Autonomous Thinking
Autonomous Thinking (also called “Jarvis Mode”) allows your Snippbot agents to think independently in the background. Instead of waiting for your prompts, agents proactively observe your workspace, notice things, and contribute when they have something genuinely valuable to say.
How It Works
Section titled “How It Works”- Background daemon ticks every 5 seconds, checking each agent
- Gate evaluation asks a lightweight model: “Should I speak right now?”
- If the gate says speak, a thought is generated and injected into the conversation
- If the gate says silent, the agent stays quiet (this is the default)
- Engagement tracking learns from your responses and adjusts frequency automatically
Enabling Autonomous Thinking
Section titled “Enabling Autonomous Thinking”-
Open the agent’s detail page — click on your agent from the Agents list in the sidebar, then open its Settings tab
-
Toggle Autonomous Thinking on in the agent settings panel
-
Set the conversation to Human mode — click the thinking indicator in the chat header
Configuration Options
Section titled “Configuration Options”| Setting | Default | Description |
|---|---|---|
| Base interval | 120s | How often the agent considers thinking |
| Min interval | 30s | Fastest thinking rate (when you’re very active) |
| Max interval | 600s | Slowest rate (when you’re away) |
| Max thoughts/day | 20 | Daily cap to prevent over-talking |
| Quiet hours | None | Time window where thinking is paused (e.g., 22:00-07:00) |
| Triggers | All | Which events prompt thinking: periodic, user return, project changes, time-based |
Adaptive Intervals
Section titled “Adaptive Intervals”The thinking interval automatically adjusts based on your activity:
- Active (message in last 5 min): Checks every ~60 seconds
- Recent (5-60 min ago): Checks at the base interval
- AFK (1+ hours): Checks at the max interval (saves ideas for your return)
When you return after being away for 1+ hours, the agent is triggered with a “user return” event and will share something substantive.
Engagement Multiplier
Section titled “Engagement Multiplier”The system tracks how you interact with autonomous thoughts:
| Your action | Score | Effect |
|---|---|---|
| Reply within 2 min | +3 | Agent thinks more often |
| Reply within 10 min | +2 | Normal engagement |
| Reply after 10 min | +1 | Slight positive |
| Continue conversation | 0 | Neutral |
| Dismiss thought | -2 | Agent thinks less often |
| Ignore for 10+ min | -3 | Agent significantly reduces frequency |
The multiplier adjusts between 0.5x (very engaged) and 2.0x (disengaged), applied to the base interval.
Human vs Bot Mode
Section titled “Human vs Bot Mode”Each conversation can be in one of two modes:
- Human mode: Autonomous thinking is active. The agent can inject thoughts.
- Bot mode: Standard chat. The agent only responds to your messages.
Toggle the mode using the thinking indicator in the chat header.
