ROADMAP.md

Aubm Roadmap

This document outlines the strategic evolution of Aubm, moving from a robust orchestration core to an enterprise-ready multi-agent operating layer.

Phase 1: Core Foundation (Completed)

• Autonomous Agent Execution: Multi-provider support (OpenAI, Groq, Gemini, etc.).
• Project Orchestration: Intelligent task scheduling and dependency management (DAG).
• Human-in-the-Loop: Approval and rejection workflows for agent outputs.
• Semantic RAG: Contextual memory injection across project tasks.
• Real-time Logs: Streaming agent thoughts and actions via SSE.
• Cost Control: Token-based budgeting and execution blocking.

Phase 2: Advanced Collaboration and Tools (Completed)

• Multi-Agent Debates: Allow agents to cross-verify each other's outputs before human review.
• Extended Toolbelt:
  • Web Browser Tool (via Playwright) for live data fetching.
  • Code Sandbox for executing and testing generated snippets.
  • File Generation (Excel, Word, and advanced PDF layouts).
• Collaborative Editing: Real-time collaborative output refining for humans.
• Mobile Experience: Capacitor-based mobile app for project monitoring (initialized).

Phase 3: Intelligence and Scale (Completed)

• Fine-tuning Loop: Feedback loop (Like/Dislike) implemented for data collection.
• Recursive Project Decomposition: Agents that can spawn sub-tasks and manage them.
• Enterprise Security:
  • SSO Integration (Google, GitHub via Supabase).
  • Advanced RLS for granular team permissions.
  • Audit logs for every LLM interaction.
• Agent Marketplace: Community-driven agent templates and specialized skill sets.

Phase 4: Autonomy and Beyond (Completed)

• Self-Healing Infrastructure: Agents that can monitor health and apply safe patches.
• Voice Interaction: Control navigation and hear project/task status updates via browser voice APIs.
• VR/AR Dashboard: Spatial DAG viewer scaffold for layered project/task visualization.

Phase 5: Production Operations (Completed)

• Operations Monitoring: Backend health summary endpoint and frontend monitoring dashboard with Supabase fallback.
• Deployment Hardening: Dockerized backend/runtime profile and production CORS configuration.
• Error Tracking: Sentry-compatible error reporting hooks for backend and frontend.
• Performance Budgeting: Frontend code splitting and bundle-size targets.

Phase 6: Distributed Scale and Intelligence (In Progress)

• Recursive Project Decomposition: Agents that can automatically break down goals.
• Numerical Consistency (Semantic Backprop): Enforce absolute figures across tasks.
• Visual Tooling: Integrated support for charts and AI illustrations.
• Vercel Deployment: Monorepo serverless configuration.
• Heuristic Output Guardrails: Prompt hardening, reviewer checks, and final-report filtering for placeholders, unsupported claims, and low-quality sections.
• Asynchronous Task Queue: Dedicated background workers (worker.py).
• Vectorized Long-term Memory: Cross-project semantic retrieval.
• Self-Optimizing Agents: Meta-prompting loops based on human feedback.

Phase 7: Structured Evidence and Entity Integrity (Next)

• Strict JSON Task Schemas: Enforce structured outputs per task type instead of free-form text.
• Mandatory source_url per Claim: Require evidence links for competitor, pricing, release, benchmark, and market claims.
• Entity Normalization Layer: Canonicalize entity names, merge aliases, and separate direct competitors from adjacent tools before final reporting.
• Semantic Deduplication: Collapse equivalent claims written differently across tasks.
• Evidence-Aware Final Report: Build the final report from normalized entities and validated claims only.

---

Last updated: May 6, 2026