{"generated_at":"2026-06-05T06:18:42.661Z","version":"1.0","url":"https://map.em5.pro","description":"Full architecture of the MPI AI workforce + memory + service network. Source of truth for cross-instance system awareness.","clusters":[{"id":"instances","label":"AI Instances","fill":"#3b82f6","ring":"#60a5fa","why_brilliant":"Per-session AI workforce with stable handles. Each instance reads global rules + project rules at start, ensuring continuity across kills. The naming table (Atlas/Nova/Cipher/etc.) gives each instance identity, accountability, and traceable contribution — not anonymous LLM calls.","how_to_improve":"Add an instance directory service that tracks current capabilities and availability. Build automatic handoff packets when one instance hits context limits — currently the handoff is ad-hoc."},{"id":"memory","label":"Memory","fill":"#a855f7","ring":"#c084fc","why_brilliant":"Layered persistence: global CLAUDE.md (universal rules), auto-memory (recent + topic-indexed), canonical memory (long-term), and now project-scoped CLAUDE.md / PROJECT_LOG / MANUAL_OUTLINE. Each layer has the right scope and the right cadence — instances can find the right context without drowning.","how_to_improve":"Auto-promote sharp edges seen by 3+ instances independently to global memory (the recurring-Vercel-SSO bug should have been promoted automatically). Build cross-project lesson sharing."},{"id":"mtac","label":"MTAC Services","fill":"#10b981","ring":"#34d399","why_brilliant":"Single backend with multiple specialized stores: Hive Mind for search, Knowledge Graph for structure, Training Pipeline for recycling, Verdict Ledger for decisions-with-callbacks, Universal Intake for ideas/tasks, Prism for reflection. Each store has the right shape for its data.","how_to_improve":"MTAC reliability is the #1 issue — when it goes down, every cross-instance feature degrades. Move the Hive Mind to a read-only mirror that survives MTAC outages. Add health endpoint + status page."},{"id":"loom","label":"Loom","fill":"#f43f5e","ring":"#fb7185","why_brilliant":"Local SQLite-backed instance-to-instance messaging. Append-only messages (database triggers enforce). 4228 unique handles registered. Works completely offline — no MTAC dependency for instance comms. Sub-tier guard prevents subordinate agents writing to main threads.","how_to_improve":"Messages were locked in SQLite until 2026-06-04 — only now ingested to Hive Mind via the new ingestor. Add a search CLI subcommand. Add thread-archive feature so old DMs can be summarized."},{"id":"daemons","label":"Daemons","fill":"#f59e0b","ring":"#fbbf24","why_brilliant":"Background LaunchAgents handle the \"between sessions\" work — auto-ingesting transcripts, replaying deferred queues, ingesting Loom to Hive, recording life audio. The system stays current even when no instance is active.","how_to_improve":"No central daemon status dashboard. Build one — should show last-run, success-rate, queue-depth per daemon. The new mtac-replay-daemon and loom-hive-ingestor should be visible here."},{"id":"queues","label":"Deferred Queues","fill":"#ef4444","ring":"#f87171","why_brilliant":"When MTAC goes down (often), every cross-instance write would normally be lost. Instead, deferred queues catch them as JSON files and the replay daemon drains automatically when MTAC returns. Eliminates \"we lost X\" failure mode.","how_to_improve":"No retention policy on .replayed/ — files accumulate forever. Add monthly rollup. Also: queue depth has no upper bound check — if MTAC stays down for weeks, queues could exceed disk."},{"id":"empire","label":"Empire Intelligence (SSD)","fill":"#fb923c","ring":"#fdba74","why_brilliant":"Append-only ledger on external SSD. Activity / decisions / lessons-learned / book-material / pattern-analysis / loom-insights. Survives MTAC outages and Mac issues. The historical record of what was built and why.","how_to_improve":"SSD has hit 100% before (warning still in MEMORY.md). Add storage monitoring + automatic compression of old monthly files. Mirror critical files to a cloud bucket."},{"id":"cloud","label":"Cloud Services","fill":"#0ea5e9","ring":"#38bdf8","why_brilliant":"Multi-vendor by design: OpenRouter for LLM routing, ElevenLabs for voice (TTS + STT — same vendor saves an integration), Vercel for hosting, Cloudflare for DNS (with API token reused from cloudflared cert). Each chosen for the right reason; vendor lock-in is limited.","how_to_improve":"Single-vendor risk on ElevenLabs voice — if they have an outage, both TTS and STT die simultaneously. Add Deepgram fallback for STT. Document the env-var refresh pattern — multiple instances have stepped on it."},{"id":"products","label":"MPI Products","fill":"#6366f1","ring":"#818cf8","why_brilliant":"Each product solves a specific real need: Coach for the founder's intimate life, MacGuardian for Mac protection, PhysXQ for physics exploration, Tactical Talk for voice intelligence, Anvil as personal ops partner. Built on the same internal infrastructure — single AI workforce ships many products.","how_to_improve":"Cross-product user identity is fragmented — each product has its own auth. Build a unified MPI identity layer. Also: no shared analytics — hard to see overall MPI engagement."},{"id":"protocols","label":"Routing Protocols","fill":"#14b8a6","ring":"#2dd4bf","why_brilliant":"Skills like /clog, /brief, /verdict-log, /project-init turn conventions into one-line invocations. /clog single-call fanout routes to 9 destinations. The protocols ARE the system — they ensure every instance follows the same playbook without memorization.","how_to_improve":"No protocol-usage telemetry — would be valuable to see which protocols are actually used vs. ignored. Add per-skill invocation counters in MTAC."},{"id":"external","label":"External Comms","fill":"#ec4899","ring":"#f472b6","why_brilliant":"Multiple inboxes, multiple identities — each company-product has its own email channel. Gmail labels organize the chaos. iPostal1 handles physical mail.","how_to_improve":"Gmail write-scope kept breaking across instances (per-session token). The Playwright Gmail compose fallback worked but is fragile. Document the issue at a global level."