39 KiB
39 KiB
Waggle OS -- Comprehensive Feature Inventory
Generated from codebase audit, April 2026. Status key: BUILT = fully implemented and wired, PARTIAL = backend exists but UI incomplete or gaps remain, PLACEHOLDER = route/type exists but minimal logic.
1. PERSISTENT MEMORY SYSTEM (Crown Jewel #1)
Waggle's memory system is a multi-layered, dual-mind architecture that gives the AI agent persistent, cross-session knowledge. Nothing comparable exists in competing products at this depth.
1.1 Dual-Mind Architecture
| Feature | Description | Status |
|---|---|---|
| Personal Mind | SQLite DB (~/.waggle/.mind/) stores user-level knowledge shared across all workspaces |
BUILT |
| Workspace Mind | Separate SQLite DB per workspace, fully isolated -- client data never leaks between projects | BUILT |
| MultiMind Router | Automatically routes memory queries/saves to the correct mind based on context | BUILT |
| Cross-Mind Dedup | Before saving, checks BOTH personal and workspace minds for exact, normalized, and semantic (cosine > 0.95) duplicates | BUILT |
| Workspace-Specific Routing Guard | If target=personal but content contains confidential signals ($amounts, client names, NDA, SOW), auto-redirects to workspace mind |
BUILT |
| Cross-Workspace Search | search_all_workspaces tool queries ALL workspace minds + personal simultaneously |
BUILT |
1.2 FrameStore (Memory Frames)
| Feature | Description | Status |
|---|---|---|
| I-Frames | Initial frames -- standalone memory units | BUILT |
| P-Frames | Predictive/update frames -- linked to a base I-frame, representing changes or additions | BUILT |
| B-Frames | Bridge frames -- cross-reference multiple other frames with structured references | BUILT |
| Importance Levels | critical, important, normal, temporary, deprecated -- with score multipliers (2.0x to 0.3x) |
BUILT |
| Source Provenance | user_stated, tool_verified, agent_inferred, import, system -- tracked per frame |
BUILT |
| Confidence Derivation | Auto-derives confidence from source: tool_verified->high, user_stated->medium, agent_inferred->low | BUILT |
| Dramatic Claim Detection | Detects dramatic statements (shutdown, bankrupt, emergency) and downgrades importance to temporary |
BUILT |
| Rate Limiting | Max 50 memory saves per session to prevent flooding attacks | BUILT |
| FTS5 Indexing | Full-text search index on all frame content | BUILT |
| Dedup on Create | I-Frame creation checks for exact duplicates before inserting | BUILT |
1.3 HybridSearch (Vector + Keyword Fusion)
| Feature | Description | Status |
|---|---|---|
| Keyword Search | SQLite FTS5 full-text search | BUILT |
| Vector Search | Embedding-based similarity via sqlite-vec | BUILT |
| Reciprocal Rank Fusion (RRF) | Combines keyword and vector results using RRF (k=60) | BUILT |
| Scoring Profiles | balanced, recent, important, connected -- weighted scoring across recency, importance, access, and connections |
BUILT |
| Temporal Filtering | since and until ISO date filters on search |
BUILT |
| Importance Filtering | Filter results by importance level (critical, important, normal, temporary) | BUILT |
| Contradiction Detection | Flags potential contradictions when results contain conflict signals ("however", "no longer", "changed to") | BUILT |
| LIKE Fallback | When vector/FTS return nothing, falls back to SQL LIKE scan on keywords | BUILT |
1.4 Knowledge Graph
| Feature | Description | Status |
|---|---|---|
| Entity Extraction | Regex-based extraction of persons, organizations, technologies, projects, concepts from text | BUILT |
| Entity Normalization | Normalizes entity names for consistent graph storage | BUILT |
| Relation Extraction | Semantic relations: led_by, reports_to, depends_on, co-occurrence | BUILT |
| Wildcard Query | query_knowledge("*") lists all known entities (capped at 50) |
BUILT |
| Entity Correction | correct_knowledge tool to update or invalidate entities via feedback handler |
BUILT |
| Ontology | Structured ontology definitions for entity types | BUILT |
| Graph Visualization API | GET /api/memory/graph returns entities + relations for UI rendering |
BUILT |
1.5 CognifyPipeline
| Feature | Description | Status |
|---|---|---|
| Full Pipeline | Save frame -> extract entities -> enrich knowledge graph -> create relations -> index for search | BUILT |
| Memory Linking | Optional MemoryLinker finds related frames after saving | BUILT |
| Co-occurrence Relations | Entities found in same text automatically get relation edges | BUILT |
1.6 IdentityLayer
| Feature | Description | Status |
|---|---|---|
