Getting Started

BI Brain

BI Brain is a project management and collaboration platform that supports Issue dependency graphs, knowledge base AI semantic search, Aiko collaboration, and other features.

---

v3.0.4 (Latest)

Release Date: 2026-08-28 | Status: Official Release Theme: Team-level user portrait + Work item data source neutralization + State machine confirmation flow + Expert Agent enhancement + Cyclomatic complexity remediation Code Range: v3.0.3..v3.0.4 (develop / origin/develop)

1.1 New Features

Module	Details
Team-level User Portrait (New Module)	1. Added user_team_portrait team portrait overlay layer, same user can have differentiated positioning in different teams 2. Supports 8 overridable fields + team-exclusive note field, multi-value fields replaced as whole, explicit emptying does not fall back to global 3. Task evaluation auto-writes back team measured values (team_id snapshot column freezes assessment-time ownership), dispatching uses four-tier degradation by sample size to avoid cross-team data impersonating this team's performance 4. Added team portrait API (list/get/put/delete, clear overlay layer) and frontend TeamPortraitPanel editor, permission distinguishes personal modification from admin operations
Work Item Data Source Neutralization (work-item)	1. Added source-neutral work_item table, identity changed to (source, external_id), no longer requires faking Jira Key, new sources won't collide with unique constraint 2. Parent/child tickets share one table structure, role computed in real-time by parent_external_id, leaf tickets (role=leaf) serve as dispatch entry, parent tickets serve as background 3. Added description and 5 parent ticket snapshot columns, child tickets carry their own background context 4. Added /api/work-items/* series endpoints, apply-plan's jira_data_id/jira_bound renamed to work_item_id/work_item_bound
Work Item Assignee Field and Team Binding	1. work_item adds assignee_name / assignee_email two columns, supports dispatching by assignee (email optional depending on source) 2. Added PUT /api/work-items/data/{id}/team binding endpoint, work items can bind/rebind/unbind Team, validates same workspace and Team not archived, re-sync preserves binding unchanged
Next State Proposal (committed after manual confirmation)	1. Added read-only proposal endpoint POST /api/bibrain/issues/{issue_id}/state-instances/proposal, no database writes throughout 2. Server assembles work item/state chain/execution trail/deliverables/candidate pool, proposal preferentially generated by BiBrain Core read-only mode (readonly enforced by auth, falls back to original path when unavailable) 3. Structured output + Pydantic validation (json_schema → json_object → pure prompt three-tier degradation, validation failure retries once with rejection reason) 4. Frontend state chain card adds "Generate Next State" entry and editable form
Task State Machine Confirmation Flow	1. Introduced pending_confirmation state to carry manual review process, enters pending confirmation after complete, only core service can append states 2. Implemented confirm / request-changes state commands; later converged to complete → completed two-stage, no second confirmation after check completes 3. Redmine tickets go directly from "pending confirmation" to done, removing per-sync online status query
Task Due Date Reminder	Issue adds due date field and reminder service (issue due reminders), expired tasks auto-push notifications
Team-isolated Dynamic Feed	Added Team-isolated department dynamic feed endpoint and team dynamic snapshot migration, dynamics filtered by team ownership for display
Expert Deliverable Ownership	1. Expert execution results auto-backfill deliverable expert ownership (uploaded_by_agent_id) on commit, no dependency on manual upload and no lost ownership 2. workspace_file supports filtering deliverable files by expert agent, and explicitly declares upload ownership (expert_agent_id) 3. Provides one-shot backfill script to handle historical NULL ownership
Generic Expert Memory API	1. Added /expert-memory/* generic memory HTTP API (pages/snapshots/config/search/context), members can browse/search channels 2. /expert-memory is the main path, /expert is legacy alias; new experts auto-bind memory_channel 3. Knowledge file full-text injection (xlsx full-text injected into context), expert run results bridge Issue and mount real reports as deliverables
Task Document Import Schema Validation	1. Added plan schema registry / schema drafts / schema binding and import content audit snapshot 2. Supports schema extraction, draft review and release flow, AI async parsing + draft persistence + evidence validation
Progress Timesheet Dashboard	1. Added progress_dashboard module, multi-dimensional aggregation filtering and pagination query by date/group/status/risk 2. Added /api/progress-dashboard and /api/w/{slug}/progress-dashboard compatible routes
Forbid Duplicate Team Names in Same Workspace	team name adds active-scope unique constraint (case-insensitive, whitespace normalized), create/rename/archive-restore all validate duplicates and return 409
Notify Team Leader After Jira Assignment	Work item auto-notifies team leader after assignment
Jira Batch Import and Version Guard	1. Added batch import endpoint, single member validation + single transaction + savepoint per-row isolation 2. Jira inbound version guard prevents old version silently overwriting new version (updated/stale response) 3. Parent-child relationship changed to use parent_jira_key indexed column derived at read time, import complexity O(N²)→O(N)
Inspection Module Migration	1. Inspection module migrated to BiBrain: git-log API + plan parsing + channel memory 2. GitLab API as backup data source, when repository project_id is missing, auto-resolves by repository name and gives explicit per-repo errors
Expert Agent (New Module)	1. Added complete Expert Agent CRUD (create/edit/delete/list), supporting role prompts and knowledge file binding 2. Added Expert Agent orchestration engine, supporting specifying experts to execute tasks within sessions, managing task lifecycle 3. Frontend adds Expert Agent management page (AdminExpertAgents) and knowledge base management page (AdminKnowledgeBases) 4. Dialog input box adds ExpertPicker selector, supporting selecting Expert Agent within dialog 5. Accompanying Alembic migrations: expert_agent table, knowledge file table, knowledge binding table
Jira Data Enhancement	1. Added Jira data parent-child relationship support (children_ids→parent_jira_key derived at read time), implementing create/move/delete operations with concurrency-safe locking 2. Jira raw data supports metadata extraction and paginated query, supports search by ticket number and type filtering, sorting changed to updated_at 3. Organization space filters deliverable files by expert (uploaded_by_agent_id) 4. Jira data storage format migrated from JSON to NDJSON text format 5. Deletion interface permission adjustment: removed DEBUG mode restriction, permission adjusted from owner/admin to any member
Dynamic Issue Decomposition	1. Requirement analysis interface adds feature_type parameter, supporting four types: epic/bug/qa/custom 2. Dynamically determines Issue count based on feature_type and requirement content, no longer using fixed templates 3. Dispatch plan Markdown format updated, feature_type written into frontmatter, old version compatible with default custom
IssueStateInstance Subtask Content	1. IssueStateInstance model adds title (required) and description (optional) fields 2. State transitions, batch creation, reject/retry all pass subtask title and description info 3. Decomposition service stores plan item title and description into initial state content
Dynamic Instance Node Addition	Supports dynamically adding instance nodes to running workflows
AIKO Session Sharing Facade	1. Added /api/session-shares session sharing API (create/get/download/incoming/outgoing/delete) 2. Capability-based access control — any logged-in user with a valid share_id can read 3. Added InboxType.CONVERSATION_SHARED notification type 4. Owner soft delete immediately invalidates the link
User-Participated Feature Query	1. Extended feature query interface, from querying features led by user to querying features related to user (including historical participation) 2. Added search and filter function by priority and title 3. Implemented user workspace permission control and query scope restriction
Kiosk User Identity	1. Introduced KIOSK_USER_ID constant to identify anonymous monitor terminal identity 2. Core operation permission adjusted from global admin to logged-in user, Kiosk identity is forbidden from accessing core operations 3. Frontend sidebar removes permission control, all logged-in users can access BiBrain Core

