v1.6.1

What's New

Features

• Smart Dispatcher - Automatic agent type selection via LLM (#20)
  • Uses Claude to analyze task descriptions and select the best agent
  • Configurable confidence threshold with fallback
  • In-memory caching with TTL for performance
  • CLI command: aistack agent auto "your task description"
  • MCP and REST API integration (agentType now optional)

Improvements

• Test coverage for Smart Dispatcher at 100% statements/functions
• Discord community badge added to README

Bug Fixes

• Fixed flaky health monitor test (timing precision in CI)

Community

• Join our Discord: https://discord.gg/uQ6fDXDs7E

---

Full Changelog: v1.5.4...v1.6.1

v1.5.4 - Documentation Sync & CLI Fixes

What's Changed

Documentation

• Complete rewrite of docs/README.md from stale v1.0.0 to current v1.5.3
• Added Slack integration documentation and configuration example
• Added missing requireConfirmationOnIntervention config property
• Fixed MCP tool count across all docs (41 → 46)
• Updated audit report with adversarial verification findings

Bug Fixes

• Fixed init command package name (npx aistack → npx @blackms/aistack)
• Fixed CLI mcp tools command to show all 46 tools (was only showing 30)

Internal

• Full adversarial documentation verification ensuring code/docs alignment

Full Changelog: v1.5.3...v1.5.4

v1.5.2 - Security Fix: Memory Isolation

🔒 Security Fix

This release fixes a critical security vulnerability (Issue #7) where agents in different sessions could access each other's memory.

Changes

Memory Access Control

• Added session-based memory isolation layer (src/memory/access-control.ts)
• All memory operations now enforce session boundaries
• Cross-session access attempts are blocked and logged

Session Cleanup

• Session end now automatically cleans up session memory
• All agents in a session are stopped when the session ends

API Changes

• MCP Tools: All memory tools (memory_store, memory_search, memory_get, memory_list, memory_delete) now require sessionId parameter
• REST Endpoints: POST and DELETE operations on /api/v1/memory now require sessionId

Test Coverage

• 34+ new tests for access control and session isolation
• 100% coverage on modified core files

Migration Notes

If you're using MCP memory tools, you'll need to update your calls to include sessionId:

// Before
await memory_store({ key: 'my-key', content: 'data' });

// After
await memory_store({ sessionId: 'your-session-id', key: 'my-key', content: 'data' });

Full Changelog: v1.5.1...v1.5.2

v1.5.0 - Resource Exhaustion Monitoring

What's New

🛡️ Resource Exhaustion Monitoring

Detect and prevent runaway agents consuming excessive resources without producing meaningful deliverables.

• Per-Agent Tracking: Track files accessed, API calls, subtasks spawned, tokens consumed
• Phase Progression: normal → warning → intervention → termination
• Configurable Thresholds: Set limits for each resource type
• Pause/Resume Control: Automatically pause agents exceeding thresholds
• Deliverable Checkpoints: Reset time-based tracking when agents produce results
• Slack Notifications: Alert on warnings, interventions, and terminations
• Prometheus Metrics: Full observability with counters, gauges, histograms

🎯 Semantic Drift Detection

Detect when task descriptions are semantically similar to ancestors (from v1.5.0-beta).

• Embedding-based Similarity: Uses OpenAI or Ollama embeddings
• Configurable Thresholds: threshold (block/warn) and warningThreshold (warn only)
• Task Relationships: Track parent_of, derived_from, depends_on, supersedes

Configuration

{
  "resourceExhaustion": {
    "enabled": true,
    "thresholds": {
      "maxFilesAccessed": 50,
      "maxApiCalls": 100,
      "maxSubtasksSpawned": 20,
      "maxTimeWithoutDeliverableMs": 1800000,
      "maxTokensConsumed": 500000
    },
    "warningThresholdPercent": 0.7,
    "autoTerminate": false,
    "pauseOnIntervention": true
  }
}

New REST API Endpoints

• GET /api/v1/agents/:id/resources - Get agent resource metrics
• POST /api/v1/agents/:id/deliverable - Record deliverable checkpoint
• POST /api/v1/agents/:id/pause - Pause an agent
• POST /api/v1/agents/:id/resume - Resume a paused agent
• GET /api/v1/system/resources - Resource exhaustion summary

Stats

• 41 MCP tools (added 8 Identity tools, 3 Task drift tools)
• 11 specialized agents
• 6 LLM providers
• 70 new unit tests for resource exhaustion (100% line coverage)

Full Changelog

v1.4.0...v1.5.0

v1.4.0 - Agent Identity v1

Agent Identity v1

This release introduces Agent Identity as a first-class concept, enabling agents to exist as persistent entities with stable identifiers, lifecycle management, and owned memory across executions.

