Extract provider resolution logic into two new modules:
- provider-catalog.ts: vendor presets, default model catalogs, protocol definitions
- provider-resolver.ts: unified resolution for all consumers (Claude SDK, AI SDK, model listing)
DB migration:
- Add protocol, headers_json, env_overrides_json, role_models_json columns to api_providers
- Create provider_models table for per-provider model catalog
Consumer refactoring:
- claude-client.ts: replace inline env-building with toClaudeCodeEnv()
- text-generator.ts: replace inline config with toAiSdkConfig()
- conversation-engine.ts, onboarding, checkin, plan routes: use shared resolveProvider()
- /api/providers/models: read from catalog + provider_models table
UI improvements:
- Merge confusing "Custom API (Anthropic-compatible)" into "Anthropic Third-party API"
- Add 3-field model name mapping (sonnet/opus/haiku) in advanced options
- ProviderForm.tsx: support new fields (protocol, env_overrides, role_models)
Bug fixes:
- Protect auth credentials from being deleted by legacy extra_env empty-string overrides
- Fix env var names: ANTHROPIC_DEFAULT_SONNET → ANTHROPIC_DEFAULT_SONNET_MODEL (SDK convention)
Tests: 229/229 passed. Verified with aiberm (no mapping) and pipellm (with mapping).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Composer restructure:
- MessageInput internal footer now only contains: file button, model
selector, slash command button, and send button
- Removed mode toggle (Code/Plan) from composer area (logic preserved)
- New ChatComposerActionBar below input with three slots: left (Design
Agent), center (permission selector), right (context indicator)
New components:
- SlashCommandButton: inserts `/` at cursor, reuses existing slash
command detection and popover logic
- ChatPermissionSelector: session-level permission toggle (default /
full_access) with danger confirmation dialog
- ContextUsageIndicator: SVG donut chart showing context usage ratio
with HoverCard displaying model, tokens, cache stats
- ChatComposerActionBar: slot-based container for action bar items
Session permission (permission_profile):
- DB: new column on chat_sessions, default 'default'
- API: PATCH sessions supports updating permission_profile
- Chat API: bypassPermissions = global setting OR session full_access
- PermissionPrompt: auto-approves and hides when full_access
- Claude client: combines global and per-session bypass flags
Bridge integration:
- permission-broker: auto-approves permissions for full_access sessions
without sending IM notifications
- conversation-engine: passes bypassPermissions to SDK for full_access
- PATCH route: auto-approves pending bridge permissions on switch to
full_access
Other:
- ImageGenToggle relabeled to "Design Agent" / "设计 Agent"
- New model-context.ts with MODEL_CONTEXT_WINDOWS lookup
- Provider models API returns optional contextWindow field
- New useContextUsage hook for token usage calculation
- i18n: added composer.*, permission.*, context.* keys (en + zh)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add "Assistant Workspace" feature that lets users designate a directory
(e.g. Obsidian vault) containing 4 Markdown files (soul.md, user.md,
memory.md, claude.md) to define AI personality, user profile, long-term
memory, and execution rules. These files are automatically injected into
the system prompt for assistant project sessions.
Core library (src/lib/assistant-workspace.ts):
- Workspace validation, initialization with template files
- File loading with 8000-char truncation, XML prompt assembly (40k limit)
- State management via .assistant/state.json
- Directory docs generation (README.ai.md + PATH.ai.md)
- Case-insensitive file name fallback (soul.md → SOUL.md)
System prompt injection (src/app/api/chat/route.ts):
- Scoped to sessions where working_directory === workspace path
- Workspace prompt prepended before session-level system prompt
- Onboarding/check-in instructions injected as <assistant-project-task>
- autoTrigger flag skips user message save and title update
Chat-based onboarding (10 questions) & daily check-in (3 questions):
- AI conducts conversation, outputs structured completion code blocks
- Frontend detects onboarding-complete/checkin-complete blocks
- Backend calls Claude API to generate file content from answers
- Fallback to raw answers if API call fails
Auto-trigger pipeline:
- Mount trigger (500ms delay) + focus fallback in MessageInput
- assistantTriggerFiredRef prevents duplicate triggers per mount
- hookTriggeredSessionId in state.json prevents cross-reload duplicates
- Daily check-in only triggers in the most recent workspace session
- hookTriggeredSessionId cleared via __clear__ sentinel after completion
API endpoints:
- GET/PUT /api/settings/workspace — path CRUD + file status
- POST /api/workspace/session — create (onboarding) or reuse (check-in)
- POST /api/workspace/onboarding — AI-generate soul.md + user.md
- POST /api/workspace/checkin — AI-update memory.md + user.md
- POST /api/workspace/hook-triggered — set/clear trigger guard
- GET /api/workspace/latest-session — latest session lookup
- POST /api/workspace/docs — refresh directory documentation
UI:
- Settings "Assistant" tab with workspace path, file status, controls
- Violet badge on workspace project group in ChatListPanel
- useAssistantWorkspace hook for state queries
Also includes:
- 102 passing tests (16 new regression tests for workspace logic)
- i18n keys for en.ts and zh.ts
- Handover documentation and execution plan
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