1.2 Changes

Module	Details
Work Item Model Migration (Breaking)	Work item storage migrated from Jira online format to source-neutral work-item/v1 envelope; apply-plan's jira_data_id / jira_bound renamed to work_item_id / work_item_bound; old jira_raw_data stops writing but retained (downgrade tier can rollback)
State Machine Convergence (Breaking)	pending_confirmation runtime state removed, complete → completed two-stage; rejection_reason column marked deprecated and no longer written, historical data retained
Skill Repository Migration	SkillRegistry's trusted_root migrated from skills-aiko to skills-core directory; removed bibrain-readonly read-only mode related logic; operator mode loads from skills-core/bibrain-data
aiko_core Default Startup Mode	aiko_core default changed to local startup, no longer requires sandbox environment; updated Dockerfile and deployment configuration
Cyclomatic Complexity Remediation	1. Added local cyclomatic complexity gate (tools/cc_report.py + baseline, only decrease not increase) and golden-master characterization test infrastructure (tests/_golden.py) 2. B1-B4 batches completed 20+ high-complexity function refactoring: bibrain_ask 121→8, build_dashboard 70→7, build_history 55→3, build_brain_context 48→6, build_agent_summary 48→8, etc. 3. Added 200+ golden snapshot cases locking behavior, zero test changes throughout refactoring
Jira Data Interface Documentation	Supplemented Jira interface search/filter function description, added type options, selective deletion, batch import and debug clear interface documentation, added server source code deployment update instructions