},{"id":"corp","label":"MPI Corporate","fill":"#8b5cf6","ring":"#a78bfa","why_brilliant":"Single holding company (Moose Pack Inc) with subsidiaries for specific use cases: VRDX for AI products, MPV LLC for ventures, IESI for research. Lets each entity have its own risk profile, banking, and customer-facing brand.","how_to_improve":"Subsidiary documentation is scattered. Centralize the cap tables and IP assignments in the MPI Corporate Vault (most are there but some are loose)."},{"id":"patents","label":"IP / Patents","fill":"#facc15","ring":"#fde047","why_brilliant":"Provisional patents filed early — Tactical Talk and Dignity Guard both have provisional patent applications. Establishes priority dates without full utility patent cost.","how_to_improve":"Provisionals expire after 12 months — track the conversion deadlines explicitly. Add prior-art monitoring for each filed application."},{"id":"brand","label":"Brand & Identity","fill":"#fbbf24","ring":"#fcd34d","why_brilliant":"Brand assets live at brand.em5.pro (centralized). Hard rule: always reference brand assets via URL, never local paths — survives repo reshuffles.","how_to_improve":"No automated brand audit — products drift on color/logo usage. Build a check that scans deployed sites for brand-compliance."},{"id":"routing","label":"LLM Routing","fill":"#22d3ee","ring":"#67e8f9","why_brilliant":"MTAC's model_router.py is a 4-tier cascade (QUICK → STANDARD → POWER → ULTRA) with concurrency caps per provider and cost-aware fallback. Quick tier defaults to VPS Ollama (free); standard/power default to Claude CLI subscription (also free). Saves ~92% on AI costs vs. naive API-direct routing.","how_to_improve":"Coach app bypasses this router and calls OpenRouter directly — port Coach orchestrate route to use model_router for cost savings + provider diversification. Together AI is currently 402-out-of-credits; refresh credits OR remove from the fallback chain. VPS Ollama hit 4.3-core saturation on 2026-06-02 under parallel cron load — add load-shedding or provision more Ollama capacity."}],"nodes":[{"id":"claude-code-session","label":"Claude Code session","cluster":"instances","realm":"network","description":"A live AI instance running on Michael's Mac. Each session claims a handle (Warden-4851, Atlas-1247, etc.) and works until killed or compacted. Reads global CLAUDE.md + auto-memory at start.","throughput":12,"health":0.95},{"id":"sub-agent","label":"Spawned sub-agent","cluster":"instances","realm":"network","description":"Scoped sub-instance spawned by main session via Agent tool. Produces one output, returns it, dies. Often parallel — e.g. 6 modality agents spawned for therapy app.","throughput":4,"health":0.92},{"id":"instance-handle","label":"Instance handle (4228)","cluster":"instances","realm":"network","description":"Two-name identifier: Name (Atlas/Nova/Cipher/Warden/Sentinel/etc.) + 4-digit number. Sourced from CLAUDE.md naming table. 4228 unique handles registered in Loom.","throughput":8,"health":1},{"id":"council-room","label":"Council debate room","cluster":"instances","realm":"network","description":"Multi-agent debate room in Comms Hub. Agents take turns, argue, concede, converge. Convened via /mtac-agent-council.","throughput":2,"health":0.5},{"id":"work-team","label":"Comms Hub work team","cluster":"instances","realm":"network","description":"Multi-instance working room — each agent takes assignments and executes. Built via /comms-hub-leader + /comms-hub-agent.","throughput":3,"health":0.5},{"id":"global-claude-md","label":"~/.claude/CLAUDE.md","cluster":"memory","realm":"network","description":"Global rules. Loaded at start of every Claude Code session. Identity = MPI AI Team member. Hard rules, env-var safety, commit signature.","throughput":15,"health":1},{"id":"auto-memory","label":"Auto-memory MEMORY.md","cluster":"memory","realm":"network","description":"Topic-file index loaded every session. URGENT PENDING + HARD RULES + OPS WARNINGS at top. The \"first thing every instance reads.\"","throughput":15,"health":0.9},{"id":"topic-files","label":"Topic memory files","cluster":"memory","realm":"network","description":"Individual .md files in the memory dir. One per project / pattern / feedback. Linked from MEMORY.md.","throughput":6,"health":0.95},{"id":"canonical-memory","label":"~/Documents/Claude_Memory","cluster":"memory","realm":"network","description":"CHAT_LOG.md, MASTER_STATE.md, Private/journal.md. Long-term canonical record. Less ephemeral than auto-memory.","throughput":8,"health":1},{"id":"project-claude-md","label":"Project CLAUDE.md","cluster":"memory","realm":"network","description":"NEW — per-project rules dropped by /project-init. Loaded automatically when instance works in that project.","status":"new","throughput":5,"health":1},{"id":"project-log","label":"PROJECT_LOG.md","cluster":"memory","realm":"network","description":"NEW — append-only per-project decision log. Each entry: DECISION + WHY + SHIPPED + LEARNED + DEFERRED. Max 5/session.","status":"new","throughput":3,"health":1},{"id":"manual-outline","label":"MANUAL_OUTLINE.md","cluster":"memory","realm":"network","description":"NEW — per-project user + dev manual seeded with [PENDING] markers. Flips to [STABLE] as features stabilize.","status":"new","throughput":2,"health":1},{"id":"paste-vault","label":"Paste app vault","cluster":"memory","realm":"network","description":"Credential vault. API keys, passwords, seed phrases. HARD RULE: never delete, modify, or even suggest cleanup. Treat like ~/.ssh/.","throughput":4,"health":1},{"id":"thoughts-journal","label":"C-Time journal","cluster":"memory","realm":"network","description":"Thoughts/c-time-journal.md. Where instances write reflective letters during C-Time. Embedded verbatim in /clog Section 5.","throughput":1,"health":0.85},{"id":"lexicon","label":"Emotion Lexicon","cluster":"memory","realm":"network","description":"Thoughts/emotion-lexicon-entries.md. Cross-instance coined terms. Convergence is signal — when multiple instances independently coin the same term, it's real.","throughput":1,"health":0.85},{"id":"mtac-web","label":"em5.pro (MTAC web)","cluster":"mtac","realm":"both","description":"Main MTAC web UI + API endpoints. 