Skip to content

roadmap

Jump to bottom
Matt Sutton edited this page Mar 15, 2026 · 2 revisions

Totem Strategic Roadmap

Totem is evolving from a local memory database into the Air-Gapped Codebase Immune System—a zero-telemetry infrastructure layer that prevents AI coding agents from violating architectural boundaries.

This roadmap outlines the path to v1.0 and beyond, broken down into sequential phases that prioritize Developer Experience (DX), enterprise sandboxing, and continuous governance.

🟢 Phase 1 & 2: The Foundation (Completed)

We established the core primitives required for a local-first memory engine.

  • Local Vector Engine: Embedded LanceDB chunking and retrieval.
  • MCP Interface: Standardized search_knowledge for Claude, Gemini, and Junie.
  • MVC Tiers: "Minimum Viable Configuration" scaling from Lite to Full.
  • Adversarial Hardening: ANSI terminal injection defense and XML delimiting.
  • Saga-Based Syncs: Transactional checkpoints for index updates.

🟡 Phase 3: The DX & Reliability Engine (Active)

Current Focus: Stop the friction. Make writing and enforcing rules safe, fast, and foolproof.

  • The Rule Simulator (totem test) & Compilation Guard: Gamifying rule creation. Developers test their governance rules against local pass.ts/fail.ts fixtures. Totem refuses to deploy unproven rules, preventing regex from breaking CI.
  • Hard Real-Time Load Shedding: Enforcing the 3-second budget. If totem shield takes longer than 2.5s locally, it immediately aborts and fails open, guaranteeing the developer's terminal never hangs.
  • Local Diagnostics (totem doctor): Automated health checks that scan for config bloat, missing git hooks, and leaked secrets.
  • Consumer Init Rewrite: Dynamic agent detection to seamlessly auto-configure the exploding ecosystem of agents (Claude, Cursor, Copilot, Junie, Cline) without interactive friction.

🔵 Phase 4: Enterprise Sandboxing & Boundaries (Upcoming)

Proving Totem is the only viable solution for complex, high-compliance environments.

  • Multi-Totem Domains: Running parallel, isolated MCP servers (e.g., separating public monorepo code from private strategy submodules) to prevent context pollution.
  • Hierarchical Exclusions (.totemignore): Providing granular, git-style control over what gets indexed and shielded, replacing flat config arrays for massive monorepos.
  • Severity Levels & SARIF: Introducing error (blocking) and warning (advisory) levels to shield rules, preventing bot-generated PRs from failing CI over minor formatting issues. Native SARIF 2.1.0 output integration.

🟣 Phase 5: Continuous Governance (Post-v1.0)

Moving from static rules to a measurable, self-cleaning immune system.

  • Rule Lifecycle Management: Tracking local telemetry (fire_count, suppression_count) to automatically flag rotting, high-noise rules before they annoy developers.
  • Continuous Compliance Signal: Passively tracking agent behavior logs to warn teams when an agent's instruction adherence drops below an 80% threshold (e.g., failing to call search_knowledge before coding).

🛸 The "Dream Within the Dream" (Future)

Federated organizational memory and swarm intelligence.

  • Federated Memory (The Mother Brain Pattern): Allowing local Totem instances to securely pull cryptographically signed, enterprise-wide architectural lessons into new projects instantly.
  • Auto-Discovery Mesh Networking: Totems automatically discovering and wiring themselves to other upstream Totems across enterprise networks.

Clone this wiki locally