1.3 Removed

Module	Details
expert_orchestration Module	Merged into expert_agent module, then deleted the independent expert_orchestration package
bibrain-readonly Skill	Removed bibrain-readonly read-only mode related skill registration logic and tests
Frontend Project Management Page	Removed Web frontend ProjectManagement page
pending_confirmation Runtime State	Removed Issue state machine's pending_confirmation runtime state and confirm / request-changes endpoints and related dead code

1.4 Bug Fixes

Module	Details
Jira Subtask ID	Fixed subtask ID rebuild function, allowing scenarios where subtasks are uploaded before parent tasks
Jira Data Deletion Permission	Removed DEBUG mode restriction, production environment can be used normally; permission adjusted from admin to member level
Transition Recommendation	Fixed issue where transition recommendation required task completion, transition recommendation no longer requires task completion
Work Item Feature Title	Fixed issue where Feature title was stored as JSON original text during work item dispatch (changed to LLM-generated title + fallback cleansing, 13 new test cases)
Expert Agent Creation	Fixed TypeError and 500 on new Expert Agent creation (completed role_prompt / description / instructions fields, restored trusted injection)
Core Stream Error	Core stream errors auto-retry and pass error messages to run when no output (CORE_KERNEL_STREAM_RETRIES, failures with existing output do not retry)
Jira Version Overwrite	Fixed issue where old version Jira data silently overwrote new version (inbound version guard, Last-Write-Wins)
Task Import	Fixed task import date serialization failure, AI parse draft not persisting issue
Database Migration	Fixed alembic dual head causing service startup failure (merged issue due reminders and task parse two heads, converged to single head)
Expert Filtering	Fixed issue where historical NULL ownership files were lost when filtering deliverable files by expert (fallback returns unowned files)
Core Session	Fixed auth issue where expert run failed to upload to public workspace (workspace_file endpoint allows Core service superuser identity)
Performance	get_settings() cache avoids re-parsing .env per request; team AIKO pool polling merged from N requests to 1 (eliminates N+1)
CORS	Allowed Tauri WebView and browser dev origin cross-origin requests
GitLab Address	Inspection repository list and data source config migrated to new GitLab (192.168.20.26)

---

v3.0.3

Release Date: 2026-08-07 Theme: Issue State Machine & Serial Workflow + Team Hierarchy Governance + Core Dynamic Skill + Filespace Named Links Code Range: v3.0.2-stable..9ff4117 (develop / origin/develop) Current Status: Candidate code materials have been updated, but the backend full regression still has P0 blockers; for internal testing only, not deployed to production.

2.1 New Features

Module	Details
Issue State Machine & Serial Workflow (New Module)	1. Added issue_state_chain / issue_state_instance state chain models, using immutable state instance history to carry Issue current state, executor, transition result and audit info 2. Added state command APIs: complete, reject, fail, retry, transfer, cancel, using lock_version to prevent concurrent duplicate transitions 3. Added GET /api/issues/{issue_id}/state-chain and GET /api/agent/active-issue-state-instances, for viewing the complete state chain and the current Aiko's active work items respectively 4. Added POST /api/teams/{team_id}/features:batch-create, creating Feature, strictly serial Issue chain, dependency edges, state chain and initial state within a single transaction 5. Added Feature workflow graph interface and detail page chart, showing serial dependencies, Issue current state, executor and complete state history
Team → Feature → Issue Hierarchy Governance (New Module)	1. Feature adds team_id and feature_type; Issue creation, listing and workflow operations uniformly verify Team, Feature, Workspace ownership consistency 2. Feature/Issue list adds Team scope filter; when manually creating Issue on frontend, Feature options link with Team and filter out completed or cancelled Features 3. Serial graphs under state machine management forbid arbitrary modification of dependencies or key ownership fields, avoiding running workflows being broken
Core Dynamic Skill	1. Core session supports querying available Skills, setting session default Skills, and explicitly overriding optional Skills in a single Run 2. Run creation freezes Skill file list, version and digest; re-verifies file size and hash before execution, preventing silent tampering during runtime 3. Added managed Skill directory and admin API, supporting upload, version update, enable/disable, soft delete and restore, with immutable protection for built-in Skills 4. Frontend adds Core Skill selector and admin Skill management page
Filespace Named Links	1. Workspace and personal filespace support creating named hyperlinks; directory interfaces return folders, files, links respectively, links do not occupy file storage 2. Named links support update, move and soft delete, inheriting parent directory's shared read/write permissions
Migration & Deployment Protection	1. Added historical Alembic branch order fix runner, ensuring config table creation migration precedes related column rename migration 2. Added one-time hierarchy database rebuild protection: only allows rebuilding PostgreSQL public schema when explicit switch, database name secondary confirmation, specified prerequisite revision and non-system database 3. Swarm deployment process adds heads, migrate, verify gates, requiring database revision to exactly match the unique head in the image before business container deployment

