TERRANOETIS

AI chat & cognition

Platform v3.0.0 Docs v3.0.0 MEASURED — cognitive services start observed at boot; LLM outcomes provider-dependent

Full description: Read the full narrative

The chat pipeline classifies a natural-language query into one of eight intents, routes it through a dual-process cognition layer (fast cached/templated answers vs deliberate multi-agent reasoning), executes tools (live data, analytical models, simulations, sandbox), and streams results over SSE. All LLM traffic passes through the Omninet provider router.

Intents & routing thresholds

Decision pointRuleSource
Intent classesquick_scan · deep_analysis · fly_to · toggle_layer · weather_check · compute · panel_command · unknownserver/agent.ts:48-49
System 1 fast pathsimilarity ≥ 0.92 → real tool for matched intent (6 mapped intents)server/cognition/cognitiveOrchestrator.ts:251,47-72
Verify band0.70 ≤ similarity < 0.92 → S1 answer + System 2 verification, 10 s timeoutserver/cognition/cognitiveOrchestrator.ts:252,295
Full System 2no match / anomaly / < 0.70 → HTN decomposition → 4-agent debate (Data Analyst, Domain Expert, Systems Thinker, Risk Assessor) + Verifier critic, 45 s hard cap, critic ≥ 0.7 to exitserver/cognition/cognitiveOrchestrator.ts:286,386-388 server/cognition/system2.ts:78-85,446-467
LLM deep classificationonly when keyword intent conf < 0.7; LLM verdict accepted at conf ≥ 0.4server/index.ts:9846-9851 server/agent.ts:795
Tool loopMAX_TOOL_ROUNDS = 3 with anti-prompt-injection synthesis promptserver/index.ts:10966-10983
Stream caphard 150 s run cap; 15 s heartbeatsserver/index.ts:9801-9811

Omninet model router

PropertyVerified valueSource
Providers9: groq, gemini, bai, deepseek, claude, ollama, openrouter, huggingface, local-gguf (20 models, tiers 1–4)server/ai-router/omninet.ts:70-84
Complexity → tiersimple→1, medium→2, complex/reasoning→3server/ai-router/omninet.ts:103-121,300-307
Rankinghealth → |tier−target| → tier → lastLatency; skips circuit-open / keyless / token-starvedserver/ai-router/omninet.ts:331-358
Resilienceper-provider token buckets; circuit opens at 5 failures, half-open 30 s; retries backoff [1s,2s,4s] ≤ 2 extra; health ping every 2 min (429/402/403 count as alive)server/ai-router/omninet.ts:281-296,563-568,337,585-589,225-250,270
Local fallbackLFM2.5-2.6B-Q4_K_M GGUF via llama-server :11436, spawned at boot when file present; admin-panel downloader with resumeserver/index.ts:789-817 server/apiMetadata.ts:gguf endpoints
Embeddingslocal Xenova all-MiniLM-L6-v2 first; API fallbacks (Gemini/OAI/Ollama), 768-dim defaultserver/ai-router/omninet.ts:675-713,648
Cost tiersModelRouter local/flash/pro + price table; /api/agent/tiers costPerQuery 0 / 0.0001 / 0.0008server/costOptimizer.ts:18-22,47-54 server/index.ts:11686-11695

Advanced agent safety rails

ControlDetailSource
Risk classes17 DESTRUCTIVE + 15 HIGH_RISK tool lists; arg regex promotes risk (delete|remove|reset|…|deploy|publish)server/advancedAgent.ts:212-234
Approval gateactions at threshold “high” require user approval before executionserver/advancedAgent.ts:240-244
Plan budgetLLM plan 1–5 steps (max 6), temp 0.3; sub-agents temp 0.4server/advancedAgent.ts:252-274,370
MCTSexploration constant 1.41, ≤100 iterations, depth 5, 30 s timeout (execution orchestrator over tools)server/cognition/mctsEngine.ts:11-16,87-89
Human overrideexplainability subsystem: pending override queue, evidence chains, bias auditsserver/index.ts:explainability routes :917-918 + module
Verification status

Startup of the cognitive services (orchestrator, omninet init, self-improver V2, feedback) was observed in server boot logs server/index.ts:787,834-845; thresholds and lists are traced to source as above. End-to-end LLM outcomes are provider/credential dependent and were not executed here.