Skip to content
MCC Who?
IT EN
← Back

Dex Assistant

Federated AI chat — Anthropic / OpenAI / Ollama

In test
Python MCP federation Anthropic OpenAI Ollama sqlite-vec
Dex Assistant

Dex Assistant is the shared AI layer of the suite. It federates over MCP to the servers of the other 3 apps (Dex Hub, Dex Editor, Net Inventory Suite) and exposes a single chat where you can ask "how many jobs failed today?" or "what is the first free IP in 192.168.1.0/24?" — the assistant decides which app to forward the query to. Providers: Anthropic Claude, OpenAI, Ollama (local). SSE streaming, multi-modal vision (attach images to chat), SQLite-persisted conversations, cost tracking, RAG with sqlite-vec for querying internal documentation, reusable skill/prompt templates.

Features

  • MCP federation: talks to Dex Hub / Dex Editor / Net Inventory Suite
  • Providers: Anthropic Claude, OpenAI, Ollama (local)
  • Real-time SSE streaming
  • Multi-modal vision (image attachments)
  • RAG with sqlite-vec vector store
  • Cost tracking per provider and model
  • Reusable skill / prompt templates
  • SQLite-persisted conversations
  • Standalone UI + proxyable widget for other apps
  • Tool namespace separator for Anthropic and OpenAI

Configuration

  • Provider and model (e.g. claude-sonnet-4-6, gpt-4o, qwen2.5:14b)
  • API keys: ANTHROPIC_API_KEY, OPENAI_API_KEY
  • Ollama endpoint (e.g. http://localhost:11434)
  • MCP URLs of the 3 sister apps
  • Vector store path (sqlite-vec)
  • Per-token pricing override
  • System prompt + enabled skills

Technical prerequisites

  • Node.js 20+
  • At least ONE of: Anthropic API key, OpenAI API key, running Ollama
  • SQL DB (default SQLite)
  • Reachable Dex Hub / Dex Editor / Net Inventory Suite via MCP (for federation)