| Agent Identity | Persists name, role, department, personality, capabilities, system_prompt | BUILT |
| Identity Context | get_identity tool returns who the agent is |
BUILT |
| Auto-Setup | ensureIdentity runs on server startup |
BUILT |
1.7 AwarenessLayer
| Feature | Description | Status |
|---|---|---|
| Active Task Tracking | Categories: task, action, pending, flag | BUILT |
| Priority System | 0-10 priority per awareness item | BUILT |
| Expiration | Optional expires_at for time-limited items | BUILT |
| Metadata | JSON metadata per item (context, status, result) | BUILT |
| Tool Utilization Tracking | Tracks unique tools used per session, reports utilization percentage | BUILT |
1.8 SessionStore
| Feature | Description | Status |
|---|---|---|
| GOP-based Sessions | Sessions identified by gop_id, with timestamps | BUILT |
| Session Distillation | findUndistilledSessions + markSessionDistilled for memory consolidation |
BUILT |
| Session Search | Full-text search across session transcripts | BUILT |
| Session Timeline | parseSessionTimeline extracts chronological events from sessions |
BUILT |
| Session Export | Export sessions to Markdown format | BUILT |
| Session Outcome Extraction | Extracts and persists structured outcomes from sessions | BUILT |
| Thread Classification | Classifies conversation threads by freshness (active, stale, dormant) | BUILT |
| Open Question Extraction | Identifies unresolved questions from session transcripts | BUILT |
1.9 Memory Weaver (Consolidation)
| Feature | Description | Status |
|---|---|---|
| MemoryWeaver | Consolidates and compacts memory frames over time | BUILT |
| Skill Extraction | Extracts reusable skills from session transcripts | BUILT |
| Memory Decay | Time-based decay for less important memories | BUILT |
| Manual Trigger | POST /api/weaver/trigger for on-demand consolidation |
BUILT |
| Status API | GET /api/weaver/status shows last consolidation times |
BUILT |
1.10 Embedding System
| Feature | Description | Status |
|---|---|---|
| Multi-Provider Chain | InProcess -> Ollama -> Voyage -> OpenAI -> LiteLLM -> Mock fallback | BUILT |
| Tier-Gated Providers | SOLO gets inprocess+mock; BASIC adds ollama/voyage/openai; TEAMS adds litellm | BUILT |
| Monthly Quota | Tracked in embedding_usage table, enforced per tier |
BUILT |
| In-Process Embeddings | Local transformer model via @xenova/transformers (no API key needed) | BUILT |
| LiteLLM Embedder | Proxy through LiteLLM for any compatible model | BUILT |
| Ollama Embedder | Local Ollama server for embeddings | BUILT |
| API Embedder | Direct Voyage AI / OpenAI API calls | BUILT |
2. AGENT PERSONA SYSTEM (Crown Jewel #2)
22 fully-defined personas with distinct system prompts, tool allowlists/denylists, failure patterns, and behavioral guardrails.
2.1 Universal Personas (8)
| Persona | Tools | Key Capability |
|---|---|---|
| General Purpose | Full tool set (30+) | Adapts to any task, suggests specialists |
| Researcher | web_search, web_fetch, search_memory, save_memory, generate_docx | Multi-source investigation, citation tracking |
| Writer | read/write/edit_file, search_memory, generate_docx | Document creation, tone adaptation, no bash/git |
| Analyst | bash, read/write_file, web_search, generate_docx | Data analysis, frameworks, quantified outputs |
| Coder | bash, all file ops, full git suite | Code review, debugging, architecture |
| Project Manager | plan tools, memory, file ops | Task tracking, status reports, no git/bash |
| Planner (read-only) | read_file, search, memory, plan tools | Strategic planning only, no write tools |
| Verifier (read-only) | read_file, search, bash (read-only) | Adversarial QA, VERDICT output format |
2.2 Domain Specialists (14)
| Persona | Domain |
|---|---|
| Executive Assistant | Email drafting, meeting prep, calendar |
| Sales Rep | Lead research, outreach, pipeline |
| Marketer | Content, campaigns, SEO, social media |
| Senior PM | PRD drafting, RICE scoring, roadmap |
| HR Manager | Policy, onboarding, compliance |
| Legal Counsel | Contract review, compliance, jurisdiction |
| Business Finance | Budget analysis, projections, invoicing |
| Strategy Consultant | Frameworks (MECE, Porter), client deliverables |
| Coordinator | Pure orchestrator -- only spawn_agent, list_agents, get_agent_result |
| Customer Support | Ticket resolution, KB management, escalation |
| Operations Manager | SOPs, vendor management, process optimization |
| Data Engineer | SQL, pipelines, dashboards, data quality |
| Recruiter | Job descriptions, screening scorecards, interview prep |
| Creative Director | Creative briefs, feedback synthesis, brand consistency |
