# Aria Roadmap

This roadmap reflects the code that is actually present in the repository today.

## In Place

- [x] Multi-agent runtime with YAML agent configs
- [x] Hot reload for agent configs and skill files
- [x] Bun + Hono gateway with chat, auth, agent, memory, task, and settings APIs
- [x] SQLite persistence for conversations, memories, tasks, daily logs, contacts, cron jobs, and learning state
- [x] Multi-provider model support through registry adapters
- [x] Per-agent prompts with persona, task persona, rules, master skill, task state, recent context, and feedback-derived preferences
- [x] Tool-use loop with permission modes and dangerous-command blocking
- [x] Shared and per-agent memory namespaces
- [x] Skill system with YAML persistence, versioning, ratings, and visibility controls
- [x] Feedback and taste-profile system
- [x] Task board with assignment, review, dependencies, and event emission
- [x] Legacy cron scheduler plus unified trigger engine
- [x] Nightly dream review runner
- [x] Tunnel server and client for local-machine execution from a remote runtime
- [x] Telegram bots with per-agent config and shared-bot switching
- [x] Discord bots with slash commands and shared-bot switching
- [x] Discord voice support
- [x] Mission Control routes for chat, agents, memory, tasks, cron, journal, contacts, rooms, editor, activity, and settings
- [x] First-run setup wizard
- [x] Google Workspace, Slack, email, YouTube, Apify, image, contacts, notes, reminders, and iMessage tool families

## Needs Polish

- [ ] Tighten and simplify the production frontend build/export story
- [ ] Add a real top-level README for operators and contributors
- [ ] Expand test coverage around gateway endpoints, triggers, and channel adapters
- [ ] Document environment-variable and integration setup in one place
- [ ] Clarify the long-term role split between legacy `crons` and unified `triggers`
- [ ] Make daemon/service installation easier to operate across Linux and macOS

## Product Gaps

- [ ] Stronger Mission Control UX around triggers, rules, dream runs, and channel setup
- [ ] Better observability for agent/tool execution, failures, and trigger history
- [ ] More explicit import/export workflows for agents, skills, and user state
- [ ] Safer and clearer approval UX outside current dev-oriented defaults
- [ ] More consistent docs around built-in agents versus user-defined agents

## Longer-Term

- [ ] More channel adapters beyond Telegram and Discord
- [ ] Better packaging and install ergonomics for non-developers
- [ ] Richer multi-agent room workflows and collaborative review flows
- [ ] Smarter model-routing policies and cost-aware execution defaults
- [ ] Better replication / backup / restore story for self-hosted deployments