2.2 Changes

Module

Details

Breaking Changes

1. Removed issue.status persistent column and old status contract: clients that continue sending status or filtering by old status fields will fail, must switch to Issue State command API, list filtering uses state_type, detail reads state chain / current state 2. Feature/Issue strengthens Team hierarchy: cross-Workspace Team/Feature associations and creation requests missing necessary Team ownership will be rejected, need to clean historical hierarchy data before release, clients select Team first then Feature of the same Team when creating 3. Graphs managed by serial state machine cannot be arbitrarily edited: modifying dependencies, Feature ownership or key graph fields during runtime will return conflict/validation errors, serial chain must be determined in atomic batch-create phase, advanced by state commands after running 4. Core Run freezes Skill manifest: modifying Skill source files does not affect created Runs, Run will fail before injection when files are tampered, Skill changes must publish new version, do not modify referenced versions in place 5. Added 6 Alembic migrations: business code depends on state chain, managed Skill and hierarchy fields, must submit/deploy migrations separately first, execute alembic heads (unique head) and alembic upgrade head on server to confirm revision before deploying business code

2.3 Removed

Module	Details
Issue Status Contract	Removed issue.status persistent column and old status field contract, status now carried by Issue State state chain and state command API

2.4 Bug Fixes

Module	Details
Issue Creation	Fixed inconsistency between Team/Feature selection and request contract when manually creating Issue on frontend
Issue Details	Fixed executor display logic on Issue detail page, prioritizes current state instance's executor with compatible fallback
Database Migration	Fixed upgrade failure caused by historical config migration branch order on new database
Filespace	Fixed Workspace file tree named links and shared permissions collaboration logic
Documentation Consistency	Updated bibrain-service, bibrain-data Skill and aggregated API docs to keep state machine and hierarchy contract consistent

2.5 Verification Results & Release Prerequisites (2026-08-07)

Item	Result	Conclusion
Backend full pytest tests/ -q	1061 passed / 61 failed / 6 skipped (1128 collected)	Not passed
v3.0.3 core incremental set	89 passed	Passed
Frontend npm run test	24 files / 154 tests passed	Passed
Frontend npm run build	Vite build successful (main chunk ~2.95 MB)	Passed
Alembic heads	e57672a20dca, single head	Code graph passed; PostgreSQL real DB pending verification
npm dependency audit	2 moderate / 1 high	Need security assessment and fix

The 61 backend failures are concentrated in Core session, Run events and Skill registration: built-in Skills still reference the deleted skills-aiko/bibrain-data directory, which is a P0 release blocker. The following prerequisites must be completed before official release is recommended:

1. Fix Core built-in Skill root directory/migration logic to ensure no reference to deleted skills-aiko/bibrain-data, and rerun Core targeted tests and backend full tests until 0 failed.
2. pyproject.toml still declares version = "3.0.2", config.py's APP_VERSION is still 0.1.0, must align with v3.0.3 before official tagging.
3. Handle the 2 moderate and 1 high vulnerability reported by npm audit, and rerun frontend tests and build after dependency changes.
4. Re-execute alembic heads on the deployment artifact to confirm unique head; new migrations must be deployed independently before business code.
5. Target PostgreSQL executes alembic upgrade head and verifies revision, then execute state machine, serial activation, Core Skill, named links and permission isolation E2E.