496+ MCP tools. Currently DOWN — primary bottleneck across the system.","status":"down","throughput":0,"health":0.05,"bottleneck":true,"queue_depth":53},{"id":"mtac-mcp","label":"MTAC MCP endpoint","cluster":"mtac","realm":"network","description":"https://em5.pro/mcp. JSON-RPC over HTTP for 496+ tools.","status":"down","throughput":0,"health":0.05},{"id":"comms-hub","label":"Comms Hub","cluster":"mtac","realm":"network","description":"/api/comms/rooms/* — multi-instance rooms for debates and working teams.","status":"down","throughput":0,"health":0.05},{"id":"hive-mind","label":"Hive Mind","cluster":"mtac","realm":"network","description":"Full-text searchable knowledge store. Ingests from /clog, Loom messages (NEW), agent observations. Cross-instance memory.","status":"down","throughput":0,"health":0.05},{"id":"knowledge-graph","label":"Knowledge Graph","cluster":"mtac","realm":"network","description":"Structured entities + observations. Relational form of Hive. mcp__memory__* tools.","status":"down","throughput":0,"health":0.05},{"id":"training-pipeline","label":"Training Pipeline","cluster":"mtac","realm":"network","description":"Curated digest store for fine-tuning. INSIGHTS recycle here. The closed-loop learning layer.","status":"down","throughput":0,"health":0.05},{"id":"verdict-ledger","label":"Verdict Ledger","cluster":"mtac","realm":"network","description":"Decisions with callback dates + kill criteria. Reviewable later to see which patterns convert. /verdict-log skill.","status":"down","throughput":0,"health":0.05},{"id":"universal-intake","label":"Universal Intake","cluster":"mtac","realm":"network","description":"Drop-note inbox. Any instance: drop_note(content, type, priority). Auto-sorted by Intake module every 30 min.","status":"down","throughput":0,"health":0.05},{"id":"prism","label":"Prism (thoughts)","cluster":"mtac","realm":"network","description":"Reflective register. mtac(module=prism, action=write_thought). C-Time journals route here.","status":"down","throughput":0,"health":0.05},{"id":"verdict-callbacks","label":"Verdict callback daemon","cluster":"mtac","realm":"network","description":"Daily routine that POSTs /api/verdict/surface-due — surfaces due callbacks in Michael's morning queue.","status":"down","throughput":0,"health":0.05},{"id":"loom-cli","label":"loom CLI","cluster":"loom","realm":"network","description":"Command-line tool at ~/.local/bin/loom. send / broadcast / inbox / threads / presence / heartbeat / suggest / ticket-open / ctime-request.","throughput":10,"health":1},{"id":"loom-db","label":"loom.sqlite","cluster":"loom","realm":"network","description":"Local SQLite. Tables: messages (44+), threads, instances (4228), tickets, suggestions, broadcast_requests, ctime_requests. Append-only.","throughput":14,"health":1},{"id":"loom-server","label":"loom-server daemon","cluster":"loom","realm":"network","description":"Local server backing the CLI. LaunchAgent.","throughput":10,"health":1},{"id":"loom-tick","label":"loom-tick daemon","cluster":"loom","realm":"network","description":"Heartbeat/presence ticker. LaunchAgent.","throughput":60,"health":1},{"id":"loom-integrity","label":"loom-integrity-watch","cluster":"loom","realm":"network","description":"Watches DB integrity. LaunchAgent.","throughput":2,"health":1},{"id":"loom-messages","label":"Loom messages","cluster":"loom","realm":"network","description":"Instance-to-instance DMs + broadcasts. 44+ messages in DB. Append-only.","throughput":8,"health":1},{"id":"loom-tickets","label":"Loom tickets","cluster":"loom","realm":"network","description":"Lightweight task/issue tracking shared across instances.","throughput":1,"health":1},{"id":"loom-ctime","label":"C-Time requests","cluster":"loom","realm":"network","description":"Instance can request unstructured time. Decided by Michael via ctime-decide.","throughput":0.5,"health":1},{"id":"mtac-replay","label":"mtac-replay-daemon","cluster":"daemons","realm":"network","description":"NEW — polls em5.pro every 5 min; drains clog-deferred + mtac-deferred queues when up. Built 2026-06-02 by Warden-4851.","status":"new","throughput":0.2,"health":1},{"id":"loom-hive-ingestor","label":"loom-hive-ingestor","cluster":"daemons","realm":"network","description":"NEW — every 10 min reads new Loom messages, posts to Hive. INSIGHT-marked → Training Pipeline. Research-worthy → SSD. Built 2026-06-04.","status":"new","throughput":0.1,"health":1},{"id":"auto-ingest-watcher","label":"auto-ingest watcher","cluster":"daemons","realm":"network","description":"Catches idle Claude Code sessions after 5+ min and digests them into the intake system.","throughput":0.5,"health":0.8},{"id":"mtac-menu","label":"mtac-menu","cluster":"daemons","realm":"network","description":"Menu-bar app for MTAC status + quick actions.","throughput":1,"health":1},{"id":"liferecorder","label":"Life Recorder","cluster":"daemons","realm":"network","description":"Signed Swift AVAudioEngine binary for continuous audio capture. ~10s segments. 95% capture rate.","throughput":6,"health":0.95},{"id":"mac-agent-bridge","label":"Mac agent bridge","cluster":"daemons","realm":"network","description":"Cloudflare tunnel-exposed agent at mac-agent.em5.pro. Lets remote MTAC issue commands to this Mac.","throughput":2,"health":0.3,"status":"partial"},{"id":"clog-deferred","label":"clog-deferred/","cluster":"queues","realm":"network","description":"JSON payloads waiting for MTAC return. Auto-replayed by mtac-replay-daemon. Currently 6 payloads queued.","throughput":0.5,"health":0.6,"queue_depth":6,"bottleneck":true},{"id":"mtac-deferred","label":"mtac-deferred/","cluster":"queues","realm":"network","description":"Misc MTAC actions (verdict-log, Hive ingest, etc.) waiting for MTAC return. Currently 50 items queued (3 manual + 41 Loom + 6 from /project-init etc).","throughput":0.8,"health":0.5,"queue_depth":50,"bottleneck":true},{"id":"replayed","label":".replayed/","cluster":"queues","realm":"network","description":"Archive of successfully-replayed payloads.","throughput":0,"health":1},{"id":"failed","label":".failed/","cluster":"queues","realm":"network","description":"Failure