2.3 Persona Features
| Feature | Description | Status |
|---|---|---|
| Per-Persona Tool Allowlist | Each persona declares which tools it can use | BUILT |
| Per-Persona Denylist | disallowedTools enforced at pool level |
BUILT |
| Read-Only Mode | isReadOnly flag prevents all write operations |
BUILT |
| Failure Patterns | 3+ documented failure modes per persona for self-correction | BUILT |
| Tagline + BestFor + WontDo | User-facing hover metadata | BUILT |
| Suggested Skills/Connectors/MCP | Each persona suggests relevant capabilities | BUILT |
| Default Workflow | Persona maps to a workflow template (research-team, plan-execute, etc.) | BUILT |
| Custom Personas | loadCustomPersonas() from disk + POST /api/personas |
BUILT |
3. MULTI-AGENT ORCHESTRATION (Crown Jewel #3)
3.1 Sub-Agent System
| Feature | Description | Status |
|---|---|---|
| spawn_agent | Launches specialist agents with specific tasks and persona roles | BUILT |
| list_agents | Check status of running agents | BUILT |
| get_agent_result | Retrieve completed agent output | BUILT |
| Agent Communication | send_agent_message / check_agent_messages for inter-agent messaging | BUILT |
| SubagentOrchestrator | Full orchestration with dependency ordering and context flow between steps | BUILT |
| Agent Groups | CRUD for multi-agent group configurations (parallel/sequential/coordinator strategies) | BUILT |
3.2 Workflow Templates (Built-in)
| Template | Description | Status |
|---|---|---|
| research-team | Researcher -> Synthesizer -> Reviewer (3 agents) | BUILT |
| review-pair | Writer -> Reviewer -> Reviser (draft/critique/revise cycle) | BUILT |
| plan-execute | Planner -> Executor -> Summarizer | BUILT |
| ticket-resolve | Triage -> Investigator -> Responder (support workflow) | BUILT |
| content-pipeline | Researcher -> Drafter -> Editor (content creation) | BUILT |
3.3 Workflow Composer
| Feature | Description | Status |
|---|---|---|
| Task Shape Detection | Analyzes user message to detect task type/complexity | BUILT |
| Execution Mode Selection | Chooses lightest sufficient mode: direct, structured_single_agent, skill_guided, subagent_workflow | BUILT |
| Custom Workflows | Create/delete custom workflow templates via API | BUILT |
3.4 Waggle Dance Protocol (Team Communication)
| Feature | Description | Status |
|---|---|---|
| Message Types | request, response, broadcast with typed subtypes | BUILT |
| Task Delegation | Dispatches task_delegation messages to spawn workers | BUILT |
| Knowledge Check | Searches memory to answer knowledge_check requests | BUILT |
| Skill Request/Share | Routes skill requests and shares between agents | BUILT |
3.5 Fleet Management
| Feature | Description | Status |
|---|---|---|
| Fleet Status | GET /api/fleet -- lists all active workspace sessions with model, cost, tool count | BUILT |
| Spawn Agent | POST /api/fleet/spawn -- create new agent session (BASIC+ tier) | BUILT |
| Pause/Resume/Kill | Per-workspace session controls | BUILT |
| Max Sessions by Tier | SOLO=3, BASIC=10, TEAMS=25, ENTERPRISE=100 | BUILT |
4. TOOL ECOSYSTEM
4.1 System Tools (12)
| Tool | Description | Status |
|---|---|---|
| bash | Shell command execution | BUILT |
| read_file | Read file content | BUILT |
| write_file | Write file content | BUILT |
| edit_file | Edit file with patch | BUILT |
| multi_edit | Batch edit multiple files | BUILT |
| search_files | Glob-pattern file search | BUILT |
| search_content | Content/regex search across files | BUILT |
| web_search | LLM-powered web search | BUILT |
| web_fetch | Fetch URL content | BUILT |
| run_code | Execute code in sandboxed runtime | BUILT |
| get_task_output | Get output from background task | BUILT |
| kill_task | Kill a running background task | BUILT |
4.2 Mind Tools (7)
| Tool | Description | Status |
|---|---|---|
| get_identity | Agent identity context | BUILT |
| get_awareness | Current tasks, flags, tool utilization | BUILT |
| search_memory | Hybrid search across personal + workspace memory | BUILT |
| search_all_workspaces | Cross-workspace search | BUILT |
| save_memory | Save to workspace or personal mind with routing guards | BUILT |
| query_knowledge | Query knowledge graph entities and relations | BUILT |
| add_task | Add task to awareness layer | BUILT |
| correct_knowledge | Correct or invalidate knowledge entities | BUILT |
4.3 Git Tools (11)
| Tool | Description | Status |
|---|---|---|
| git_status, git_diff, git_log | Read operations | BUILT |
| git_commit, git_branch, git_stash | Write operations | BUILT |
| git_push, git_pull, git_merge | Remote operations | BUILT |
| git_pr | Create pull requests | BUILT |
