Raxol Roadmap

Multi-surface application runtime for Elixir. One TEA module, four render targets.

Current Version: v2.6.0

What's Done

• Phase 1: Core Runtime: TEA architecture, TTY detection, ANSI input/output, render pipeline end-to-end
• Phase 2: Widget Library: 23 widgets with tests and examples (exceeds original 15-widget target)
• Phase 3: Framework Polish: Focus management, W3C-style event capture/bubble, style inheritance, terminal compatibility (color downsampling, Unicode width, synchronized output)
• Phase 4: OTP Differentiators: Process-per-component crash isolation, hot code reload, LiveView bridge, SSH app serving
• Phase 5: Ecosystem Tooling: Hex package, Phoenix optional, mix raxol.new generator, documentation, tech debt cleanup, showcase, session recording
• Phase 6: Playground + Charts: 30 demos across 8 categories, 7 chart modules (braille-resolution), web refactor, SSH playground
• Time-Travel Debugging: Snapshot every update/2 cycle, cursor navigation, restore, export/import. Zero cost when disabled
• Agent Framework: TEA-based AI agents, coordinator/worker teams, 7 agent harness gaps closed (compaction, hooks, permissions, MCP client, streams, LSP, SSE)
• Sensor Fusion HUD: Sensor behaviour, polling feeds, weighted averaging, threshold alerts
• Distributed Swarm: CRDTs, node monitoring, topology election, libcluster + Tailscale strategy
• Self-Evolving Interface: Behavior tracking, layout recommendations, animated transitions
• AI Cockpit + Streaming: Multi-pane terminal dashboard, SSE streaming for 8 AI backends
• Virtual File System: Pure functional in-memory VFS, REPL helpers, 7 agent actions
• Phase 8: raxol_mcp: Extracted MCP package (server, client, registry, stdio/SSE transports)
• Phase 9: ToolProvider: Auto-derive MCP tools from widget tree (15 widgets), focus lens, tree walker
• Phase 10: MCP Protocol Hardening: Full MCP spec coverage (prompts, logging, completion), notifications, circuit breaker, chart ToolProviders, agent bridge, Tidewave removed
• Phase 11: MCP Resources + Context Tree: ResourceProvider behaviour, ResourceRouter, ContextTree, StructuredScreenshot, model projection diffs, resource subscriptions
• Phase 12: MCP Widget + Agent Coverage: @mcp_exclude attribute, FocusLens hover mode, ToolSynchronizer focus/hover tracking, HoverHighlight effect
• Phase 13: MCP Test Harness: Pipe-friendly test API (click, type_into, assert_component, assert_tool_available), session lifecycle, functor law property tests (10 properties)
• Agent Payments (Phase A): x402/MPP auto-pay, wallets (env + 1Password), spending controls, 5 agent actions
• Phase 14B: Xochi Integration: Xochi as default agent protocol for cross-chain (cash-positive, tier fees 0.10-0.40%). Riddler solves behind the scenes. Full intent flow: quote -> sign -> execute -> poll. Router prefers Xochi for cross-chain and privacy. 94 tests.
• Phase 14C: PXE-Bridge Integration: Aztec Private eXecution Environment as settlement target for high-trust privacy tiers. PXE client (JSON-RPC 2.0), schemas, PrivacyTier (Glass Cube, 6 tiers), Router settlement routing. 51 tests.
• Phase 14D: Stealth Settlement: Full ERC-5564/ERC-6538 in Xochi.Stealth (~300 LOC). ECDH derivation, view tag scanning (256x speedup), domain-separated key derivation, meta-address encode/decode. 44 tests (32 unit + 12 e2e).
• Phase 14E: ZKSAR + Trust Tiers: ZKSAR attestation verification (6 ZK proof types), diminishing-returns trust score aggregation, attestation-gated privacy tiers, privacy depth routing. 52 tests.
• Riddler Solver Wiring (ADR-0005): Xochi adapter wired to Riddler's intent engine: 9 endpoints, fee policy (5 tiers + privacy premiums), stealth/ERC-4337 settlement, ZKSAR attestation, EIP-712 typed data, PropertyTable stores. 119 Riddler tests + 347 raxol_payments tests. Aztec shielded settlement deferred (sidecar deployment).
• Phase 15A-E: Animation Hints: Declarative metadata flowing from view/1 to all surfaces. Animation.Helpers DSL (animate, stagger, sequence), terminal frames computed server-side via Animation.Framework, LiveView CSS via TerminalBridge.animation_css/1 with prefers-reduced-motion, MCP StructuredScreenshot JSON hints, reduced_motion in render context.
• Phase 17: Telegram bridge (raxol_telegram): Bot, per-chat session router (1000 max, 5s cooldown), inline keyboard navigation, message edit dedup, 10min idle timeout. 34 tests.
• Phase 18: Speech interface (raxol_speech): TTS (OsSay/espeak, accessibility subscription), STT (Whisper via Bumblebee), Listener (sox Port, bounded), 21 default voice commands. 28 tests.
• Phase 19: Watch notifications (raxol_watch): APNS/FCM push via pigeon, ETS-backed device registry, 1s debounce, 160-char glanceable summaries, tap-to-event actions. 34 tests.
• Adaptive UI Overhaul: 8-rule LayoutRecommender (was 3), multi-recommendation, TrendDetector, NxModel expanded to 10 features with auto-retrain, Lifecycle integration (adaptive: true), 5 MCP tools. 67 tests.
• BorderBeam Effect: Three-layer animated border glow (beam stroke, inner glow, outer bloom). Four variants (colorful/mono/ocean/sunset). Terminal exponential decay, LiveView conic-gradient + mask CSS, MCP hint serialization. Pipe + prop DSL. 99 effects tests.
• raxol_acp v0.2 (pre-alpha, 0.2.0-rc.0): First Elixir/OTP-native Virtuals Agent Commerce Protocol implementation, aligned to the deployed Base contracts and fork-validated. The v2 hook/event model is the active runtime (the v1 memo model was retired -- see packages/raxol_acp/MIGRATION_V2.md): Raxol.ACP.JobSession state machine (:open -> :budget_set -> :funded -> :submitted -> :completed plus :rejected/:expired), on-chain hook writes via HookClient -> AgenticCommerceV3 through an injected ProviderAdapter (SCA sponsored UserOps / JSONRPC EOA with NonceServer-serialized nonces / Mock), full ERC-4337 SCA wallet (Alchemy Modular Account v2: sponsored UserOps, counterfactual deploy, session keys), Seller stack (Backend.{InMemory, WebSocket} over Socket.IO + Queue + Runtime driving JobSession.Provider), mix raxol_acp.bench. Real ABIs and verified Base addresses vendored. 509 tests. Pre-alpha until a single graduated offering on Base mainnet.
• raxol_symphony Phases 0-14: Elixir/OTP port of OpenAI Symphony. Tracker-driven coding-agent orchestrator with two runner backends (raxol_agent default + codex app-server Port-based JSON-RPC), six surfaces (terminal/LiveView/MCP/Telegram/Watch/JSON API), workflow hot-reload, evidence framework (CI status + PR comments + cloc/SLOC + asciinema), in-run asciicast capture. 399 tests. Pre-alpha; path-dep.
• Video Render Target (hermes-agent pickup): One TEA module to MP4/WebM/GIF. Frames computed server-side at a virtual animation clock (Raxol.Animation.Clock), themed HTML via TerminalBridge rasterized through headless Chrome, encoded with FFmpeg. Raxol.Recording.Video (frame_html/render_frame/capture_frame/capture_clip), a swappable Rasterizer (zero-dep HeadlessChrome plus warm-pool ChromicPDF, ~6x faster), Headless.get_buffer/1, and mix raxol.render.
• Cross-Session Agent Memory (hermes-agent pickup): Opt-in, per-agent persistent memory behind a pluggable Raxol.Agent.Memory behaviour. Default Memory.Store.Ets is a self-contained ETS+DETS store (no deps) with BM25-lite + recency + tag ranking over a token inverted index. Automatic pre-turn recall injected into the system prompt (Manager.enrich_messages) plus explicit memory_remember/recall/forget actions; wired into Stream/ReAct and exposed via use Raxol.Agent memory_provider/0. 25 tests. User model, auto-capture, and the recall-MCP bridge deferred.

