Agent List

KULVEX ships with 17 domain agents, each handling a specific category of requests.

Priority	Agent	Domain	Examples
10	`url_analyze`	URL analysis	”analyze this link”, paste a URL
15	`signal`	Signal messaging	”send a Signal message to…“
20	`overview`	System overview	”status”, “overview”, “dashboard”
25	`fashion`	Fashion/style	”outfit suggestions”, “what to wear”
30	`image`	Image generation	”generate an image of…”, “create a picture”
35	`home`	Home automation	”turn on lights”, “set temperature”
40	`weather`	Weather	”what’s the weather”, “forecast”
45	`presence`	Who’s home	”who’s home”, “is anyone here”
50	`solar`	Solar energy	”solar production”, “energy today”
55	`security`	Security system	”arm alarm”, “camera status”
60	`remote_node`	Remote nodes	”GPU status on node 32”
65	`code`	Code tasks	”write a function”, “explain this code”
70	`system`	System info	”disk usage”, “memory”, “uptime”
75	`search`	Web search	”search for…”, “find information about”
80	`backup`	Backups	”backup status”, “run backup”
85	`twitter`	Twitter/X	”post a tweet”, “check mentions”
90	`message`	General messaging	”send a message via Telegram”

Agent Anatomy

Each agent is a Python class in core/agents/domain/ with:

class WeatherAgent(BaseAgent):
    name = "weather"
    priority = 40
    patterns = [r"weather", r"forecast", r"temperature outside"]
    tools = ["get_weather", "get_forecast"]
 
    async def handle(self, message, context):
        # Simple query → template response
        if self._is_simple(message):
            return self._template(await self._get_weather())
 
        # Complex query → tool execution + LLM synthesis
        data = await self._execute_tools(message)
        return await self._synthesize(message, data)

Template vs Synthesis

Template responses — Pre-formatted, zero LLM. Used for “what’s the weather?” type queries. Instant.
Synthesis — Runs data through a minimal LLM prompt (~200 chars) for natural language. Used for complex queries. ~1-2s.

Overview Agentic Chat