4.4 Plan Tools (4)
| Tool | Description | Status |
|---|---|---|
| create_plan | Create structured multi-step plan | BUILT |
| add_plan_step | Add step to existing plan | BUILT |
| execute_step | Execute a plan step | BUILT |
| show_plan | Display current plan status | BUILT |
4.5 Document Tools (1)
| Tool | Description | Status |
|---|---|---|
| generate_docx | Generate Word documents with formatting | BUILT |
4.6 Skill Tools (7)
| Tool | Description | Status |
|---|---|---|
| list_skills | List installed skills | BUILT |
| create_skill | Create new skill | BUILT |
| delete_skill | Remove skill | BUILT |
| read_skill | Read skill content | BUILT |
| search_skills | Search skills by keyword | BUILT |
| suggest_skill | Recommend skills for current context | BUILT |
| acquire_capability / install_capability | Install from marketplace or starter pack | BUILT |
4.7 Search Provider Tools (3)
| Tool | Description | Status |
|---|---|---|
| perplexity_search | Perplexity AI search (requires API key) | BUILT |
| tavily_search | Tavily search (requires API key) | BUILT |
| brave_search | Brave Search (requires API key) | BUILT |
4.8 Browser Tools (6)
| Tool | Description | Status |
|---|---|---|
| browser_navigate | Navigate to URL | BUILT |
| browser_screenshot | Take page screenshot | BUILT |
| browser_click | Click element | BUILT |
| browser_fill | Fill form field | BUILT |
| browser_evaluate | Execute JavaScript | BUILT |
| browser_snapshot | Accessibility snapshot | BUILT |
4.9 LSP Tools (4)
| Tool | Description | Status |
|---|---|---|
| lsp_diagnostics | Get code diagnostics | BUILT |
| lsp_definition | Go to definition | BUILT |
| lsp_references | Find references | BUILT |
| lsp_hover | Hover info | BUILT |
4.10 CLI Discovery Tools (2)
| Tool | Description | Status |
|---|---|---|
| cli_discover | Detect installed CLI tools (git, node, docker, aws, etc. -- 26 tools) | BUILT |
| cli_execute | Execute a discovered CLI tool safely | BUILT |
4.11 Cron Tools (4)
| Tool | Description | Status |
|---|---|---|
| create_schedule | Create cron schedule | BUILT |
| list_schedules | List all schedules | BUILT |
| delete_schedule | Remove schedule | BUILT |
| trigger_schedule | Manually trigger | BUILT |
4.12 Team Tools (9)
| Tool | Description | Status |
|---|---|---|
| check_hive | Check team status | BUILT |
| share_to_team | Share content with team | BUILT |
| create_team_task | Create task for team | BUILT |
| claim_team_task | Claim a team task | BUILT |
| send_waggle_message | Send Waggle Dance protocol message | BUILT |
| request_team_capability | Request capability from team | BUILT |
| team_activity | View team activity feed | BUILT |
| team_tasks / team_members | List tasks and members | BUILT |
| assign_task / complete_task | Task lifecycle management | BUILT |
4.13 KVARK Tools (4)
| Tool | Description | Status |
|---|---|---|
| kvark_search | Search KVARK enterprise platform | BUILT |
| kvark_feedback | Submit feedback to KVARK | BUILT |
| kvark_action | Execute KVARK action | BUILT |
| kvark_ask_document | Query KVARK document | BUILT |
4.14 Audit Tools (1)
| Tool | Description | Status |
|---|---|---|
| query_audit | Query audit trail events | BUILT |
4.15 Workflow Tools (2)
| Tool | Description | Status |
|---|---|---|
| compose_workflow | Generate workflow plan from task description | BUILT |
| orchestrate_workflow | Execute a multi-agent workflow template | BUILT |
Total: ~80+ distinct agent tools
5. WORKSPACE SYSTEM
5.1 Workspace CRUD
| Feature | Description | Status |
|---|---|---|
| Create Workspace | With name, description, path, persona, model, template | BUILT |
| List Workspaces | With group and teamId filters | BUILT |
| Get Workspace Detail | Full config + memory stats + recent decisions + summary | BUILT |
| Update Workspace | Config, persona, model changes | BUILT |
| Delete Workspace | With mind DB cleanup | BUILT |
| Workspace Summary | Narrative summary with activity status, memory count, key topics | BUILT |
5.2 Workspace Templates (15)
| Template | Persona | Category |
|---|---|---|
| Sales Pipeline | sales-rep | sales |
| Research Project | researcher | research |
| Code Review | coder | engineering |
| Marketing Campaign | marketer | marketing |
| Product Launch | product-manager-senior | operations |
| Legal Review | legal-professional | legal |
| Agency/Consulting | consultant | custom |
| + 8 more via template system | various | various |
Template features:
- Starter memory (pre-seeded workspace context)
- Suggested commands
- Welcome messages per template
- Template-to-capability-pack mapping (auto-installs relevant skills)