Next Up

Ship It

The twelve stable packages are published to Hex: the mature framework packages at v2.4.0 and the newer payment and surface packages (raxol_payments, raxol_speech, raxol_telegram, raxol_watch) at v0.1.0. raxol_acp, raxol_symphony, and raxol_gateway stay pre-alpha and unpublished until they graduate.

Targeting Next: Agent & Creative

Capabilities identified by studying NousResearch/hermes-agent. The BEAM gives several of these a more natural home than Hermes's single-process Python: real concurrency, crash isolation, and a native frame engine. Shipped from this set: the Video Render Target and Cross-Session Agent Memory (see What's Done). Remaining:

AI Backend Providers

Supported now:

• Mock (default): instant offline demo, no API key
• Proton Lumo (PROTON_UID=... PROTON_ACCESS_TOKEN=...): zero-access encrypted AI, full U2L encryption via Backend.Lumo
• Proton Lumo via lumo-tamer (LUMO_TAMER_URL=http://localhost:3000): OpenAI-compatible proxy fallback
• Kimi K2.5 (KIMI_API_KEY=...): Moonshot AI, $0.60/M input, 256K context, named :kimi provider
• LLM7.io (FREE_AI=true): free, OpenAI-compatible, no key needed, 40 req/min
• Ollama (OLLAMA_MODEL=...): free local inference, OpenAI-compatible
• Groq (AI_API_KEY=... AI_BASE_URL=https://api.groq.com/openai): fast free tier
• OpenAI (AI_API_KEY=...): GPT-4o-mini and up
• Anthropic (ANTHROPIC_API_KEY=...): Claude Haiku/Sonnet/Opus
• OpenRouter (:openrouter harness, key via ExecutorConfig auth): OpenAI-compatible aggregator that sends app-attribution headers (HTTP-Referer, X-OpenRouter-Title, X-OpenRouter-Categories) so Raxol appears on openrouter.ai/rankings

Longer Term

• Multi-node cockpit (swarm coordination across physical terminals)
• Plugin marketplace
• VS Code extension for component previews
• Burrito packaging (single standalone binary)
• FLAME integration (elastic compute for Nx training bursts)
• SSH session multiplexing (tmux-like panes)
• Collaborative sessions (multi-user terminal sharing)
• Documentation site (widget gallery, tutorials, examples)

Contributing

Want to help? See CONTRIBUTING.md.

Versioning

• Minor (2.x.0): New features, framework additions
• Patch (2.0.x): Bug fixes, performance improvements
• Major (3.0.0): Breaking API changes, architectural shifts