CircuVolt AI
raj713335/CircuVolt_AI
Previsão de saúde de baterias EV com passaportes digitais e rotas circulares otimizadas
Ver no GitHubPlataforma agentic que prevê a saúde de baterias EV, atribui classificações de segunda vida/reciclagem e gera passaportes de materiais com QR-code. Integra LangGraph, MCP e A2A Protocol para orquestrar pipelines de economia circular com recomendações personalizadas e seguras.
Instalação
npx convez-ai raj713335/CircuVolt_AI1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Previsão de State of Health (SoH) em baterias de veículos elétricos
- Geração de passaportes de materiais conforme regulações EU (2023/1542)
- Recomendação de rotas de segunda vida e reciclagem otimizadas por valor
Veredito

Solução robusta e multi-protocol para circularidade de baterias EV, com arquitetura bem-estruturada (FastAPI + React + LangGraph). Especializada demais para uso geral, mas referência sólida para integrações de IA agentic em domínios de sustentabilidade industrial.
README
A tradução para o português deste README ainda está na fila.
CircuVolt AI
Battery Second-Life + Material Passport Intelligence Platform
An agentic AI platform that predicts EV battery State of Health, assigns a second-life/recycling grade, generates a QR-based material passport, and recommends the safest, highest-value circular-economy pathway built with LangGraph, MCP, A2A Protocol, AG-UI Protocol, and CopilotKit.
Architecture Diagram
Layer Breakdown
-
Frontend Layer (React + Vite)
- Uses
React Three Fiberfor high-performance 3D rendering (explodable battery models). - Uses
Rechartsfor advanced analytics (MathWorks-style degradation curves and radar charts). - Consumes Server-Sent Events (SSE) to stream LLM responses in real-time.
- Uses
-
Backend Layer (FastAPI)
- SOH Inference Engine: Computes battery degradation matrices, internal resistance drops, and generates synthetic forecast data.
- Recyclability Rules Engine: Evaluates JSON payloads of component properties against hardcoded EU battery regulations and material science heuristics.
- LLM Orchestrator: Takes the deterministic outputs from the Rules Engine and passes them as context to a Large Language Model to generate human-readable, actionable redesign recommendations via an SSE stream.
-
External Integrations
- LLM API: Powers the deep generative insights.
- Material Database: Stores static lookup tables for material lifecycle carbon footprints, recovery rates, and market values.
Protocol Integration
MCP (Model Context Protocol)
All circularity intelligence tools are exposed as MCP-compatible tools via mcp_server/server.py. Any MCP client can discover and invoke:
predict_battery_soh: ML-powered SOH predictiongrade_battery: Second-life grading with safety overridesgenerate_material_passport: EU Reg 2023/1542 compliant passportplan_recovery: Disassembly sequence optimizationcalculate_circularity: Circularity score calculationanalyze_recyclability: Design improvement suggestions
MCP Resources provide context: grading rules, material composition data.
A2A (Agent-to-Agent Protocol)
Implements Google's A2A specification for inter-agent communication:
- Agent Card at
/.well-known/agent.json: discovery endpoint with skills - Task Send at
/a2a/tasks/send: synchronous task processing - Task Subscribe at
/a2a/tasks/sendSubscribe: streaming SSE responses - Task Status at
/a2a/tasks/{id}: check task progress
Other AI agents can discover and interact with CircuVolt AI agents autonomously.
