NvAssistant
NvAssistant is a full-stack AI assistant platform combining a real-time chat interface with a powerful multi-agent orchestration engine. From single questions to enterprise-scale workflows.
Real-Time AI Chat Interface
A rich, streaming chat experience that brings your AI agent to life with live reasoning, tool execution visibility, and file management.
SSE Streaming
Real-time Server-Sent Events stream reasoning, tool calls, skill activations, and results as they happen — enabling live "thinking" UIs.
File Upload & Persistence
Upload CSV, XLSX, PDF, DOCX, images and more. Files persist across questions in a session — no re-uploading needed.
Rich Tool Call Events
See exactly what your AI is doing — human-readable tool call descriptions with hint extraction and deduplication labels.
Image Generation
Generate images from text prompts using Amazon Titan via Bedrock. Auto-detected from natural language requests.
Generated File Downloads
Files created by the agent — reports, presentations, code — are automatically tracked and available for instant download.
Multi-Session Management
Isolated user sessions with full conversation history, persistent attachments, and stateless architecture for horizontal scaling.
Multi-Model AI Agent Framework
A production-grade agent engine that connects to any LLM, executes tools via MCP, and orchestrates complex multi-step workflows.
Multi-Model Support
Seamlessly route between Ollama local models and AWS Bedrock (Claude, Qwen, Meta) with automatic API format detection.
MCP Tool Integration
Connect to any MCP server via STDIO or SSE. Auto-discover and execute tools from filesystems, APIs, databases, and more.
Sub-Agent Orchestration
DAG-based orchestrator decomposes complex requests into parallel and sequential tasks across 13 specialized sub-agents.
Token Usage Tracking
Per-call token metrics with cost analysis across sessions and models. Budget-based model downgrade for cost optimization.
Error Resilience
Automatic retry on transient errors, XML tool-call parsing fallback, and intelligent result truncation for 45% token savings.
Agent Skills System
17 pre-loaded skills from Anthropic's repository. Dynamic activation based on context. Create custom skills with YAML + markdown.
Enterprise-Grade Infrastructure
Built for production with PostgreSQL persistence, Kubernetes deployment, and horizontal scaling. Every session, every tool call, every token — tracked and durable.
- PostgreSQL — 10 tables for full observability
- Kubernetes-ready with horizontal pod autoscaling
- Docker Compose for rapid local development
- Stateless agents — no sticky sessions required
- RESTful API with OpenAPI documentation
- Comprehensive logging with structured JSON events
Local Models
Claude / Qwen
Persistence
13 Agents. One Mission.
Each sub-agent is a domain expert — purpose-built for research, analysis, coding, compliance, and reporting.
Data Analyst
Researcher
Market Researcher
Competitive Analyst
Financial Modeler
Regulatory Researcher
Writer
Report Writer
Compliance Writer
Code Reviewer
Code Scanner
Remediation Coder
Evaluator
Ready to Build with
NvAssistant?
Deploy your own AI assistant platform with enterprise-grade orchestration, multi-model support, and production-ready infrastructure.