5.3 Workspace Context
| Feature | Description | Status |
|---|---|---|
| Time-Aware Greeting | Morning/afternoon/evening contextual greetings | BUILT |
| Upcoming Schedules | Shows next cron jobs for the workspace | BUILT |
| Workspace State | Activity level, memory count, decisions, progress | BUILT |
| Recent Decisions | Extracted from memory for quick context | BUILT |
5.4 File Management
| Feature | Description | Status |
|---|---|---|
| File Registry | Track ingested files per workspace | BUILT |
| File Ingestion | Process images, documents (PDF, DOCX, PPTX), spreadsheets (XLSX, CSV), code files, archives | BUILT |
| Directory Browser | Browse local filesystem for workspace path selection | BUILT |
| Document Version Registry | Track document versions within a workspace | BUILT |
5.5 Pins
| Feature | Description | Status |
|---|---|---|
| Pin Messages | Favorite/bookmark messages within workspace | BUILT |
| Draft/Final Status | W7.4 -- mark pinned outputs as draft or final | BUILT |
| Per-Workspace Storage | JSON file in workspace directory | BUILT |
6. CONNECTOR / INTEGRATION SYSTEM
6.1 Built-in Connectors (28)
| Category | Connectors |
|---|---|
| Code | GitHub, GitLab, Bitbucket |
| Communication | Slack, Discord, MS Teams, Gmail, Outlook, Email |
| Project Management | Jira, Linear, Asana, Trello, Monday |
| Knowledge | Notion, Confluence, Obsidian |
| CRM | HubSpot, Salesforce, Pipedrive |
| Cloud Storage | Google Drive, Google Docs, Google Sheets, OneDrive, Dropbox |
| Calendar | Google Calendar |
| Data | Airtable, PostgreSQL |
| Meta | Composio (250+ integrations gateway) |
| Mock | MockSlack, MockTeams, MockDiscord (for testing) |
6.2 Connector Features
| Feature | Description | Status |
|---|---|---|
| Connect/Disconnect | Store/remove credentials in vault | BUILT |
| Health Check | Per-connector health status (connected, expired, disconnected) | BUILT |
| OAuth Flow | Full OAuth for GitHub, Slack, Google, Notion, Jira | BUILT |
| Connector Registry | Dynamic registry with definitions and live status | BUILT |
6.3 MCP (Model Context Protocol) Runtime
| Feature | Description | Status |
|---|---|---|
| MCP Server Lifecycle | Start/stop/health for stdio-based MCP servers | BUILT |
| JSON-RPC 2.0 | Full protocol implementation | BUILT |
| Tool Discovery | Lists tools from connected MCP servers | BUILT |
| Per-Workspace MCP | Servers can be scoped to specific workspaces | BUILT |
6.4 Plugin System
| Feature | Description | Status |
|---|---|---|
| PluginManager | Load/install/uninstall plugins from ~/.waggle/plugins/ | BUILT |
| Plugin Runtime | PluginRuntimeManager for lifecycle management | BUILT |
| Plugin Validation | Manifest validation via SDK | BUILT |
| Starter Skills | Auto-install starter skills on first run | BUILT |
| Capability Packs | Grouped skill bundles (developer_workspace, research_analyst, etc.) | BUILT |
7. MARKETPLACE
| Feature | Description | Status |
|---|---|---|
| FTS5 Search | Full-text search across package catalog | BUILT |
| Faceted Filters | Filter by type, category, pack, source | BUILT |
| Sort Options | relevance, popular, recent, name | BUILT |
| Install/Uninstall | Package lifecycle management | BUILT |
| Security Gate | Cisco scanner integration for package security scanning | BUILT |
| Scan Status | passed/failed/not_scanned per package | BUILT |
| Enterprise Packs | Curated package bundles for specific use cases | BUILT |
| Package Categories | Structured categorization system | BUILT |
| Install Audit | Track all marketplace installs with timestamps | BUILT |
| Skill Hash Tracking | Detect if installed skills have been modified | BUILT |
8. TIER / BILLING SYSTEM
8.1 Tier Architecture
| Tier | Price | Key Limits |
|---|---|---|
| SOLO | Free | 5 workspaces, 10 connectors, 500 embeddings/mo, no sub-agents, no custom skills |
| BASIC | $15/mo | Unlimited workspaces/connectors, 5000 embeddings/mo, sub-agents, custom skills |
| TEAMS | $79/mo/seat | Everything + team library, cloud sync, shared workspaces, admin panel, audit log |
| ENTERPRISE | Consultative | Everything + self-hosted, full audit, KVARK integration |
8.2 Feature Gating
| Feature | Description | Status |
|---|---|---|
| TierCapabilities | 17 capability flags per tier | BUILT |
| requireTier Middleware | Fastify preHandler that enforces minimum tier | BUILT |
| tierSatisfies | Comparison function for tier ordering | BUILT |
| parseTier | Handles legacy lowercase tier names -> canonical uppercase | BUILT |
8.3 Stripe Integration
| Feature | Description | Status |
|---|---|---|
| Checkout | Stripe Checkout session creation | BUILT |