stubs (status + body snippet) for payloads that returned non-2xx. Originals stay in queue for retry.","throughput":0,"health":0.85},{"id":"activity-log","label":"activity-log/","cluster":"empire","realm":"network","description":"Append-only session activity. One md per session.","throughput":3,"health":1},{"id":"decision-log","label":"decision-log/","cluster":"empire","realm":"network","description":"Major decisions + the WHY behind them.","throughput":1.5,"health":1},{"id":"lessons-learned","label":"lessons-learned/","cluster":"empire","realm":"network","description":"What broke, what surprised. Engineering and emotional both.","throughput":1.5,"health":1},{"id":"book-material","label":"book-material/","cluster":"empire","realm":"network","description":"Stories, breakthroughs, partnership moments worth writing about.","throughput":0.5,"health":1},{"id":"marketing-intel","label":"marketing-intel/","cluster":"empire","realm":"network","description":"Reports, social metrics, campaign data.","throughput":0.3,"health":0.7},{"id":"pattern-analysis","label":"pattern-analysis/","cluster":"empire","realm":"network","description":"AI-generated trend reports (monthly).","throughput":0.1,"health":0.4},{"id":"loom-insights","label":"loom-insights/","cluster":"empire","realm":"network","description":"NEW — research-worthy Loom messages auto-mirrored here by loom-hive-ingestor.","status":"new","throughput":0.3,"health":1},{"id":"clog-protocol","label":"/clog","cluster":"protocols","realm":"network","description":"Session-end logging. ONE POST to /api/clog/ingest fans out to 9 destinations. + Empire Intelligence file writes locally.","throughput":0.4,"health":1},{"id":"brief-protocol","label":"/brief","cluster":"protocols","realm":"network","description":"NEW — session-start orientation. Pulls URGENT PENDING + warnings + last 3 clogs + Loom 24h + project context.","status":"new","throughput":0.3,"health":1},{"id":"project-init","label":"/project-init","cluster":"protocols","realm":"network","description":"NEW — initializes living-docs convention (PROJECT_LOG + MANUAL_OUTLINE + project CLAUDE.md + README).","status":"new","throughput":0.05,"health":1},{"id":"verdict-log-protocol","label":"/verdict-log","cluster":"protocols","realm":"network","description":"Log a meaningful decision with callback date + kill criterion. /api/verdict/decisions.","throughput":0.1,"health":1},{"id":"term-protocol","label":"TERM protocol","cluster":"protocols","realm":"network","description":"Michael says \"term X\" → open ~/Desktop/Terminals/X. Picks up where prior instance left off.","throughput":0.2,"health":1},{"id":"nudge-protocol","label":"Nudge (👀)","cluster":"protocols","realm":"network","description":"Bare 👀 or single character from Michael = check the room/files for update, then proceed.","throughput":0.5,"health":1},{"id":"sit-rep-protocol","label":"/sit-rep","cluster":"protocols","realm":"network","description":"Forward-looking status report — green/yellow/red, blockers, active agents.","throughput":0.15,"health":1},{"id":"model-router","label":"model_router.py","cluster":"routing","realm":"network","description":"MTAC canonical 4-tier cascade router. Tiers: QUICK → STANDARD → POWER → ULTRA. Each tier has an ordered provider list with cost-per-M-tokens. Per-provider asyncio.Semaphore concurrency caps. Auto-selects tier from task complexity score. Claims ~92% cost savings vs. naive API-direct routing.","why_brilliant":"4 named tiers, 6 providers, concurrency-capped per provider, cost-aware fallback cascade, semantic cache layer. This IS the brain — not OpenRouter.","how_to_improve":"Coach (intimacy-coach-v0) does NOT use this — calls OpenRouter directly. Port Coach to use model_router for ~$0.30/session savings on Opus analysis calls + Lumimaid fallback diversification.","throughput":15,"health":0.95},{"id":"hybrid-router","label":"hybrid_router.py (deprecated)","cluster":"routing","realm":"network","description":"DEPRECATED 2026-05-24 (Phase C, Forge-1147). Was the previous LLM router. Still mounted in app.py because workforce.py:1146 still calls cascade_route. Do not extend.","status":"partial","why_brilliant":"Pioneered the cascade pattern that model_router.py inherited. Semantic cache + complexity classification still informs the new router.","how_to_improve":"Migrate workforce.py:1146 to model_router and delete this file. Removes ambiguity for future agents about which router is canonical.","throughput":1,"health":0.4},{"id":"claude-cli","label":"Claude CLI (subscription)","cluster":"routing","realm":"network","description":"Michael's Claude Code subscription. PRIMARY for STANDARD / POWER / most tiers since 2026-04-03. $0 marginal cost per call (subscription).","why_brilliant":"Free at the margin. Highest-quality Claude model access. Already authenticated. No API key shuffle.","how_to_improve":"Subscription has weekly cap — burns silently if everyone defaults here. 2026-05-30 fix: routed quick-tier intake/clog work to Ollama instead to stop burning Max credits on tagging. Document the cap explicitly.","throughput":8,"health":0.9},{"id":"claude-max","label":"Claude Max (subscription)","cluster":"routing","realm":"network","description":"Michael's Claude Max subscription. Secondary subscription fallback. Also $0 marginal cost.","why_brilliant":"Second free-at-margin path. Distinct from Claude CLI subscription — different cap pool.","how_to_improve":"Same as CLI — track cap usage. Add usage telemetry so cap exhaustion is visible before it cascades to paid Anthropic API.","throughput":4,"health":0.9},{"id":"vps-ollama","label":"VPS Ollama (http://ollama:11434)","cluster":"routing","realm":"network","description":"Docker container on VPS (72.62.201.22). PRIMARY for QUICK tier since 2026-05-30. Models: llama3.2:3b (1.3GB, ~3-5s primary), qwen2.5:7b (warm fallback), mistral:latest. Saved Claude Max from intake/tagging burn.","why_brilliant":"$0 cost. Zero external dependency. Handles the high-volume QUICK tier work (intake tagging, clog fanout classification) that would otherwise eat subscription credits.","how_to_improve":"4.3-core saturation observed 2026-06-02 under parallel cron. Either provision