2.6 Version Upgrade Operation Guide

⚠️ This release adds 6 Alembic migrations and removes the issue.status column, must deploy migrations separately first, confirm database revision before deploying business code, do not deploy migrations and business code in the same batch.

1. Sync code and dependencies

   git fetch && git checkout develop && git pull
   uv sync

2. Deploy database migrations first (must be before business code)

   uv run alembic heads          # must be unique head: e57672a20dca
   uv run alembic upgrade head   # verify database revision after upgrade

3. After confirming database revision is correct, deploy business code that depends on the new table structure

4. Frontend build

   cd web && npm install && npm run build

5. Post-release verification: execute state machine, serial activation, Core Skill, named links and permission isolation E2E

---

v3.0.2

Release Date: 2026-07-17 | Iteration Cycle: W5 Iteration Theme: Integration Configuration Center + Alibaba Cloud AI Gateway Consumer Governance + Production Deployment Coverage: BiBrain-Server (FastAPI Backend) + Web Frontend + Production Deployment (Podman Four-Node)

2.1 New Features

Module	Details
Integration Configuration Center (New Module)	1. Unified integration configuration page: added Settings-side integration configuration page and supporting APIs, unifying third-party credentials, personal MCP, and OA accounts under the configuration center 2. Personal MCP credential management: supports manual entry and encrypted storage of third-party credentials such as DingTalk (Contacts/Calendar/Robot), GitLab, Gerrit, Jira, Redmine; sensitive fields stored in user_profile.secrets_enc (encrypted), non-sensitive items stored in integrations (JSONB) 3. Lark OAuth binding: added /api/auth/lark/oauth (start/callback/disconnect) authorization flow, one-click binding of Lark user tokens for use by Lark MCP tools 4. Profile configuration schema interface: GET /api/auth/profile/schema drives frontend form rendering via backend field definitions (string/password/oauth/number/boolean); new backend fields automatically appear on the frontend 5. Configuration resolution internal endpoint: POST /internal/config/resolve allows Aiko client to pull decrypted runtime configuration (credentials/MCP URL) using PAT
OA Login Service Refactoring	1. OA login integrated with configuration center: refactored OA login service, automatically merges OA account information (oa_user_id/oa_password, etc.) into user profile configuration upon login, unified under configuration center 2. OA login interface protocol fix and alignment
Alibaba Cloud AI Gateway Governance (New Module)	1. Consumer governance: /api/gateway adds Alibaba Cloud AI gateway consumer management — CRUD + authorization rules + RBAC gate + API Key authentication 2. ModelApi list and authorization binding: added GET /model-apis and POST /authorizations, service layer orchestration + client layer flattening nested versioned_http_apis + consumer×API Cartesian product atomic batch binding 3. Model API Key binding management 4. Gateway consumer management frontend: added gateway consumer management console and frontend client 5. Authorization gate: bibrain-core authorization consumer verification and RBAC coverage for functional interface calls
Central Intelligence / Assessment / Dashboard	1. Central intelligence Q&A streaming output: BiBrain Core Q&A changed to streaming (SSE) output 2. Token usage statistics: updated ticket type definitions and added token usage statistics; issue summary associates total token count of related tasks 3. Runtime model switching configuration: added /api/admin/llm-config management endpoint; global admin can query/switch models used by central intelligence online 4. Data dashboard polish: weekly tasks, historical completed, average completion time correction, ranking, titles and color scheme, and many other optimizations 5. Health check: added /api/health health check endpoint
GitLab Integration	1. Added GitLab comment forwarding interface 2. Webhook interface compatible with GitLab raw payload format
Deployment	Production Podman four-node deployment: added deploy/production/ — entry node (Nginx/Frontend) + 3 backend replicas + PostgreSQL four-node topology, including per-service compose, Dockerfile, env examples, and deployment guide

2.2 Changes

Module

Details

Breaking Changes

1. Removed Pillow dependency: no longer supports image OCR, chart content recognition, or layout restoration; document parsing retains only text/code/Office/PDF 2. Removed PyMuPDF (mupdf) dependency: PDF parsing switched to pypdf, complex layout restoration capability weakened, need to re-run uv sync 3. OA login refactoring integrated with configuration center: OA account information now carried by configuration center (user_profile.integrations/secrets_enc), requires running this iteration's migration; old clients using device-login/login are not affected 4. Added Alibaba Cloud AI gateway governance: central intelligence functional interface calls need to pass gateway consumer authorization verification, need to configure corresponding consumers and authorization rules for central intelligence