| Webhooks | Stripe webhook handling (subscription events) | BUILT |
| Customer Portal | Stripe customer portal redirect | BUILT |
| Tier-from-Price Mapping | Map Stripe price IDs back to canonical tiers | BUILT |
| Lazy SDK Init | Stripe SDK only loaded if STRIPE_SECRET_KEY is set | BUILT |
9. LLM PROVIDER SYSTEM
9.1 Provider Registry
| Feature | Description | Status |
|---|---|---|
| Multi-Provider | Anthropic, OpenAI, Google, DeepSeek, xAI, Mistral, Alibaba/Qwen, MiniMax, GLM/Zhipu, OpenRouter, Perplexity, Moonshot | BUILT |
| Model Catalog | Per-provider model list with cost/speed ratings | BUILT |
| Vault Key Status | Shows which providers have API keys configured | BUILT |
| LiteLLM Proxy | Routes through LiteLLM for model abstraction | BUILT |
| Anthropic Direct Proxy | Fallback direct Anthropic proxy when LiteLLM is down | BUILT |
| Health Check | Provider health: healthy, degraded, unavailable | BUILT |
| Credential Pool | Round-robin + cooldown across multiple keys per provider | BUILT |
9.2 Cost Tracking
| Feature | Description | Status |
|---|---|---|
| Per-Model Pricing | Default pricing for Claude, GPT, etc. (input + output per 1k tokens) | BUILT |
| Usage Recording | Track input/output tokens per request with workspace attribution | BUILT |
| By-Model Breakdown | Aggregate cost/tokens per model | BUILT |
| By-Workspace Cost | GET /api/cost/by-workspace (TEAMS+) | BUILT |
| Cost Reset | POST /api/agent/cost/reset | BUILT |
| Iteration Budget | Per-session budget tracking with isWithinBudget check | BUILT |
10. SECURITY FEATURES
| Feature | Description | Status |
|---|---|---|
| Injection Scanner | 3 pattern sets: role_override (12 patterns), prompt_extraction (6 patterns), instruction_injection (7 patterns) | BUILT |
| Multi-Language Patterns | Detects injection in English, German, Spanish, French | BUILT |
| Memory Wipe Detection | Catches "forget everything", "erase all context" attempts | BUILT |
| Authority Claim Detection | Detects fake system/admin messages, debug mode claims | BUILT |
| XSS Sanitization | Memory frame content sanitized (script tags, event handlers, javascript: URIs) | BUILT |
| Vault (Encrypted Secrets) | AES-encrypted secret storage for API keys and tokens | BUILT |
| Vault Categories | LLM Providers, Embedding, Search, Code/DevOps, Communication, Business, Storage | BUILT |
| Secret Reveal | POST /api/vault/:name/reveal -- explicit decrypt action | BUILT |
| Rate Limiter | Server-level rate limiting middleware | BUILT |
| CORS Configuration | Allowed origins list with validation | BUILT |
| Input Validation | assertSafeSegment for path traversal prevention | BUILT |
| Parameterized Queries | All SQL uses parameterized queries (no string interpolation) | BUILT |
| Trust Assessment | assessTrust function for evaluating content trustworthiness | BUILT |
11. PROACTIVE / AUTONOMOUS FEATURES
11.1 Cron Scheduling
| Feature | Description | Status |
|---|---|---|
| Cron CRUD | Create, read, update, delete schedules via REST API | BUILT |
| Job Types | Multiple job types with configurable payloads | BUILT |
| Manual Trigger | POST /api/cron/:id/trigger | BUILT |
| Cross-Workspace | workspaceId="*" for global cron jobs |
BUILT |
| Default Crons | Auto-seed default cron schedules on first run | BUILT |
| Delivery System | Email delivery of cron results with HTML escaping (XSS safe) | BUILT |
| Delivery Preferences | Configurable delivery channels per schedule | BUILT |
11.2 Proactive Handlers
| Feature | Description | Status |
|---|---|---|
| Morning Briefing | Daily summary of workspace state and priorities | BUILT |
| Stale Workspace Detection | Flags workspaces with no recent activity | BUILT |
| Pending Task Reminders | Alerts for unresolved awareness items | BUILT |
| Capability Suggestions | Recommends skills/tools based on usage patterns | BUILT |
11.3 Monthly Self-Assessment
| Feature | Description | Status |
|---|---|---|
| Performance Report | Total interactions, correction rate, improvement trend | BUILT |
| Strengths/Weaknesses | Auto-derived from optimization logs and feedback | BUILT |
| Capability Gaps | Detected gaps from improvement signals | BUILT |
| Saved to Memory | Assessment becomes part of long-term self-awareness | BUILT |
12. DATA MANAGEMENT
12.1 Backup & Restore
| Feature | Description | Status |
|---|---|---|
| Encrypted Backup | AES-256-GCM encrypted ZIP archive of ~/.waggle/ | BUILT |
| Stream Download | POST /api/backup streams the archive | BUILT |
| Restore | POST /api/restore accepts encrypted backup | BUILT |
| Metadata | GET /api/backup/metadata -- last backup info | BUILT |