Key Features

Persistent Identity Model

• Stable agent_id (UUID) that persists across executions
• Lifecycle states: created → active → dormant → retired (no hard deletes)
• Agent capabilities and metadata storage
• Full audit trail for all identity lifecycle events

Agent-Scoped Memory

• Memory namespaces owned by agents with agentId parameter
• includeShared option to access shared memory
• Memory isolation between agents

REST API

New endpoints at /api/v1/identities:

• Create, list, get, update identities
• Lifecycle operations: activate, deactivate, retire
• Audit trail retrieval

MCP Tools

8 new tools for identity management:

• identity_create, identity_get, identity_list, identity_update
• identity_activate, identity_deactivate, identity_retire, identity_audit

Spawner Integration

• identityId option to spawn with existing identity
• createIdentity option for ephemeral identities
• Stopping an agent transitions its identity to dormant

Example Usage

# Create identity via REST
curl -X POST /api/v1/identities \
  -d '{"agentType":"coder","displayName":"main-coder"}'

# Spawn agent with identity
curl -X POST /api/v1/agents \
  -d '{"type":"coder","identityId":"<uuid>"}'

# Store agent-scoped memory
aistack memory store -k learning -c "I learned X" --agent-id <uuid>

Test Coverage

• 7 new test files with 90%+ coverage on identity code
• All 1728 tests passing

Full Changelog: v1.3.1...v1.4.0

v1.3.1 - CI/CD Performance Optimization

🚀 Performance Improvements

This release focuses on CI/CD pipeline optimization, delivering 70% faster build times.

CI/CD Optimizations

• 70% faster CI runs - From ~8 minutes to ~2.5 minutes
• Removed Node 20 matrix duplication - Now testing only on Node 22
• Parallel job execution - Type check, lint, test, and build run concurrently
• Test parallelization - Unit tests now use Vitest thread pool for parallel execution
• Build caching - TypeScript compilation cached between runs
• Playwright optimization - Increased CI workers from 1 to 2

Bug Fixes

• Fixed SQLite I/O conflicts by separating unit and integration test execution
• Skipped 1 config test incompatible with worker threads (process.chdir limitation)

Test Results

• ✅ 1488 tests passed, 1 skipped
• ✅ All CI jobs passing
• ✅ Coverage reporting working correctly

Breaking Changes

None - This is a patch release with CI/CD improvements only.

---

Full Changelog: v1.3.0...v1.3.1

v1.3.0

What's Changed

• feat: add adversarial agent with iterative review loop by @blackms in #1

New Contributors

• @blackms made their first contribution in #1

Full Changelog: v1.2.0...v1.3.0

v1.2.0 - IDE-like Task Workflow UI

What's New

IDE-like Task Workflow UI

This release introduces a comprehensive project workflow system for managing development tasks through a visual interface.

Features

• Project Management: Create and manage projects linked to filesystem folders
• Task Workflow: Phase-based task management with Kanban board visualization
  • Draft → Specification → Review → Development → Completed
• Specification Documents: Create, edit, and manage specification documents
  • Support for architecture, requirements, design, and API specs
  • Approval/rejection workflow with comments
• Folder Picker: Browse and select project directories through the UI
• Real-time Updates: WebSocket events for live status updates

Backend

• New database tables: projects, project_tasks, specifications
• New API endpoints:
  • /api/v1/projects - Project CRUD operations
  • /api/v1/projects/:id/tasks - Task management with phase transitions
  • /api/v1/tasks/:taskId/specs - Specification management
  • /api/v1/filesystem - Directory browsing and validation

Frontend

• ProjectsPage - Project listing with grid layout
• ProjectDetailPage - Kanban board for task phases
• TaskDetailPage - Specification management with viewer/editor
• New Zustand stores for state management

Testing

• 26 unit tests for database operations
• 52 integration tests for new API routes
• Overall test coverage: 97.63%

Full Changelog: v1.1.0...v1.2.0

v1.1.0 - Web Interface

🌐 Web Interface for AgentStack

This release introduces a complete web-based dashboard for monitoring and managing AgentStack.

Features

Backend API (src/web/)

• REST API with custom router (no Express dependency)
• WebSocket server for real-time updates
• Routes for agents, memory, tasks, sessions, workflows, system
• CORS and error handling middleware
• CLI command: aistack web start

Frontend Dashboard (web/)

• React + TypeScript + Material UI
• Vite dev server with API proxy
• Dashboard with system metrics
• Agent management (spawn, execute, stop)
• Memory explorer with search
• Task queue management
• Workflow launcher with progress tracking
• Real-time chat interface with agents
• Dark theme UI

Usage

Start the web server:

# Start backend API
aistack web start

# Or using npm
npm run web

# Start frontend dev server (in separate terminal)
npm run dev:web

Access the dashboard at http://localhost:5174

Test Coverage

Added comprehensive tests for the web module achieving 94% coverage.

v1.0.9

Full Changelog: v1.0.8...v1.0.9