatedStepToolsDelta
- Strip operationToolSet/toolManifestMap/tools/toolSourceMap from done.finalState
- Add reconstructToolsetBaseline() and reconstructActivatedStepTools() utilities
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🐛 fix: correct snapshot delta recording and restore context-engine output
- P1: messagesDelta now always stores only appended messages (afterMessages.slice),
fixing duplication when isBaseline was true (step 0 / compression reset)
- P2: Restore context_engine_result.output (processedMessages) — needed by
inspect CLI for --env, --system-role, and -m commands
- Add P1 regression test for message deduplication
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ✨ feat: inject referenced topic context into last user message
When users @refer_topic in chat, inject the referenced topic's summary
or recent messages directly into the context, reducing unnecessary tool calls.
* 🐛 fix: include agentId and groupId in message retrieval for context engineering
Signed-off-by: Innei <tukon479@gmail.com>
* ✨ feat: skip topic reference resolution for messages with existing topic_reference_context
Added logic to prevent double injection of topic references when messages already contain the topic_reference_context. Updated tests to verify the behavior for both cases: when topic references should be resolved and when they should be skipped.
Signed-off-by: Innei <tukon479@gmail.com>
---------
Signed-off-by: Innei <tukon479@gmail.com>
* 🐛 fix(desktop): remove electron-liquid-glass to fix click event blocking
The electron-liquid-glass native addon was blocking all click events in the
Electron desktop app window. Remove the dependency and restore vibrancy-based
transparency with semi-transparent body background via `.desktop` CSS class.
* 🔨 chore(desktop): remove electron-liquid-glass from native modules config
* style: update claude 4.6 series 1M contextWindow
* chore: cleanup bedrock search tag
chore: cleanup bedrock retired model
chore: cleanup bedrock retired model
chore: cleanup bedrock retired model
* fix: fix ci test
* 💄 style: remove platform-specific Spotlight reference from searchLocalFiles
Replace "using Spotlight (macOS) or native search" with "using native search"
since the actual search implementation is platform-dependent and the LLM
doesn't need to know the specific backend.
Fixes LOBE-5778
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ⚡️ perf: remove duplicate API descriptions from tool system prompt
API identifiers and descriptions are already in the tools schema passed
via the API tools parameter. Repeating them in the system prompt wastes
tokens. Now only tools with systemRole (usage instructions) are injected.
Also rename XML tags: plugins→tools, collection→tool,
collection.instructions→tool.instructions
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 💄 style: inject tool description when no systemRole instead of skipping
Tools without systemRole now show their description as <tool> children.
Tools with systemRole use <tool.instructions> wrapper as before.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 💄 style: always emit <tool> tag, fallback to "no description"
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* update tools
* fix
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ✨ feat: inject all installed skills into <available_skills> for AI discovery
Previously, only skills explicitly added to the agent's plugins list appeared
in <available_skills>. Now all installed skills are exposed so the AI can
discover and activate them via activateSkill.
Changes:
- Frontend: use getAllSkills() instead of getEnabledSkills(plugins)
- Backend: pass skillMetas through createOperation → RuntimeExecutors → serverMessagesEngine
- Add skillsConfig support to serverMessagesEngine
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🐛 fix: use DB + builtin skills for available_skills instead of provider manifests
lobehubSkillManifests are tool provider manifests (per-provider, containing
tool APIs), not skill metadata. Using them for <available_skills> incorrectly
showed provider names (e.g. "Arvin Xu") as skills.
Now fetches actual skills from AgentSkillModel (DB) + builtinSkills for correct
<available_skills> injection.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 💄 style: use XML structure for online-devices in system prompt
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ♻️ refactor: extract online-devices prompt to @lobechat/prompts package
Move device XML prompt generation from builtin-tool-remote-device into
the shared prompts package for reusability and consistency.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ✅ test: add failing tests for Remote Device suppression when auto-activated
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ⚡️ perf: suppress Remote Device tool when device is auto-activated
When a device is auto-activated (single device in IM/Bot or bound device),
the Remote Device management tool (listOnlineDevices, activateDevice) is
unnecessary — saves ~500 tokens of system prompt + 2 tool functions.
- Add autoActivated flag to deviceContext
- Move activeDeviceId computation before tool engine creation
- Disable Remote Device in enableChecker when autoActivated
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* update system role
* update system role
* ♻️ refactor: use agentId instead of slug for OpenAPI responses model field
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🐛 fix: use JSON round-trip instead of structuredClone in InMemoryAgentStateManager
structuredClone fails with DataCloneError when state contains non-cloneable
objects like DOM ErrorEvent (from Neon DB WebSocket errors).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🐛 fix: only inject available_skills when tools are enabled
Restore plugins guard to prevent skills injection when tool use is
disabled (plugins is undefined), fixing 28 test failures.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ✅ test: update system message assertions for skills injection
Use stringContaining instead of exact match for system message content,
since available_skills may now be appended after the date.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
🐛 fix: use jsonb ? operator instead of ->> to avoid Neon rt_fetch bug
The ->> operator in WHERE clauses triggers a Neon-specific
`rt_fetch used out-of-bounds` error. Switch to the ? operator
which is semantically equivalent for checking key existence.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* 🐛 fix(desktop): update bundled agent-browser to v0.20.1 and align native-mode docs
Upgrade desktop bundled agent-browser to 0.20.1 and remove obsolete AGENT_BROWSER_NATIVE runtime override since native mode is now default. Update builtin agent-browser skill descriptions to reflect the new default behavior.
