EVOMETAX

EvoMetaX — the long-term state engine

EvoMetaX is our long-term state AI agent foundation for life-cycle health — memory, multimodal modeling, agent workflows, risk tiering and explainable governance.

Technology, made for love
TECHNOLOGY FOUNDATION

A four-domain foundation

Organized like a body — brain thinks, cerebellum acts, mouth & ears speak and listen, memory remembers. Bounded, independent, orchestrated.

APPLICATIONS · SOLUTIONS
PerinatalMaternalYouthAdultSportsElderly
AI AGENT WORKFLOWS
MaternityFollow-upMental healthNutritionEldercareCare team
EVOMETAX CORE ENGINE
EvoMetaX — Long-term State Engine
Long-term MemoryMultimodal ModelingRisk TieringAgent OrchestrationExplainable AI
MULTIMODAL DATA
DialogueCheck-insReportsVitalsSleep/MoveNutritionMood/Scales
GOVERNANCE · END-TO-END
Reviewed KBRisk tieringExplainableIsolationPrivate deployFull audit
Brain

Thinking & Orchestration

The single source of truth for every conversation — managing persona, history and context. A self-built orchestrator performs intent recognition, parallel recall of knowledge/profile/memory, context assembly and tool routing; config-driven multi-persona, hot-reload and multi-model with no vendor lock-in.

Input
User message
Orchestrate
Intent
Recall
KnowledgeProfileMemory
Assemble
Persona + Context
Output
LLMTool routing
Cerebellum

Tools & Skills

A tool-execution layer on the open MCP (Model Context Protocol): retrieval, computation and system integrations are packaged as independent, hot-pluggable tool services — decoupled from the brain, focused on stable, observable execution, extensible per scenario.

From Brain
Tool call
MCP Gateway
MCP Server
Tools · hot-plug
SearchComputeExternal
Return
Result
Mouth & Ears

Real-time Voice

A self-hosted real-time voice platform: streaming ASR + streaming TTS (with voice cloning) + end-to-end voice dialogue. Low first-packet latency, barge-in, multi-turn memory and on-the-fly persona switching — Chinese-optimized, runs on a single GPU.

Audio in
Mic stream
Streaming ASR
Speech → Text
Brain
Dialogue
Streaming TTS
Text → Speech
Audio out
Speaker
Memory

Memory & Profile

A unified memory service: one API over three engines — conversation memory, structured user profiles and a knowledge graph — building long-term, cross-time state profiles and prompt-ready memory context, with multi-tenant isolation, audit and backup.

Unified API
Memory service
Engines
ConversationProfileKnowledge graph
Stores
VectorGraphSQL
Output
Memory context
CAPABILITIES

Six EvoMetaX modules

Long-term memory

History, current state, trends and personalized baselines.

Multimodal modeling

Fuses dialogue, check-ins, reports, vitals, sleep, exercise, nutrition and mood.

Agent workflows

Dedicated agents for maternity, follow-up, postpartum, mental health, nutrition, rehab and care.

Risk tiering

Education / follow-up / referral / crisis intervention.

Explainable governance

Reviewed knowledge, expert rules, cited advice, versioning and full audit.

Privacy, security & compliance

Permission & memory isolation, data masking, private deployment and full-chain audit to meet healthcare data compliance.

FAQ

FAQ

Can it be self-hosted and integrate with hospital systems?

Yes — private deployment with permission/memory isolation and data masking.

How is medical advice kept trustworthy?

Hospital-reviewed knowledge, expert rules, cited advice, human review and full audit.

Does AI replace doctors?

No — we assist experts and provide out-of-clinic infrastructure, not diagnosis or treatment.

Want a deeper look at EvoMetaX?

We’d be glad to give a technical walkthrough.