The starter kit in normal words.
The default suite is telegram-starter. It gives Spark a chat door, runtime routing, memory, Mission Control, and research loops.
The installer only includes modules that expose Spark's installable module contract. Public labs are available for builders, but not every lab is part of the automatic starter bundle yet.
How the pieces talk.
Telegram is where the user talks. Spark Intelligence Builder routes the work. The LLM thinks. Memory remembers. Mission Control handles bigger tasks.
The modules.
Spark Telegram Bot
Receives commands and normal messages.
Uses long polling, not webhooks.
Spark Intelligence Builder
Routes Agent and Mission roles.
Connects providers, memory, and tools.
Domain Chip Memory
Saves useful facts when memory is wired.
Recalls context when Spark needs it.
Spawner UI
Mission Control for bigger work.
Shows boards and tracked execution.
Spark Researcher
Runs evidence and advisory workflows.
Supports bounded improvement loops.
Spark Character
Keeps role and voice stable.
Tracks permission behavior.
Spark Live
Runs Spark services quietly.
Can autostart after login.
Public builder labs.
Spark Domain Chip Labs
Domain chips, benchmark packs, specialization paths, autoloop policies, tool integrations, and publish packet standards.
Spark Personality Chip Labs
Modular personality chips and profile experiments that connect to Spark's character layer.
Spark Voice Comms
Voice communication hooks, voice profiles, and local or paid provider examples.
Network boundary
Spark Swarm network submission and hosted workspace features are upcoming. They are not required for local Spark experiments and are not public installer dependencies yet.
What users should try first.
Prove chat
Send /start, /myid, and a normal message in Telegram.
Prove memory
Send /remember launch test memory, then /recall launch test memory.
Prove missions
Send /run summarize the current Spark setup, then /board.
Keep improving
Add provider-specific examples from real onboarding runs as they land. Until then, keep provider docs conservative and use spark providers status as the source of truth.