Made-with: Cursor
* ✨ feat: enable agent-browser skill on Windows desktop
Made-with: Cursor
* 🔧 refactor: remove isWindows from ToolAvailabilityContext interface
Updated the ToolAvailabilityContext interface to remove the isWindows property, simplifying the context checks in the isBuiltinSkillAvailableInCurrentEnv function.
Signed-off-by: Innei <tukon479@gmail.com>
---------
Signed-off-by: Innei <tukon479@gmail.com>
* ✨ feat: add subscribeStreamEvents to InMemoryStreamEventManager and use factory for stream route
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* 🐛 fix: remove duplicate agentExecution types and fix stream route test mock
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* 🐛 fix: prevent Enter key submission during IME composition in LoginStep
* ♻️ refactor: extract useIMECompositionEvent hook for IME composition tracking
Made-with: Cursor
---------
Co-authored-by: Innei <tukon479@gmail.com>
* ✨ feat: add slash action tags in chat input
Made-with: Cursor
* ✨ feat: enhance editor with new slash actions and localization updates
- Added new slash actions: change tone, condense, expand, polish, rewrite, summarize, and translate.
- Updated localization files for English and Chinese to include new action tags and slash commands.
- Removed deprecated useSlashItems component and integrated its functionality directly into InputEditor.
Signed-off-by: Innei <tukon479@gmail.com>
* ✨ feat: add slash placement configuration to chat input components
- Introduced `slashPlacement` prop to `ChatInputProvider`, `StoreUpdater`, and `InputEditor` for customizable slash menu positioning.
- Updated initial state to include `slashPlacement` with default value 'top'.
- Adjusted `ChatInput` and `InputArea` components to utilize the new `slashPlacement` prop.
This enhancement allows for better control over the user interface in chat input interactions.
Signed-off-by: Innei <tukon479@gmail.com>
* ✨ feat: implement command bus for slash action tags processing
Add command bus system to parse and execute slash commands (compact context,
new topic). Refactor action tag categories from ai/prompt to command/skill.
Add useEnabledSkills hook for dynamic skill registration.
* feat: compress command
Signed-off-by: Innei <tukon479@gmail.com>
* refactor: compress
Signed-off-by: Innei <tukon479@gmail.com>
* fix: skill inject
* ✨ feat: slash action tags with context engine integration
Made-with: Cursor
* ✨ feat: add topic reference builtin tool and server runtime
Made-with: Cursor
* ✨ feat: add topic mention items and update ReferTopic integration
Made-with: Cursor
* 🐛 fix: preserve editorData through assistant-group edit flow and update RichTextMessage reactively
- EditState now forwards editorData from EditorModal to modifyMessageContent
- modifyMessageContent accepts and passes editorData to updateMessageContent
- RichTextMessage uses useEditor + effect to update document on content change instead of key-based remount
- Refactored RichTextMessage plugins to use shared createChatInputRichPlugins()
* ✨ feat(context-engine): add metadata types and update processors/providers
Made-with: Cursor
* ✨ feat(chat-input): add slash action tags and restore failed input state
* 🔧 chore: update package dependencies and enhance Vite configuration
- Changed @lobehub/ui dependency to a specific package URL.
- Added multiple SPA entry points and layout files to the Vite warmup configuration.
- Removed unused monorepo packages from sharedOptimizeDeps and added various dayjs locales for better localization support.
Signed-off-by: Innei <tukon479@gmail.com>
* 🔧 chore: update @lobehub/ui dependency to version 5.4.0 in package.json
Signed-off-by: Innei <tukon479@gmail.com>
* 🐛 fix: correct SkillsApiName.runSkill to activateSkill and update trimmed content assertions
* 🐛 fix: resolve type errors in context-engine tests and InputEditor slashPlacement
* 🐛 fix: update runSkill to activateSkill in conversationLifecycle test
* 🐛 fix: avoid regex backtracking in placeholder parser
* ✨ feat(localization): add action tags and tooltips for slash commands across multiple languages
Signed-off-by: Innei <tukon479@gmail.com>
* 🐛 fix: preserve file attachments when /newTopic has no text content
* cleanup
Signed-off-by: Innei <tukon479@gmail.com>
---------
Signed-off-by: Innei <tukon479@gmail.com>
* fix command issue
* add run command UI
* fix API key
* add apikey page
* add apikey
* 🐛 fix: update apiKey model tests to use new sk-lh- prefix format
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* fix: add the activatedSkills to improve the execScripte tools
* feat: change the activePath into call market endpoint
* fix: clean the code
* feat: fixed the execScript in desktop ts error