2.3 Removed

Module	Details
Image Parsing Dependency	Removed Pillow dependency, offline image OCR, chart content recognition, and layout restoration capabilities
PDF Parsing Dependency	Removed PyMuPDF (mupdf) dependency, PDF parsing switched to pypdf, weakened high-fidelity layout restoration

2.4 Bug Fixes

Module	Details
Data Dashboard	Fixed bugs in dashboard data statistics
Dependency Security	Fixed high-risk vulnerabilities in software packages (dependency upgrades)
Alibaba Cloud AI Gateway	Fixed gateway model API authorization request
OA Login	Fixed OA login service interface protocol
Value Assessment	Retained assessment metric deliverable field alias (deliverables alias)

2.5 Module Development Progress Snapshot (W5)

Module	W4 Previous Progress	W5 Current Progress
RBAC Permission System	85%	88%
Value Assessment System	80%	82%
BiBrain Core Central Intelligence	60%	70% (Q&A streaming + gateway authorization)
Monitoring & Dashboard / Data Dashboard	70%	85%
Integration Configuration Center	-	80% (New Module)
Alibaba Cloud AI Gateway Governance	-	75% (New Module)
IM Integration (Lark / DingTalk)	-	55% (New Module, Lark OAuth + DingTalk MCP credentials)
Production Deployment (Podman Four-Node)	-	80% (New Module)

2.6 Version Upgrade Operation Guide

1. Pull code

   git fetch && git checkout develop && git pull

2. Install dependencies (remove Pillow / PyMuPDF, gateway-related dependencies)

   uv sync            # or pip install -e .

3. Database migration

   alembic upgrade head

4. Configure integration center and gateway (optional)

   • Lark binding requires configuring Lark app App ID / App Secret and callback URL in .env
   • Alibaba Cloud AI gateway consumer governance requires configuring gateway instance access credentials
   • Central intelligence runtime model switching see /api/admin/llm-config

5. Frontend build

   cd web && npm install && npm run build

6. Production deployment (optional, Podman four-node): refer to deploy/production/README.md

---

v3.0.1

Release Date: 2026-07-02 | Iteration Cycle: W4 Iteration Theme: Unified Permission System + Value Assessment Closed Loop + Central Intelligence Implementation Coverage: BiBrain-Server (FastAPI Backend) + Web Frontend + Aiko/Central Intelligence Skill (bibrain-service / bibrain-data)