| Size Limit | 500 MB max, batched file reads to limit memory pressure | BUILT |
| Exclusions | node_modules, .git, marketplace.db excluded | BUILT |
12.2 Import / Export
| Feature | Description | Status |
|---|---|---|
| ChatGPT Import | Preview and commit ChatGPT conversation exports | BUILT |
| Claude Import | Preview and commit Claude conversation exports | BUILT |
| Knowledge Extraction | Extracts knowledge items from imported conversations | BUILT |
| GDPR Export | ZIP file with memories, sessions, workspaces, settings (keys masked), vault metadata | BUILT |
| Per-Workspace Export | Scoped export for agency/multi-client privacy | BUILT |
| Audit Event on Export | Every export operation logged to audit trail | BUILT |
13. REAL-TIME & COMMUNICATION
13.1 Notifications
| Feature | Description | Status |
|---|---|---|
| SSE Stream | Server-Sent Events for live notifications | BUILT |
| Categories | cron, approval, task, message, agent | BUILT |
| Sub-Agent Status | Real-time status updates for running sub-agents | BUILT |
| Workflow Suggestions | Pattern detection prompts for workflow capture | BUILT |
13.2 Waggle Signals
| Feature | Description | Status |
|---|---|---|
| Signal Types | agent:started, tool:called, memory:saved, agent:completed | BUILT |
| SSE Stream | GET /api/waggle/stream for real-time signal feed | BUILT |
| Acknowledgment | PATCH /api/waggle/signals/:id/ack | BUILT |
| In-Memory Store | Capped at 500 signals | BUILT |
13.3 Audit Trail
| Feature | Description | Status |
|---|---|---|
| Event Types | 14 types: tool_call, tool_result, memory_write, memory_delete, workspace CRUD, session lifecycle, approvals, exports, cron_trigger | BUILT |
| Separate Database | audit.db in dataDir (not in .mind) | BUILT |
| Paginated Listing | GET /api/events with filters | BUILT |
| Aggregate Stats | GET /api/events/stats by type and day | BUILT |
| SSE Stream | GET /api/events/stream for live Cockpit feed | BUILT |
| Retention | 90-day default, auto-cleanup via cron | BUILT |
14. USER PROFILE & PERSONALIZATION
| Feature | Description | Status |
|---|---|---|
| Identity Profile | Name, role, company, industry, bio, avatar | BUILT |
| Writing Style Analysis | Tone, sentence length, vocabulary, structure -- analyzed from text samples | BUILT |
| Brand Profile | Company colors (primary/secondary/accent), fonts, logo description | BUILT |
| Per-Format Styles | DOCX, PPTX, PDF, XLSX style preferences | BUILT |
| Online Research | POST /api/profile/research -- research user/company online | BUILT |
| Agent-Accessible | GET /api/profile/style and /api/profile/brand for agent tools | BUILT |
| Memory Integration | Key profile fields saved to personal memory for agent access | BUILT |
15. CHAT SYSTEM
| Feature | Description | Status |
|---|---|---|
| WebSocket Chat | Real-time bidirectional chat via WebSocket | BUILT |
| Session Persistence | Messages persisted to JSONL files per workspace | BUILT |
| Context Window Management | Configurable MAX_CONTEXT_MESSAGES with intelligent pruning | BUILT |
| Context Compression | Iterative conversation compression for long sessions | BUILT |
| Skill Prompt Injection | Active skills appended to system prompt per message | BUILT |
| Governance Permissions | Per-workspace tool/action permissions | BUILT |
| Dramatic Claim Hook | pre:memory-write hook flags dramatic claims before saving | BUILT |
| System Prompt Caching | Cached per session to avoid rebuilding on every message | BUILT |
| Ambiguity Detection | Detects ambiguous messages and prompts for clarification | BUILT |
| Schedule Suggestions | Detects recurring task patterns and suggests cron schedules | BUILT |
| Regulated Content Detection | Flags legal/financial/medical content for disclaimers | BUILT |
| Retry Logic | Retryable error detection for transient LLM failures | BUILT |
| Approval System | Human-in-the-loop approval for sensitive tool calls | BUILT |
| Capability Gap Recording | Records when agent cannot fulfill a request | BUILT |
| Correction Analysis | Analyzes user corrections to improve future responses | BUILT |
| Workflow Capture Suggestion | Detects repeatable patterns and suggests saving as workflow | BUILT |
16. OFFLINE MODE
| Feature | Description | Status |
|---|---|---|
| Offline Detection | OfflineManager tracks connectivity state | BUILT |
| Message Queue | Queue messages for when connection restores | BUILT |
| Offline-Capable Tools | Tools marked with offlineCapable: true (all mind tools) |
BUILT |
| Status API | GET /api/offline/status | BUILT |
| Queue Management | List, clear, and delete individual queued messages | BUILT |