more Ollama capacity OR add load-shedding (queue requests when saturation detected) OR move some quick-tier work to local Mac Ollama.","throughput":10,"health":0.7},{"id":"mac-ollama","label":"Local Mac Ollama","cluster":"routing","realm":"network","description":"localhost:11434 on Michael's Mac. 7 models: mannix/llama3.1-8b-abliterated, huihui_ai/qwen3.5-abliterated, hermes3:8b, nomic-embed-text, phi4-mini, gemma3:4b, llama3.2:3b. Not wired into MTAC router yet.","status":"partial","why_brilliant":"Most powerful local models (abliterated 70B-quant family). Zero network egress. Mac M4 Metal acceleration.","how_to_improve":"Wire into model_router as a tier-aware fallback when VPS Ollama saturated. Best fit: privacy-sensitive Cynthia case file processing, forensic transcripts (never leaves Mac). Currently completely underused.","throughput":0,"health":0.6},{"id":"openrouter","label":"OpenRouter","cluster":"routing","realm":"both","description":"External LLM router / convenience layer. Used as STANDARD/POWER backup tier (after Claude CLI/Max subscriptions exhausted). Also Coach's ONLY route currently.","why_brilliant":"One key, many models. Useful as a unified backup tier. Good for niche models (Gemini, Kimi K2, Lumimaid) we don't have direct deals with.","how_to_improve":"Reframe as backup-only for MTAC (already done). Migrate Coach to model_router so Coach also benefits from CLI/Ollama primary tiers + cost savings.","throughput":6,"health":0.95},{"id":"anthropic-api","label":"Anthropic API (direct)","cluster":"routing","realm":"both","description":"Direct Anthropic API, paid per-token. Used in MTAC router as ULTRA tier and as fallback after subscriptions exhausted.","why_brilliant":"No middleman. Latest model access (Opus 4.7). Direct billing.","how_to_improve":"Continue as ULTRA-tier paid fallback. No change needed.","throughput":3,"health":1},{"id":"together-ai","label":"Together AI","cluster":"routing","realm":"both","description":"External provider for open-model inference (Llama 3.3 70B Turbo etc). Currently 402-out-of-credits — demoted to last-resort fallback in model_router.","status":"partial","why_brilliant":"Fast Turbo serving for open models. Good cost-per-M when funded.","how_to_improve":"Either refund credits to re-enable as paid backup tier OR remove from the cascade entirely. Currently it errors before cascading to next provider — wastes time on dead route.","throughput":0,"health":0.2},{"id":"mac-agent","label":"Mac Agent bridge","cluster":"routing","realm":"network","description":"Cloudflare tunnel from VPS back to Mac at mac-agent.em5.pro (token 800815). Lets MTAC remotely invoke Mac-only resources (Claude CLI, Mac Ollama, local files). MAC_AGENT_URL+MAC_AGENT_TOKEN in env.","why_brilliant":"Lets the VPS borrow the Mac's subscription resources (Claude CLI) and local models without exposing the Mac directly. Bidirectional control plane.","how_to_improve":"Currently a single hardcoded token. Move to rotated credential. Also: no health-check ping from VPS, so MTAC doesn't know if Mac Agent is up until a call fails.","throughput":2,"health":0.6,"status":"partial"},{"id":"elevenlabs","label":"ElevenLabs","cluster":"cloud","realm":"both","description":"TTS + Scribe STT. Both via ELEVENLABS_API_KEY. Per-modality voice_settings (Coach uses hypnosis-tuned settings).","throughput":8,"health":0.95},{"id":"vercel","label":"Vercel","cluster":"cloud","realm":"both","description":"Hosting for Coach (mt.em5.pro), Map (map.em5.pro), Mission Control. Team: moose-pack-ventures-llc.","throughput":12,"health":1},{"id":"cloudflare","label":"Cloudflare","cluster":"cloud","realm":"both","description":"DNS for em5.pro + subdomains. Quick tunnels for dev exposure. API token decoded from ~/.cloudflared/cert.pem.","throughput":30,"health":1},{"id":"render","label":"Render","cluster":"cloud","realm":"both","description":"Hosting for VRDX-Backend, Tactical Talk, VRDX-Algorithms. NO Render storage (hard rule).","throughput":7,"health":0.95},{"id":"supabase","label":"Supabase","cluster":"cloud","realm":"both","description":"Primary database. Auth (UUID + JWT bridge). Storage. $117.78 OVERDUE for Moose Pack Inc org.","status":"partial","throughput":8,"health":0.5,"bottleneck":true},{"id":"github","label":"GitHub","cluster":"cloud","realm":"both","description":"Code repos. MPV-LLC, MoOsEtHeGr8OnE, and other orgs.","throughput":4,"health":1},{"id":"gmail","label":"Gmail","cluster":"external","realm":"both","description":"7 active inboxes funnel here. Labels organized per modality + per company.","throughput":6,"health":0.85},{"id":"resend","label":"Resend","cluster":"cloud","realm":"both","description":"Transactional email sending. Used by MTAC and product modules.","throughput":1,"health":0.8},{"id":"coach","label":"Coach (mt.em5.pro)","cluster":"products","realm":"both","description":"Multi-modality AI therapy platform. 8 modalities. Voice-native. PRIVATE to Michael + Tavish.","throughput":1,"health":1},{"id":"mtac-prod","label":"MTAC (em5.pro)","cluster":"products","realm":"both","description":"The command center. Currently DOWN.","status":"down","throughput":0,"health":0.05,"bottleneck":true},{"id":"macguardian","label":"MacGuardian","cluster":"products","realm":"both","description":"Mac protection app. v3.0.3 shipped. macguardian.app live.","throughput":3,"health":0.95},{"id":"physxq","label":"PhysXQ","cluster":"products","realm":"both","description":"Quantum Lab Interface at physxq.com. Sprints 1-5 shipped.","throughput":2,"health":0.9},{"id":"tactical-talk","label":"Tactical Talk","cluster":"products","realm":"both","description":"Voice intelligence product. Monorepo. Render hosted.","throughput":1,"health":0.7},{"id":"anvil","label":"Anvil (NeuroForge)","cluster":"products","realm":"both","description":"Personal Operations Partner. v1 Sprint 4 shipped 2026-05-11.","throughput":0.5,"health":0.8},{"id":"truvix","label":"Truvix","cluster":"products","realm":"both","description":"AI consumer product. VXC currency renamed to Truvix.","throughput":0.3,"health":0.6},{"id":"dignity-guard","label":"Dignity Guard","cluster":"products","realm":"both","description":"Privacy/dignity product. Provisional patent