AG-UI Protocol + CopilotKit
The AG-UI (Agent-User Interaction) protocol streams agent state to the frontend:
- TEXT_MESSAGE events: streaming text responses
- TOOL_CALL events: real-time tool invocation visibility
- STATE_SNAPSHOT/DELTA : agent state synchronization
- RUN lifecycle events: start/finish tracking
CopilotKit implements AG-UI natively and provides:
CopilotSidebar: conversational AI assistant paneluseCopilotAction: frontend actions invokable by the agentuseCopilotReadable: state shared with the agent
LangGraph + LangChain
- LangGraph ReAct Agent : multi-step reasoning with tool use
- LangChain tools :
@tooldecorated functions wrapping ML models - Streaming : real-time token streaming via CopilotKit runtime
- Fallback : works without OpenAI API key using direct tool execution
Key Features
- Battery SOH Prediction : XGBoost/GradientBoosting ML model with SHAP explainability
- Second-Life Grading : A/B/C/D grades with safety override rules
- Material Passport : EU Regulation 2023/1542 compliant with QR code
- Disassembly Optimizer : Step-by-step recovery with carbon/economic analysis
- Circularity Scoring : Multi-factor score (0-100)
- Design Recyclability : Improvement suggestions with scoring
- AI Copilot : Open-ended conversational interface via CopilotKit
- Dynamic Agent Builder : Create custom agents through the UI with Auto-Verification for external MCP servers and A2A endpoints.
- Multi-Agent Interoperability : Native A2A protocol support where each custom agent publishes its own
/.well-known/agent.jsonand proxies tasks to specialized sub-agents. - Interactive Vehicle Studio : Interactive 3D component exploration and advanced, fully dynamic vehicle-to-vehicle circularity comparisons.
How to Run
Prerequisites
- Python 3.10+
- Node.js 18+
- (Optional) OpenAI API key for LLM-powered agent
Backend
git clone https://github.com/raj713335/CircuVolt_AI
cd CircuVolt_AI
pip install -r requirements.txt
# Optional: set OpenAI key for full agent capabilities
cp .env.example .env
# Edit .env and add your OPENAI_API_KEY
python main.py
- API:
http://localhost:8000 - Docs:
http://localhost:8000/docs - Agent Card:
http://localhost:8000/.well-known/agent.json
MCP Server (standalone)
cd CircuVolt_AI
python mcp_server/server.py
Frontend
git clone https://github.com/raj713335/CircuVolt_AI_UI
cd CircuVolt_AI_UI
npm install
npm run dev
- App:
http://localhost:5173
3. Application Screenshots
Tech Stack
| Layer | Technology |
|---|---|
| Agent Framework | LangGraph (ReAct pattern) |
| LLM Orchestration | LangChain + LangChain-OpenAI |
| Tool Protocol | MCP (Model Context Protocol) |
| Agent Communication | A2A (Google Agent-to-Agent Protocol) |
| UI Protocol | AG-UI (via CopilotKit) |
| Frontend | React, Vite, Tailwind CSS, CopilotKit, Recharts |
| Backend | FastAPI, Python |
| ML | Scikit-learn, GradientBoosting, NumPy, Pandas |
| Database | SQLite (SQLAlchemy ORM) |
| QR Generation | Python qrcode library |
Adoption roadmap for AI-enabled battery circularity in automotive
---
config:
theme: redux-color
---
timeline
title Adoption roadmap for AI-enabled battery circularity in automotive
2026 : Standardize battery genealogy, pack IDs, and service-history capture
: Build grading lakehouse from BMS, CAN, telematics, lab, and repair data
: Pilot AI-assisted SOH grading on retirements and warranty returns
2027 : Expose BMS SoH and expected-lifetime data to authorized users
: Deploy battery-passport-ready data models and interoperability layer
: Launch residual-value and routing workflows for reuse, repair, and recycle
2028 : Scale transfer-learning SOH models across mixed chemistries and OEM programs
: Deploy robotic assistance for screws, covers, connectors, and hazard-prone steps
: Add chemistry-aware intake screening for direct-recycling feedstock
2029 : Connect grading, marketplace, BESS repurposing, and recycler routing in one control plane
: Close loop between recovered-material quality and upstream grading decisions
2030 and beyond : Use lifecycle data to influence pack design-for-disassembly and design-for-recycling
: Run closed-loop optimization across first life, second life, and recovery