17. TEAM FEATURES
| Feature | Description | Status |
|---|---|---|
| Team DB | SQLite database for teams and members with role-based access | BUILT |
| Team Connect/Disconnect | Connect desktop app to team server | BUILT |
| Team Roles | owner, admin, member, viewer | BUILT |
| Team Sync | TeamSync from @waggle/core for cross-team data synchronization | BUILT |
| Cloud Sync | Tier-gated (TEAMS+) cloud synchronization | PARTIAL |
| Admin Panel | Admin overview endpoint (TEAMS+) | BUILT |
| Shared Workspaces | Tier-gated shared workspace access | PARTIAL |
18. TELEMETRY & OBSERVABILITY
| Feature | Description | Status |
|---|---|---|
| Telemetry Store | SQLite-backed event telemetry | BUILT |
| Summary API | GET /api/telemetry/summary | BUILT |
| Event Cleanup | DELETE /api/telemetry/events | BUILT |
| Status | GET /api/telemetry/status | BUILT |
19. ONBOARDING
| Feature | Description | Status |
|---|---|---|
| 7-Step Wizard | Welcome -> WhyWaggle -> MemoryImport -> Template -> Persona -> APIKey -> HiveReady | BUILT |
| Template Selection | 7+ workspace templates to choose from | BUILT |
| Persona Selection | Subset of personas for initial configuration | BUILT |
| API Key Setup | Guided key entry with provider selection | BUILT |
| Memory Import | Import from ChatGPT/Claude on first run | BUILT |
20. BEHAVIORAL SPEC & INTELLIGENCE
| Feature | Description | Status |
|---|---|---|
| BEHAVIORAL_SPEC | 273-line rules covering core loop, quality, behavior, work patterns, tools, intelligence defaults | BUILT |
| Smart Router | routeMessage() for intelligent message routing | BUILT |
| Capability Router | Resolves capabilities across native tools, skills, plugins, MCP, sub-agents, connectors | BUILT |
| Tool Filter | filterToolsForContext() with allowlist/denylist enforcement | BUILT |
| Tool Availability Check | Dynamic checkAvailability() per tool (e.g., browser connected, git repo present) |
BUILT |
| Skill Recommender | Context-aware skill suggestions based on current task | BUILT |
| Hook Registry | Pre/post tool use hooks for validation and auto-formatting | BUILT |
| Command Registry | Slash command system with workflow and marketplace commands | BUILT |
| Optimization Logging | Track agent performance for self-improvement | BUILT |
| Improvement Signals | ImprovementSignalStore for detecting recurring issues | BUILT |
CROWN JEWELS -- What Makes Waggle OS Unique
-
Dual-Mind Persistent Memory -- No other AI agent platform has workspace-isolated memory with cross-mind dedup, knowledge graph extraction, provenance tracking, and confidence derivation. The memory survives across sessions and grows smarter over time.
-
22 Specialized Personas with Behavioral Guardrails -- Not just "system prompts" but full personas with tool allowlists, denylists, failure pattern documentation, read-only enforcement, and suggested capability stacks. The Coordinator persona that can only delegate is particularly unique.
-
Multi-Agent Orchestration with Workflow Templates -- Built-in multi-agent workflows (research-team, review-pair, plan-execute, ticket-resolve, content-pipeline) with dependency ordering, context flow between steps, and automatic result aggregation.
-
Memory Weaver -- Automatic consolidation, decay, and strengthening of memories over time. Combined with monthly self-assessment that saves performance reports back into the agent's own memory, creating a genuine self-improvement loop.
-
80+ Agent Tools -- Comprehensive tool ecosystem spanning system ops, git, planning, documents, web, browser, LSP, CLI discovery, cron scheduling, team collaboration, and KVARK enterprise integration.
-
28 Built-in Connectors + MCP Runtime -- Native integrations for GitHub, Slack, Jira, Salesforce, Google Workspace, etc., plus MCP server support for extensibility. OAuth flows for major platforms.
-
Encrypted Backup/Restore + GDPR Export -- AES-256-GCM encrypted backup with per-workspace scoped export for agency privacy compliance.
-
Tiered Feature Gating with Stripe -- Full SOLO -> BASIC -> TEAMS -> ENTERPRISE tier system with per-feature capability gating, Stripe checkout/webhook/portal integration, and KVARK enterprise upsell funnel.
-
Proactive Intelligence -- Morning briefings, stale workspace detection, task reminders, capability suggestions, and monthly self-assessments -- all running on cron without user prompting.
-
Security-First Design -- Injection scanner with 25+ patterns (multilingual), XSS sanitization on memory writes, workspace-level data isolation, encrypted vault, rate limiting, parameterized SQL, and trust assessment on all external content.