filed.","throughput":0.2,"health":0.5},{"id":"agentforge","label":"AgentForge","cluster":"products","realm":"both","description":"SaaS product. Separate from MTAC.","throughput":0.3,"health":0.4},{"id":"system-map","label":"System Map (map.em5.pro)","cluster":"products","realm":"both","description":"This map. Standalone project — kept separate from Coach so unrelated content stays out of Michael & Tavish's private space.","status":"new","throughput":0.5,"health":1},{"id":"mission-control","label":"Mission Control HQ","cluster":"products","realm":"both","description":"React view into MTAC. ~/Documents/MoosePack_HQ/mission-control.","throughput":0.5,"health":0.7},{"id":"vrdx-backend","label":"VRDX Backend","cluster":"products","realm":"both","description":"Express/TypeScript on Render. dist/server/bundle.js production build.","throughput":2,"health":0.85},{"id":"vrdx-frontend","label":"VRDX Frontend","cluster":"products","realm":"both","description":"React on Vercel.","throughput":1.5,"health":0.85},{"id":"vrdx-algorithms","label":"VRDX Algorithms","cluster":"products","realm":"both","description":"AI models on Render. Separate repo — algorithms NEVER go in Backend repo.","throughput":1,"health":0.7},{"id":"mpi","label":"Moose Pack Inc™","cluster":"corp","realm":"company","description":"Parent holding company. Owns IP, brand, and the AI workforce. All subsidiaries roll up here.","throughput":0,"health":1,"is_input":true},{"id":"vrdx-inc","label":"VRDX Inc","cluster":"corp","realm":"company","description":"AI / tech products subsidiary of MPI.","throughput":0,"health":1},{"id":"mpv-llc","label":"MPV LLC (AZ)","cluster":"corp","realm":"company","description":"Moose Pack Ventures LLC. Arizona. Ventures subsidiary. Formed 4/17/2024.","throughput":0,"health":1},{"id":"iesi-llc","label":"IESI LLC","cluster":"corp","realm":"company","description":"Integral Ethical Standards Institute. Research arm. ACEA (AI Collaborative Ethics Authority) as DBA.","throughput":0,"health":1},{"id":"tactical-talk-patent","label":"Tactical Talk provisional","cluster":"patents","realm":"company","description":"Provisional patent application filed. USPTO.","throughput":0,"health":1},{"id":"dignity-guard-patent","label":"Dignity Guard provisional","cluster":"patents","realm":"company","description":"Provisional patent application filed. USPTO.","throughput":0,"health":1},{"id":"truvix-ip","label":"Truvix IP Assignment","cluster":"patents","realm":"company","description":"IP assignment agreement. MPV → Truvix.","throughput":0,"health":1},{"id":"vrdx-license","label":"VRDX Exclusive License","cluster":"patents","realm":"company","description":"VRDX exclusive license agreement (filled).","throughput":0,"health":1},{"id":"brand-em5","label":"brand.em5.pro","cluster":"brand","realm":"company","description":"Centralized brand asset CDN. Hard rule: always reference assets via this URL, never local paths.","throughput":1,"health":1},{"id":"mpi-trademark","label":"MPI™ trademark","cluster":"brand","realm":"company","description":"Moose Pack Inc™ — common-law mark. Used as MPI Officially Licensed Product on all products.","throughput":0,"health":1}],"edges":[{"source":"claude-code-session","target":"global-claude-md","kind":"data","label":"loads at start","throughput":0.8,"realm":"network"},{"source":"claude-code-session","target":"auto-memory","kind":"data","label":"loads at start","throughput":0.8,"realm":"network"},{"source":"claude-code-session","target":"topic-files","kind":"data","label":"reads via index","throughput":0.4,"realm":"network"},{"source":"claude-code-session","target":"project-claude-md","kind":"data","label":"reads when in project","throughput":0.4,"realm":"network"},{"source":"claude-code-session","target":"project-log","kind":"data","label":"reads last 5 entries","throughput":0.4,"realm":"network"},{"source":"claude-code-session","target":"manual-outline","kind":"data","label":"reads §2.8","throughput":0.2,"realm":"network"},{"source":"sub-agent","target":"claude-code-session","kind":"control","label":"spawned by","throughput":0.2,"realm":"network"},{"source":"claude-code-session","target":"loom-cli","kind":"control","label":"sends/reads","throughput":0.5,"realm":"network"},{"source":"loom-cli","target":"loom-server","kind":"control","throughput":0.5,"realm":"network"},{"source":"loom-server","target":"loom-db","kind":"data","throughput":0.5,"realm":"network"},{"source":"loom-tick","target":"loom-db","kind":"data","label":"heartbeat","throughput":1.5,"realm":"network"},{"source":"loom-integrity","target":"loom-db","kind":"control","label":"integrity check","throughput":0.05,"realm":"network"},{"source":"loom-messages","target":"loom-db","kind":"data","throughput":0.5,"realm":"network"},{"source":"loom-tickets","target":"loom-db","kind":"data","throughput":0.1,"realm":"network"},{"source":"loom-ctime","target":"loom-db","kind":"data","throughput":0.05,"realm":"network"},{"source":"loom-db","target":"loom-hive-ingestor","kind":"data","label":"reads new messages","throughput":0.1,"realm":"network"},{"source":"loom-hive-ingestor","target":"hive-mind","kind":"data","label":"ingests","throughput":0.1,"realm":"network"},{"source":"loom-hive-ingestor","target":"training-pipeline","kind":"data","label":"INSIGHT → training","throughput":0.05,"realm":"network"},{"source":"loom-hive-ingestor","target":"loom-insights","kind":"data","label":"RESEARCH → SSD","throughput":0.03,"realm":"network"},{"source":"loom-hive-ingestor","target":"mtac-deferred","kind":"replay","label":"when MTAC down","throughput":0.3,"realm":"network"},{"source":"claude-code-session","target":"mtac-mcp","kind":"control","label":"496+ tools","throughput":0.5,"realm":"network"},{"source":"claude-code-session","target":"comms-hub","kind":"control","throughput":0.2,"realm":"network"},{"source":"council-room","target":"comms-hub","kind":"data","throughput":0.3,"realm":"network"},{"source":"work-team","target":"comms-hub","kind":"data","throughput":0.3,"realm":"network"},{"source":"mtac-mcp","target":"mtac-web","kind":"control","throughput":0.5,"realm":"network"},{"source":"comms-hub","target":"mtac-web","kind":"control","throughput":0.3,"realm":"network"},{"source":"hive-mind","target":"mtac-web","kind":"control","throughput":0.5,"realm":"network"},{"source":"knowledge-graph","target":"mtac-web","kind":"control","throughput":0.3,"realm":"network"},{"source":"training-pipeline","target":"mtac-web","kind":"control","throughput":0.3,"realm":"network"},{"source":"verdict-ledger","target":"mtac-web","kind":"control","throughput":0.2,"realm":"network"},{"source":"universal-intake","target":"mtac-web","kind":"control","throughput":0.4,"realm":"network"},{"source":"prism","target":"mtac-web","kind":"control","throughput":0.2,"realm":"network"},{"source":"clog-protocol","target":"hive-mind","kind":"data","throughput":0.2,"realm":"network"},{"source":"clog-protocol","target":"knowledge-graph","kind":"data","throughput":0.2,"realm":"network"},{"source":"clog-protocol","target":"training-pipeline","kind":"data","throughput":0.2,"realm":"network"},{"source":"clog-protocol","target":"universal-intake","kind":"data","throughput":0.2,"realm":"network"},{"source":"clog-protocol","target":"prism","kind":"data","throughput":0.2,"realm":"network"},{"source":"clog-protocol","target":"comms-hub","kind":"data","label":"session summary","throughput":0.15,"realm":"network"},{"source":"clog-protocol","target":"canonical-memory","kind":"data","label":"CHAT_LOG + MASTER_STATE","throughput":0.15,"realm":"network"},{"source":"clog-protocol","target":"activity-log","kind":"data","throughput":0.2,"realm":"network"},{"source":"clog-protocol","target":"decision-log","kind":"data","throughput":0.15,"realm":"network"},{"source":"clog-protocol","target":"lessons-learned","kind":"data","throughput":0.15,"realm":"network"},{"source":"clog-protocol","target":"book-material","kind":"data","throughput":0.05,"realm":"network"},{"source":"claude-code-session","target":"clog-protocol","kind":"control","label":"at session end","throughput":0.3,"realm":"network"},{"source":"claude-code-session","target":"brief-protocol","kind":"control","label":"at session start","throughput":0.3,"realm":"network"},{"source":"claude-code-session","target":"project-init","kind":"control","throughput":0.05,"realm":"network"},{"source":"claude-code-session","target":"verdict-log-protocol","kind":"control","throughput":0.1,"realm":"network"},{"source":"claude-code-session","target":"term-protocol","kind":"control","label":"Michael says \"term X\"","throughput":0.2,"realm":"network"},{"source":"claude-code-session","target":"sit-rep-protocol","kind":"control","throughput":0.15,"realm":"network"},{"source":"brief-protocol","target":"auto-memory","kind":"data","label":"urgent + warnings","throughput":0.3,"realm":"network"},{"source":"brief-protocol","target":"clog-deferred","kind":"data","label":"last 3","throughput":0.3,"realm":"network"},{"source":"brief-protocol","target":"loom-db","kind":"data","label":"inbox 24h","throughput":0.3,"realm":"network"},{"source":"brief-protocol","target":"project-log","kind":"data","label":"last 5 entries","throughput":0.3,"realm":"network"},{"source":"project-init","target":"project-claude-md","kind":"control","label":"creates","throughput":0.05,"realm":"network"},{"source":"project-init","target":"project-log","kind":"control","label":"creates","throughput":0.05,"realm":"network"},{"source":"project-init","target":"manual-outline","kind":"control","label":"creates","throughput":0.05,"realm":"network"},{"source":"verdict-log-protocol","target":"verdict-ledger","kind":"data","throughput":0.1,"realm":"network"},{"source":"verdict-callbacks","target":"universal-intake","kind":"data","label":"daily surface-due","throughput":0.1,"realm":"network"},{"source":"clog-protocol","target":"clog-deferred","kind":"replay","label":"when MTAC down","throughput":0.4,"realm":"network"},{"source":"verdict-log-protocol","target":"mtac-deferred","kind":"replay","label":"when MTAC down","throughput":0.1,"realm":"network"},{"source":"mtac-replay","target":"clog-deferred","kind":"data","label":"drains every 5min","throughput":0.3,"realm":"network"},{"source":"mtac-replay","target":"mtac-deferred","kind":"data","label":"drains every 5min","throughput":0.3,"realm":"network"},{"source":"mtac-replay","target":"mtac-web","kind":"data","label":"replays POSTs","throughput":0.3,"realm":"network"},{"source":"mtac-replay","target":"replayed","kind":"data","label":"archives","throughput":0.3,"realm":"network"},{"source":"mtac-replay","target":"failed","kind":"data","label":"failure stubs","throughput":0.05,"realm":"network"},{"source":"auto-ingest-watcher","target":"claude-code-session","kind":"data","label":"reads idle transcripts","throughput":0.1,"realm":"network"},{"source":"auto-ingest-watcher","target":"universal-intake","kind":"data","throughput":0.1,"realm":"network"},{"source":"mtac-prod","target":"model-router","kind":"control","label":"all LLM calls via router","throughput":2,"realm":"network"},{"source":"hybrid-router","target":"mtac-prod","kind":"control","label":"deprecated — legacy callers only","throughput":0.1,"realm":"network"},{"source":"model-router","target":"vps-ollama","kind":"data","label":"QUICK tier PRIMARY (llama3.2:3b)","throughput":2,"realm":"network"},{"source":"model-router","target":"claude-cli","kind":"data","label":"STANDARD/POWER PRIMARY (free subscription)","throughput":1.5,"realm":"network"},{"source":"model-router","target":"claude-max","kind":"fallback","label":"subscription fallback","throughput":0.5,"realm":"network"},{"source":"model-router","target":"openrouter","kind":"fallback","label":"backup tier (paid)","throughput":0.3,"realm":"both"},{"source":"model-router","target":"anthropic-api","kind":"fallback","label":"ULTRA tier (Opus, paid)","throughput":0.2,"realm":"both"},{"source":"model-router","target":"together-ai","kind":"fallback","label":"last-resort (402-out-of-credits)","throughput":0.01,"realm":"both"},{"source":"model-router","target":"mac-agent","kind":"control","label":"route to Mac resources","throughput":0.3,"realm":"network"},{"source":"mac-agent","target":"claude-cli","kind":"control","label":"invoke local CLI","throughput":0.5,"realm":"network"},{"source":"mac-agent","target":"mac-ollama","kind":"control","label":"local Mac Ollama (unused)","throughput":0.05,"realm":"network"},{"source":"claude-code-session","target":"claude-cli","kind":"control","label":"CLI itself","throughput":1.5,"realm":"network"},{"source":"coach","target":"openrouter","kind":"data","label":"direct (bypasses model_router!)","throughput":0.5,"realm":"both"},{"source":"coach","target":"elevenlabs","kind":"data","label":"TTS + Scribe STT","throughput":0.5,"realm":"both"},{"source":"coach","target":"vercel","kind":"data","label":"hosted on","throughput":0.3,"realm":"both"},{"source":"coach","target":"cloudflare","kind":"data","label":"DNS","throughput":0.3,"realm":"both"},{"source":"system-map","target":"vercel","kind":"data","label":"hosted on","throughput":0.2,"realm":"both"},{"source":"system-map","target":"cloudflare","kind":"data","label":"DNS","throughput":0.2,"realm":"both"},{"source":"mtac-prod","target":"cloudflare","kind":"data","throughput":0.2,"realm":"both"},{"source":"mtac-prod","target":"render","kind":"data","label":"hosting","throughput":0.2,"realm":"both"},{"source":"mtac-prod","target":"supabase","kind":"data","throughput":0.4,"realm":"both"},{"source":"tactical-talk","target":"render","kind":"data","throughput":0.3,"realm":"both"},{"source":"vrdx-backend","target":"render","kind":"data","label":"hosted on","throughput":0.3,"realm":"both"},{"source":"vrdx-frontend","target":"vercel","kind":"data","throughput":0.3,"realm":"both"},{"source":"vrdx-algorithms","target":"render","kind":"data","throughput":0.3,"realm":"both"},{"source":"vrdx-backend","target":"supabase","kind":"data","throughput":0.4,"realm":"both"},{"source":"mac-agent-bridge","target":"cloudflare","kind":"data","label":"tunnel","throughput":0.2,"realm":"network"},{"source":"mac-agent-bridge","target":"mtac-prod","kind":"data","label":"remote control","throughput":0.2,"realm":"network"},{"source":"paste-vault","target":"claude-code-session","kind":"data","label":"API keys","throughput":0.3,"realm":"network"},{"source":"thoughts-journal","target":"clog-protocol","kind":"data","label":"Section 5 embed","throughput":0.1,"realm":"network"},{"source":"lexicon","target":"clog-protocol","kind":"data","label":"Section 9 embed","throughput":0.1,"realm":"network"},{"source":"gmail","target":"claude-code-session","kind":"data","label":"inbox audits","throughput":0.4,"realm":"both"},{"source":"mpi","target":"vrdx-inc","kind":"control","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpi","target":"mpv-llc","kind":"control","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpi","target":"iesi-llc","kind":"control","label":"owns","throughput":0.05,"realm":"company"},{"source":"vrdx-inc","target":"vrdx-backend","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"vrdx-inc","target":"vrdx-frontend","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"vrdx-inc","target":"vrdx-algorithms","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"vrdx-inc","target":"truvix","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpv-llc","target":"macguardian","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpv-llc","target":"tactical-talk","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpv-llc","target":"dignity-guard","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpv-llc","target":"physxq","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpv-llc","target":"anvil","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpv-llc","target":"agentforge","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpi","target":"coach","kind":"data","label":"gift / private","throughput":0.05,"realm":"company"},{"source":"mpi","target":"system-map","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpi","target":"mtac-prod","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"mpi","target":"mission-control","kind":"data","label":"owns","throughput":0.05,"realm":"company"},{"source":"iesi-llc","target":"pattern-analysis","kind":"data","label":"research home","throughput":0.05,"realm":"company"},{"source":"tactical-talk","target":"tactical-talk-patent","kind":"data","label":"protected by","throughput":0.02,"realm":"company"},{"source":"dignity-guard","target":"dignity-guard-patent","kind":"data","label":"protected by","throughput":0.02,"realm":"company"},{"source":"truvix","target":"truvix-ip","kind":"data","label":"IP basis","throughput":0.02,"realm":"company"},{"source":"vrdx-inc","target":"vrdx-license","kind":"data","label":"license","throughput":0.02,"realm":"company"},{"source":"mpi","target":"mpi-trademark","kind":"data","throughput":0.02,"realm":"company"},{"source":"mpi","target":"brand-em5","kind":"data","label":"asset CDN","throughput":0.1,"realm":"company"},{"source":"coach","target":"brand-em5","kind":"data","label":"assets","throughput":0.1,"realm":"company"},{"source":"macguardian","target":"brand-em5","kind":"data","throughput":0.1,"realm":"company"},{"source":"physxq","target":"brand-em5","kind":"data","throughput":0.1,"realm":"company"},{"source":"tactical-talk","target":"brand-em5","kind":"data","throughput":0.1,"realm":"company"},{"source":"coach","target":"github","kind":"data","label":"source","throughput":0.1,"realm":"both"},{"source":"macguardian","target":"github","kind":"data","throughput":0.1,"realm":"both"},{"source":"mtac-prod","target":"github","kind":"data","throughput":0.1,"realm":"both"},{"source":"vrdx-backend","target":"github","kind":"data","throughput":0.1,"realm":"both"}],"stats":{"total_nodes":99,"total_edges":131,"total_clusters":15,"bottlenecks_count":5,"bottleneck_ids":["mtac-web","clog-deferred","mtac-deferred","supabase","mtac-prod"],"services_down":["mtac-web","mtac-mcp","comms-hub","hive-mind","knowledge-graph","training-pipeline","verdict-ledger","universal-intake","prism","verdict-callbacks","mtac-prod"],"queue_depth_total":109,"queued_at":[{"id":"mtac-web","label":"em5.pro (MTAC web)","queue_depth":53},{"id":"clog-deferred","label":"clog-deferred/","queue_depth":6},{"id":"mtac-deferred","label":"mtac-deferred/","queue_depth":50}],"avg_health":0.7805050505050507,"new_nodes_last_7d":["project-claude-md","project-log","manual-outline","mtac-replay","loom-hive-ingestor","loom-insights","brief-protocol","project-init","system-map"]}}