2.1 New Features

Module	Details
Permissions & Security (New RBAC Permission System)	1. Introduced Casbin domain-based RBAC unified authorization layer, replacing scattered hand-written auth, unified permission points and role matrix 2. Annotation-based functional permission auto-registration, supports functional group batch authorization 3. Workspace ownership-based data gate, unified resource ownership verification 4. Built-in role convergence: workspace owner equals admin, deprecated per-ws g2 role projection; role management only operable by global admin 5. Refactored admin console dual-mode shell, role management page adds role deletion entry 6. Added "current user permission query" and "user directory search" read-only interfaces
Value Assessment System (New Assessment Closed Loop Module)	1. Five-dimensional task quality assessment system (delivery/quality/efficiency/collaboration/autonomy) + digital employee comprehensive scoring 2. Automatic assessment closed loop: tasks automatically scored upon completion and written back to user profile, forming execution-assessment-profile update chain 3. Frontend page displays task value assessment results
BiBrain Core Central Intelligence (Super Digital Employee)	1. First version of central-side DeepAgents Agent, integrates bibrain-service/data skills, standard backend API call chain 2. Web-side dedicated multi-turn dialogue interaction page 3. Added 2 global information query interfaces to support intelligent Q&A
Aiko Autonomous Evolution	Aiko evolution outputs stored in isolated skill folders by agent_id, with storage interfaces and frontend management page
Monitoring & Dashboard V2 (New Visualization Module)	1. Multi-dimensional data analysis panels: workload, progress, token consumption, agent running status monitoring 2. New global Dashboard, adds AI dialogue card, user profile card, displays position, skills, workload, availability data 3. Configurable database connection pool monitoring metrics
Wiki Knowledge Base (formerly knowledge refactored)	1. File-based Wiki architecture: Issue attachments stored in single copy and automatically included in knowledge base 2. Issue closure automatically triggers Wiki document construction
Workspace & Collaboration Enhancement	1. Added admin workspace management full interface set and frontend pages 2. Supports on-demand one-click launch of multi-person collaboration group chat 3. Added designated Issue execution process and deliverable query interfaces 4. Platform-wide file upload record retention function
Agent / Aiko Collaboration (Existing Capability Iteration)	1. High-level aggregated interface bibrain_api: provides scenario-oriented encapsulated interfaces for Aiko, reducing multiple rounds of atomic API calls 2. Agent visualization monitoring: new Agent monitoring page, supports canvas drag-and-drop, Issue hover display, real-time log display 3. Aiko completion/deliverable submission process: bibrain-http skill adds deliverable processing workflow and completion summary comment template, standardizes Aiko task closure 4. pm-assistant one-time dependency setup: task creation process supports one-time dependency configuration, simplifies Aiko orchestration 5. Deprecated old aiko-code-fixer skill, added bibrain-data skill, responsible for data statistics, task decomposition and dispatch
Issue Collaboration (Existing Capability Iteration)	1. Issue dependency graph: new issue dependency graph function and dependency visualization 2. Task date and planning attributes: Issue supports date and planning attributes 3. Batch operations and status extensions: Issue supports batch deletion, dimension distribution constraints, pending confirmation status 4. Agent assignment display optimization: enhanced visualization of Agent assignment information in Issue list 5. Issue comment/attachment complete file upload chain
Infrastructure	1. Database standard PostgreSQL, with complete Alembic migration scripts, this release adds casbin_rule permission table 2. WebSocket event broadcast after transaction commit, ensures multi-end data consistency 3. Improved database connection pool Metrics monitoring configuration 4. Added multiple design documents, validation report archiving specifications
Concept Unification	Business layer unifies project to feature, full-stack alignment with R&D semantics, frontend/backend/Skill synchronously adapt to modeling conventions
User Authentication Basics	OA login, JWT authentication, PAT tokens, OAuth integration
Frontend Web App	Full upgrade: permission management console, central intelligence dialogue page, new monitoring Dashboard, Wiki knowledge base, workspace management, role configuration and other pages fully covered
Redmine / GitLab Integration	User mapping, project/Issue synchronization, Commit query

2.2 Changes

Module

Details

Breaking Changes

1. Casbin is a mandatory hard dependency, RBAC enabled by default; unauthorized interfaces uniformly return 403, need to execute database migration + initialize admin script 2. Role system refactoring: deprecated per-ws g2 roles, workspace owner equals admin, old multi-role projection logic invalidated 3. Knowledge base module renamed: route changed from /api/knowledge to /api/wiki, all original knowledge-related interfaces migrated 4. Added runtime dependencies deepagents/langchain-openai/tavily-python/aiomysql/pymupdf, deployment needs to re-execute uv sync/pip install dependencies 5. Aiko skill package adjusted, aiko-code-fixer offline, must upgrade bibrain-service, bibrain-data skill packages 6. Database switched from SQLite to PostgreSQL: deployment method changed; connection strings and migration scripts are incompatible 7. project → feature concept renaming: affects APIs, data models, frontend routes 8. Aiko completion process removes assignee modification: old logic that automatically changed assignee when Aiko completes is deprecated 9. Issue tasks can only be dispatched by "dimension": directly dispatching non-dimensional tasks to Aiko will be rejected

2.3 Removed

Module	Details
Aiko Skills	Offline old aiko-code-fixer skill package
Permission Logic	Deprecated each route's hand-written independent auth logic, deprecated per-ws g2 role projection mechanism
Knowledge Base	Deprecated old knowledge module architecture, unified migration to wiki system

2.4 Bug Fixes

Module	Details
Value Assessment Module	1. Fixed Issue assessment cannot auto-trigger issue 2. Fixed Agent and user identity confusion bug in assessment process
Knowledge Base Wiki	Fixed file construction status display abnormal issue
Message Notification	Fixed no notification push after task dispatch issue
Aiko Execution Push	Fixed Aiko real-time push execution process logic
Permission Control	Fixed ws-admin interface permission privilege escalation vulnerability, added resource ownership verification gate
BiBrain Core	Optimized core module global error capture handling mechanism
Display Issues	Fixed floating Issue display issues
Redmine Integration	Fixed Redmine creation synchronization issues
Database	Fixed database-related bugs, added connection pool monitoring
Log Display	Changed display logs to use real data, removed mocks
Routing	Fixed route refresh and matching order issues
Date & Time	Unified UTC timezone conversion for date and time
Project Allocator	Fixed project allocator disabled state judgment
Cache Synchronization	Optimized cache synchronization after delete operations

