Complete a prompt
llm://host/chat/command/complete
{
"model": "qwen2.5",
"prompt": "Summarise: ...",
"maxTokens": 256
}
AI connector
Local LLM
Local LLM bridges a self-hosted model runtime (Ollama, llama.cpp or vLLM) into llm:// routes, so flows and agents can request completions and list models with the same URI contract — and project them to MCP tools / A2A skills, keeping inference on-device.
What it does
- Summarise a flow step's output with a local model.
- List installed models from an agent.
- Keep inference on the LAN with no external API.
URI schemes
llm://
Requirements
a local LLM runtime (Ollama / llama.cpp / vLLM)
Routes
Examples
Install command
curl -fsSL 'https://connect.ifuri.com/install?connectors=llm' | bash
This connector is planned. The installer is visible for contract design, but execution is disabled until the connector package is available.
Registry entry
{
"id": "llm",
"name": "Local LLM",
"status": "planned",
"uriSchemes": [
"llm"
],
"routes": [
"llm://host/chat/command/complete",
"llm://host/model/query/list"
],
"install": {
"mode": "planned",
"pipSpec": "urirun-connector-llm"
},
"hubUrl": "https://connect.ifuri.com/connectors/llm"
}