Roadmap

Roadmap

AI Governance Infrastructure — When governance is structural, agent capabilities become safely tradeable.


Phase Overview

Phase 1-3d         v0.1              v0.2                  v0.3                    v1.0
Foundation         Initial Release   Plugin System+OAuth   Bot Enhancement         Production Ready
Completed          Completed         Completed             Completed               ──────────▶

Completed

v0.1 — Initial Release

  • Three-layer resource model (ResourceType -> Resource -> Bot)
  • Unified plugin contract (ToolDef + ScopeDimension + CostDef)
  • ToolCatalog loading tool index from database
  • Declarative scope checker
  • 3 built-in types + GitHub reference plugin (8 tools)
  • Full documentation + bilingual (English/Chinese)
  • Code review: 3 rounds, 49 fixes

v0.2 — Plugin System + OAuth + SDK Infrastructure

  • auth_methods data model — AuthMethod enum, AuthMethodDef, unified credential creation UI
  • OAuth infrastructure — OAuthProvider management, authorize/callback flow, automatic token refresh
  • Plugin auto-discovery — PluginManifest, PluginLoader, PluginManager, REST API, 24 tests
  • .mst import + hot-reload — Upload, validate, extract, reload plugins at runtime
  • Plugin i18n + SDK + CLI — Locale translations, monstrum_sdk package, CLI scaffolding
  • OAuth SDK — Multi grant type support (AuthCode + PKCE + ClientCreds + DeviceCode)
  • Token lifecycle management — Background refresh, per-credential locking, expiry notifications
  • SDK infrastructure — HttpExecutorBase, PluginClient, Platform SDK, 49 tests

v0.3 — Bot Enhancement + Feature Completion

  • Bot capabilities — Async task lifecycle, local OS execution, web access (multi-engine search), bot memory tools, Skills system, proactive messaging, scheduled task tools
  • Session management — Session persistence, task progress reporting, channel binding, group chat formatting, conversation compression
  • Memory system — Database migration, partitioned memory (origin field), MemoryContext three-layer cache, LLM-based memory extraction
  • Prompt management — 9 centralized prompt keys, three-layer resolution, DB storage + API + UI
  • Workflow improvements — AST condition evaluator, parallel fail-fast, DAG cycle detection, timeout budgets, variable piping, event/schedule triggers, execution cancellation
  • Events & delegation — Unified event system (PlatformEvent + EventSubscription + EventDispatcher), Bot event tools, DelegateConstraints
  • Infrastructure — Docker Sandbox, Gateway message policies, Agent Mode (reactive/planning/adaptive)

v1.0 — Production Ready

Planned Features

IDFeatureDescriptionPriority
B2Streaming responses (SSE/WebSocket)Real-time LLM output push to frontendHigh
B5Budget alertsNotify users when thresholds are reachedMedium
RA1-4Resource AliasMulti-credential tool routingMedium
R1BotRole permission modelRBAC role preset permission templatesMedium
DX3Error message improvementsErrors include context and fix suggestionsLow
P6Frontend bundle optimizationRoute lazy loading + antd tree-shakingLow

Template Marketplace Enhancements

FeatureDescription
Template versioningTemplates support version numbers, users can deploy specific versions
Ratings and reviewsCommunity feedback mechanism for quality filtering
One-click deploy optimizationAuto-detect required ResourceTypes and Providers

Long-term Backlog

FeatureDescription
Bot cloningCopy existing Bot configuration to create new Bot
Credential expiry remindersNotify N days before expiration
Resource health checksPeriodic heartbeat, offline alerts
Workflow Dry-RunPreview execution path without running
Human handoffTransfer conversation to human operator
Task subtree visualizationNested call parent-child relationship display

Milestones

VersionGoalStatus
v0.1Initial release: full features + docs + three-layer resource modelReleased
v0.2Plugin system + OAuth + SDK infrastructureCompleted
v0.3Bot enhancement + session persistence + partitioned memory + prompt management + workflow + events + Docker sandboxCompleted
v1.0Production ready: security scanning, horizontal scaling, enterprise SSOLong-term