2.5 Module Development Progress Snapshot (W4)

Module	W3 Previous Progress	W4 Current Progress
Agent (Aiko) Task Scheduling	85%	90%
Real-time Communication & Message Push	80%	85%
Knowledge Base / Wiki	65%	75%
Frontend Web App	60%	75%
RBAC Permission System	-	85% (New Module)
Value Assessment System	-	80% (New Module)
BiBrain Core Central Intelligence	-	60% (New Module)
Aiko Autonomous Evolution	-	55% (New Module)
Monitoring & Dashboard v2	-	70% (New Module)

2.6 Version Upgrade Operation Guide

1. Sync latest code

git fetch && git checkout develop && git pull

---

v3.0.0

Release Date: 2026-05-29 | Base Version: -

2.1 New Features

Module	Details
Agent / Aiko Collaboration	1. High-level aggregated interface bibrain_api: provides scenario-oriented encapsulated interfaces for Aiko, reducing multiple rounds of atomic API calls 2. Agent visualization monitoring: new Agent monitoring page, supports canvas drag-and-drop, Issue hover display, real-time log display 3. Aiko completion/deliverable submission process: bibrain-http skill adds new deliverable processing workflow and completion summary comment template, removes assignee modification logic, standardizes Aiko task closure 4. pm-assistant one-time dependency setup: task creation process supports one-time dependency configuration, simplifies Aiko orchestration
Issue Collaboration	1. Issue dependency graph: new issue dependency graph function and dependency visualization 2. Task date and planning attributes: Issue supports date and planning attributes 3. Batch operations and status extensions: Issue supports batch deletion, dimension distribution constraints, pending confirmation status 4. Agent assignment display optimization: enhanced visualization of Agent assignment information in Issue list 5. File upload support: Issue comment/attachment file upload channel connected
Knowledge Base	1. New knowledge and agent_file data models 2. Knowledge base supports AI semantic search + large model answering
Infrastructure	1. Database switched to PostgreSQL: migrated from SQLite to PostgreSQL, solves multi-Aiko concurrent heartbeat and message flow write lock bottlenecks 2. Included migration scripts, deployment scripts, and data import scripts 3. Post-transaction broadcast: WebSocket events broadcast after transaction commit, ensures multi-end consistency 4. Personnel list import script: supports batch import from management platform 5. Configuration system improvement: database connection configuration and environment variable updates
Concept Unification	1. Project → Feature concept refactoring: business layer unifies project to feature, aligns with R&D semantics 2. bibrain-http skill synchronously updates Feature/Issue modeling conventions and status field specifications
User Authentication & Permissions	OA login, JWT authentication, PAT tokens, OAuth integration
Workspace & Project Management	Workspace CRUD, member invitation, Project progress tracking
Server Skeleton	FastAPI + SQLAlchemy async + Alembic + SQLite WAL
Frontend Web App	Login, Issue list/details, project management, knowledge base, GitLab, Inbox
Redmine / GitLab Integration	User mapping, project/Issue synchronization, Commit query

2.2 Changes

Module

Details

Breaking Changes

1. Database switched from SQLite to PostgreSQL: deployment method changed; connection strings and migration scripts are incompatible 2. project → feature concept renaming: affects APIs, data models, and frontend routes 3. Aiko completion process removes assignee modification: old logic that automatically changed assignee when Aiko completes is deprecated 4. Issue tasks can only be dispatched by "dimension": directly dispatching non-dimensional tasks to Aiko will be rejected

2.3 Removed

Module	Details
None	None

2.4 Bug Fixes

Module	Details
Display Issues	Fixed floating Issue display issues
Redmine Integration	Fixed Redmine creation synchronization issues
Database	Fixed database-related bugs
Log Display	Changed display logs to use real data, removed mocks
Routing	Fixed route refresh and matching order issues
Date & Time	Unified UTC timezone conversion for date and time
Project Allocator	Fixed project allocator disabled state judgment
Cache Synchronization	Optimized cache synchronization after delete operations